Public Version

Home , Girls Chase Boys, SCI Fidelity Records, Sirius XM Indie

Electronically Filed Docket: 19-CRB-0005-WR (2021-2025) Filing Date: 09/26/2019 10:44:23 PM EDT

Before the UNITED STATES COPYRIGHT ROYALTY JUDGES Washington, D.C.

) In the Matter of ) ) DETERMINATION OF RATES AND TERMS ) Docket No. 2011-1 FOR PREEXISTING SUBSCRIPTION AND ) CRB PSS/Satellite II SATELLITE DIGITAL AUDIO RADIO ) SERVICES )

WRITTEN DIRECT TESTIMONY OF STEVEN BLATTER

(On behalf of Sirius XM Radio Inc.)

I. Introduction

1. I am Senior Vice President and General Manager of Music Programming at Sirius

XM Radio Inc. (“Sirius XM”). I have worked professionally in radio, as well as in marketing,

promotion and online for over 25 years. In terrestrial radio, I was responsible for the

programming of local stations in New York and Los Angeles. My prior experience also includes

the creation and oversight of content syndicated to hundreds of local radio stations across North

America. Prior to the merger of Sirius Satellite Radio Inc. (“Sirius”) and XM Satellite Radio

Holdings, Inc. (“XM”), I was at Sirius for approximately five years.

2. I testified on behalf of Sirius in the prior proceeding before the Copyright Royalty

Board (“CRB”) to set rates for the period from 2007 through 2012 (“Satellite I”). In my prior

written testimony, which is incorporated herein (and which has been designated for inclusion in

this proceeding according to the governing regulations), I covered three key areas: (1) the range

and coverage of the Sirius music channels; (2) how each music channel is developed and

programmed and Sirius’ creative contribution to that process; and (3) the promotional benefits of

radio, including satellite radio. To the extent that there have been any material changes since the last proceeding before the CRB, I will update each of these categories in my testimony here. In addition, I will provide testimony describing how airplay on Sirius XM channels has a direct impact on the sale of recorded music. The correlation between Sirius XM airplay and music sales – which has been repeatedly acknowledged to me and my staff by record company executives, artists, and their managers – is substantially greater than it was when I testified in the

Satellite I proceeding five years ago.

II. Description Of Sirius XM’s Music Programming And Current Music Channels

A. Sirius XM’s Services And Music Programming Structure

3. As is described in more detail in the accompanying testimony of James Meyer, even after the merger of Sirius and XM, Sirius XM maintains two separate distribution platforms. For the most part, the content offerings on the two services, including the music channels, are almost identical. Although there remain “Sirius” and “XM” platforms, by illustration, a subscriber listening to the 80s on 8 channel on the Sirius platform hears the same channel as a subscriber listening to the 80s on 8 channel on the XM platform.

4. I am responsible for all of Sirius XM’s music programming operations and I have a significant role in other areas related to music programming, such as artist and talent relations.

I currently supervise approximately 285 full and part-time employees, including six Music

Programming heads, two Programming Operations supervisors, and one head of Production.

Each Music Programming head is responsible for a genre or two of music, including Jazz,

Standards, Latin, Pop, Rock, Urban/Hip Hop, and Country, and supervises the programming personnel responsible for Sirius XM’s various music channels within each genre.

5. The Sirius satellite platform offers 71 full-time music stations without commercials (64 are produced by Sirius XM in the United States; the other seven are produced

2 by our Canadian affiliate).1 The XM satellite platform offers 71 full-time music stations without commercials (67 are produced by Sirius XM in the United States; the other four are produced by our Canadian affiliate).2 The most current versions of the channel line-ups are attached as Sirius

XM Direct Exhibit (“SXM Dir. Ex.”) 1. These line-ups do not include limited-engagement channels, which are offered for shorter periods of time and which I discuss below in paragraphs

56-58.

6. As can be seen on the channel line-ups, most of the music stations continue to be era- or genre-based. For example, our thirteen pop stations (fourteen on XM) include Sirius XM

Hits 1 (Top 40 hits), ’40s on 4, ’50s on 5, ’60s on 6, and so on. Our 22 rock channels include a wide mix of classic rock, alternative and indie rock, hard rock and heavy metal, and several more mellow, “adult” rock formats. Each of the Hip-Hop/R&B, Dance, Country, Christian,

Jazz/Standards, and Classical categories likewise offers between three and seven stations covering several listening options – including a variety of niche channels (like Bluegrass, Reggae and Show Tunes) that cannot be found on terrestrial radio.

7. As I described in my Satellite I testimony, a number of our music channels are devoted almost exclusively to music recorded before 1972 (and thus, as I understand, not subject to the statutory license at issue in this proceeding). These include ’40s on 4, ’50s on 5, ’60s on 6, and Elvis Radio. Other music stations on our services make substantial (though not exclusive) use of pre-1972 recordings, including Classic Vinyl, the Grateful Dead Channel, Underground

1 Sirius XM recently introduced sixteen additional full-time music stations that are available to subscribers that purchase the latest Sirius XM radio and one of our new “SiriusXM” branded packages which come with a suite of 20+ “Xtra” channels. The required radio includes the new Sirius XM Edge and Lynx products. Over time, more and more of our subscribers will be able to access these stations. 2 There are slight differences between the full-time music channels of XM and Sirius. For example, there are a few music channels offered on XM that are not available through Sirius because the XM platform is capable of hosting a few more channels than the Sirius platform, and there are a few Clear Channel- programmed radio stations available on XM that are not available on Sirius.

3 Garage, Siriusly Sinatra, Radio Margaritaville, On Broadway, The Bridge, ’70s on 7, and Soul

Town.

B. Development Of Channel Style And Identity

8. Sirius XM goes to great lengths to create an identity for each of its music channels. In many cases, our distinctive channel identities have been crafted over the course of eight or ten years. During this time, we have strategically built true national “brands” for each of these channels, as follows:

While the era or genre will define the universe of music upon which the programmers will draw, the channel style and personality will shape the overall energy level of the channel, for example, whether it is edgy and energetic or conservative and mellow.

The channel identity also helps determine the appropriate on-air hosts for the channel – and is in turn shaped by the personalities of those hosts.

Each channel has a concise name that captures its format and identity such as “Octane,” (uptempo hard rock) or “BPM” (dance). Moreover, a distinctive channel logo is associated with the channel name so as to permit instant visual identification in our marketing materials and on receivers. Further, the name and logo are supported by additional positioning or attitudinal statements that are used to reinforce the brand image of the channel. For example, the slogan of our flagship Top 40 Station, Sirius XM Hits 1, is “The Sound of Generation Now.”

Sirius XM music channels also typically include production elements or interstitial audio pieces heard between the songs that further develop the image and personality of the channel for the intended audience. Each music channel has its own station voice and slogan, and some even employ custom jingles that help enhance the mood and flow of the channel.

Interactive contests and promotions are aired on the appropriate channels to further build listener loyalty and create an even stronger bond with the audience.

The identity of a channel guides the types of special programming that may be created and scheduled. Artists often participate in interviews and perform live on certain rock channels, while on our pop channels artists will guest-host special programs like countdown shows and new music showcases.

9. All of these factors – which go beyond simply playing music – work together to enhance the personality of each station, make the listening experience more engaging to the

4 target audience, and create meaningful and satisfactory channel options for the subscriber. As a result of Sirius XM’s success in building these national channel brands, our listeners develop valuable relationships with their favorite Sirius XM channels.

C. The Music Selection Process For Sirius XM Music Channels

10. Our stations are far from mere “jukeboxes” of songs from a given era or genre.

Rather, each channel has a clearly defined style and identity that shapes the programming and presentation of the channel in multiple ways. As I mentioned in my Satellite I testimony (at paragraph 25 through 30), the creation and maintenance of a particular music channel requires a great deal of planning and resources. Selecting, sequencing and packaging music is a difficult and ongoing task carried out by our music programming specialists. It is not enough just to rotate down an alphabetical or random list of pieces that are within the universe defined by a channel’s format. Instead, trained music programming experts who are deeply familiar with the music bring to bear both scientific and artistic judgment to curate an optimal musical flow and mood on a song-by-song basis. Each Music Programmer also utilizes software called

MusicMaster that helps in maximizing the rotation and sequencing of songs for each music channel. While this software assists programmers in managing their music libraries and facilitates the music scheduling process, it is no substitute for the informed judgment of programmers with in-depth knowledge of the genre. Every hour of music scheduled across all

Sirius XM music channels is carefully reviewed and specially selected by a Sirius XM programmer before it is then presented to our listeners.

D. Sirius XM’s Direct Licensing Initiative

11. One additional factor that will play into our music selection going forward is

Sirius XM’s recent direct licensing initiative. As discussed, the music selected and played across

5 the Sirius XM platform is curated by our programmers using their professional judgment, as well as proprietary and non-proprietary research data as it applies to a particular channel. Where appropriate, the programmer also takes into account the relationship that the artist and/or their representatives have with Sirius XM. This is sometimes measured by the artist’s willingness to participate in our programming with specials like live performances, interviews, guest DJs sessions and other unique opportunities. Going forward, we also will take into account whether the artist’s record label has a direct licensing relationship with Sirius XM.

12. Sirius XM has recently entered into a number of direct licenses with more than 60 record companies. I have reviewed the list of companies, and can attest that their catalogs include songs that are already featured on Sirius XM music channels. For example:

The band LA Guns, whose music is now directly licensed via our agreement with Cleopatra Records, is a mainstay of our Hair Nation (hard rock) channel – we feature thirteen tracks by the band in rotation. Many other classic 80s hair metal tracks regularly played on Hair Nation, including “Turn Up the Radio” by Autograph and “Seventeen” by Winger, are likewise licensed by Cleopatra, whose Deadline Records imprint specializes in the genre.

The band Local Natives has two songs – “Wide Eyes” and “Airplanes” – that are now covered by Sirius XM’s direct license with Frenchkiss Records. These songs have been playing on Sirius XMU and/or Alt Nation for over a year.

Sirius XM’s direct license with Sci Fidelity Records covers more than twenty songs by the artist String Cheese Incident. These songs are actively being played on Sirius XM’s Jam On channel.

Our direct license with PS Classic covers many soundtracks, including Sondheim on Sondheim, A Little Night Music, and Finian’s Rainbow. Currently, more than fifteen songs covered by our license with PS Classic are in active rotation on Sirius XM channels.

13. Our primary objective is to provide our subscribers with the most satisfactory listening experience possible. While the merits of the artist and song always come first, we will favor a directly licensed song when the programmer feels that it will fit just as well as other non- directly licensed tracks.

6 E. Sirius XM’s Top Quality On-Air Talent

14. Sirius XM makes major investments in identifying and developing top quality on- air talent, and our programmers and other creative workers regularly support that talent.3 Sirius

XM employs over 200 credible and authentic on-air personalities to present music and timely information about each artist and song played. The relationships that Sirius XM’s hosts develop with their audience are strong and valuable, as the listeners come to trust the on-air hosts who

“sell” the music they play in a passionate and engaging manner.

15. Sirius XM’s 200-plus music programming hosts include well-known musicians like Tom Petty, Bob Dylan, Lou Reed, Nancy Sinatra, Peter Frampton and Steven Van Zandt. In the hip-hop genre, our hosts include DJ Sway, the Aphiliates, DJ Kay Slay, Tony Touch, and

Whoo Kid – who might not be household names like Bob Dylan, but are well-known by hip-hop fans. Sirius XM is particularly influential in the electronic dance genre, where the world’s most popular DJs host regular shows on Sirius XM’s electronic dance channels such as Electric Area and BPM. These DJs include Avicii, David Guetta, Skrillex, Laidback Luke, Tiesto, Armin Van

Buuren, Afrojack, Kaskade, Bob Sinclar, Above and Beyond, Steve Aoki, Paul Oakenfold, Ferry

Corsten, Josh Wink, Eddie Halliwell, Markus Schulz, Sander Van Doorn, Carl Cox, John

Digweed, Robbie Rivera, Benny Benassi, George Acosta, Adam Beyer, Dave Aude, Crystal

Method, Pretty Lights, and Nervo. SXM Dir. Ex. 18 contains a full list of celebrity hosts on

Sirius XM music channels.

16. Sirius XM listeners recognize that the hosts and programmers behind the stations

– people whose taste the listeners have come to know, appreciate, and rely on – have hand-

3 In contrast, the largest radio operator in the United States – Clear Channel Communications – reportedly just cut a significant number of DJs from its payroll. See Brian Stelter, Clear Channel Cuts DJs Across the Country, N.Y. Times, Oct. 27, 2011 (attached as SXM Dir. Ex. 17).

7 selected the music they hear. Curation causes our listeners to take notice of and be open to the songs we select.

F. Sirius XM’s Unique Artist Channels And Specialty Programming

17. We also have developed channels and programs in most of the main music categories in conjunction with well-known artists such as Jimmy Buffett (“Radio

Margaritaville”), Eminem (“Shade 45”), and Steven Van Zandt (“Underground Garage”). Since

I last testified in Satellite I, we have developed additional channels with several artists, including

Pearl Jam (“Pearl Jam Radio”). Most recently, Sirius XM announced that Ozzy Osbourne will launch his own music channel to be titled “Ozzy’s Boneyard.” SXM Dir. Ex. 19. We also continue to offer channels devoted to the music of Elvis Presley (“Elvis Radio”), Frank Sinatra

(“Siriusly Sinatra”), Bruce Springsteen (“E Street Radio”), and the Metropolitan Opera (“Met

Opera Radio”), among others.

18. The channels that we have developed with these artists involve exclusive relationships that often entail not just lending their name to a channel, but significant contributions by the artists themselves. For example, Willie Nelson is deeply involved in

Willie’s Roadhouse, and has at times asked to review and approve every artist whose music is played on his channel. Similarly, Jimmy Buffet also has significant involvement with his channel, Margaritaville, and Eminem’s manager, Paul Rosenberg, has daily interactions with

Sirius XM regarding the programming of Eminem’s Shade 45 channel. With respect to Elvis

Radio, the Presley estate is in regular contact with Sirius XM and our dedicated channel programmer, who is located at the Elvis headquarters in Memphis, Tennessee.

8 19. In addition to bringing their creative talents to bear on shaping the overall listening experience, these artists often make available a range of unreleased recordings that listeners otherwise could not access.

20. We also program a variety of specialty shows that play music that listeners would not typically hear elsewhere. For example, on Sirius XM’s channel The Spectrum, we have a show called “Nordic Rox” that plays music by Scandinavian artists, “Celtic Crush” plays all

Celtic music, and the “Cool Bobby B’s Doo Wop Shop” show on our ’50s on 5 channel plays exclusively Doo Wop music. A complete list of Sirius XM’s 53 specialty programs is attached as SXM Dir. Ex. 20.

III. The Sirius XM Music Channels Compared To Terrestrial Radio: Promotional Benefits Of Wider Variety And More Exposure To New Music

21. In my Satellite I testimony, I discussed (at paragraphs 21-23) the fact that free terrestrial radio tends to be limited in its ability to offer anything other than the most mainstream music formats that appeal to adults. For example, a single format focused on playing both new alternative rock and hip-hop will likely appeal to male listeners who are 16 to 24 years-old. In most local radio markets, it is extremely difficult to find enough advertisers interested in that demographic to support a financially viable radio station. As a result, the tendency is for each local market to have several stations with musically conservative formats, supplemented, perhaps, with a few specialty formats supported by institutions such as colleges or local ethnic concentrations. Even in a major urban area such as Washington, D.C., it is difficult for most listeners to receive more than fifteen different music formats, and channel options in much of the country can be far fewer.

9 A. Terrestrial Radio Has Become Even More Conservative During The Past Five Years

22. Since I last testified, terrestrial radio stations have become even more conservative in embracing a range of music. Because local stations are forced to play music that appeals to relatively broad audiences and fear that listeners will change channels when unfamiliar music is played, they typically will not take the risk of playing songs with potentially narrow appeal and are less willing to introduce new songs into their limited playlists.

23. Some of the omissions forced by the constraints on terrestrial radio are striking, and the omissions have only grown more notable over the past five years. For example, two of the largest local markets in the United States, New York and San Francisco, lack any country music format. Many major markets, including New York, Los Angeles, Chicago, San Francisco,

Dallas, Houston, Atlanta, Philadelphia, Washington, D.C. and Boston, have no dance music stations. Several of the largest markets, including Houston, Atlanta, Washington, D.C., and

Boston, lack an oldies format.

24. A prime example of the increasingly conservative nature of terrestrial radio is the current state of rock radio. In March 2011, Rolling Stone published an article by contributing editor Steve Knopper titled, “Where Did the Rock Hits Go?” The article, which explored the reasons for the decline in rock music sales, attributed part of the decline to changes in radio:

Thanks to important stations switching formats in recent years, a Number One rock-radio hit reaches just 12 million listeners, compared with 81 million for Top 40. This gap is far wider than it was in 2009 than 2006, according to Nielsen SoundScan. And many of the stations that are left have seen their ratings fall.

SXM Dir. Ex. 21. Knopper wrote another article on this continuing trend in November 2011, published in Rolling Stone and titled “Rock Radio Takes Another Hit.” He there noted the continued presence of this gap: “[a] Number One rock hit reaches just 13 million listeners,

10 compared to 138 million for a Number One Top 40 hit.”4 SXM Dir. Ex. 22. With terrestrial radio stations changing formats to play more mainstream music, less new rock music is playing on terrestrial radio, and those stations that do play rock music are reaching fewer listeners.

B. Sirius XM Can Provide A Mix of Music That Terrestrial Radio Cannot

25. Currently, Sirius XM is reaching more than 21 million subscribers with its national reach, which, along with our broad channel capacity, allows us to support many different musical channels with distinctive formats. Even where a particular niche genre – bluegrass, for example – may not attract enough listeners in a given city to justify devoting a local station to the format on a full-time basis, the audience for the genre on a national basis is large enough, and our channel capacity great enough, to allow us to devote an entire channel to the genre. SXM Dir. Ex. 24 highlights the 40 music formats on Sirius XM that are generally not available on terrestrial radio.

26. Even where we offer “mainstream” channels comparable to those one would expect to find on the local radio dial (for example, classic rock), our national footprint, subscription model, and lack of advertising make us much less risk-averse than terrestrial radio programmers. Because our music stations are supported by subscription fees, not advertising, we provide listeners with mixes of music that often do not fit with the advertising interests of automobile dealerships, supermarkets, and other businesses that provide the core advertising for terrestrial radio stations. The absence of commercials also translates into a listening experience that is more sustained and satisfying than advertiser-supported radio. See SXM Dir. Ex. 25 at

Slide 16 (study conducted by Arbitron Inc., Edison Research, and Scarborough Research

4 For another discussion on the decline of rock music radio, see also Christine Pawlak, We Won’t Rock You, Slate, November 15, 2011, available at, http://mobile.slate.com/articles/arts/culturebox/2011/11/alternative_rock_radio_the_sad_unwarranted_dec line_of_fm_rock_stations_.html (attached as SXM Dir. Ex. 23).

11 demonstrates that Sirius XM listeners “love” using the satellite radio in their car more than, inter alia, AM/FM radio). The greater variety and listener satisfaction explains why subscribers are willing to purchase Sirius XM radios and pay subscription fees when music can be heard for free on terrestrial radio.

27. Because our music channels are listener-driven (and not advertiser-driven), Sirius

XM can dig much deeper into an artist’s music catalog than what is typically heard on terrestrial radio. With 71 music channels available to our subscribers, we can be more specialized and play lesser-known songs and artists that terrestrial radio would typically ignore. This music is then heard by listeners who selected a channel that is specifically committed to playing a variety of music that is wide and deep. For these reasons, listeners are much more likely to discover and purchase music when listening to Sirius XM than they are listening to terrestrial radio.

C. Sirius XM Plays And Promotes More New Music Than Terrestrial Radio

28. Because Sirius XM does not have the same constraints on its playlists as terrestrial radio stations, Sirius XM is able to place emerging songs and artists on our channels long before they might appear on terrestrial radio. For example:

Sirius XM’s Indie Rock channel, Sirius XMU, first played Amy Winehouse’s “Rehab” on November 6, 2006. In contrast, terrestrial radio stations did not begin playing the song until January 16, 2007. “Rehab” went on to become Winehouse’s most famous song (it won three Grammy Awards in 2008), and her album “Back to Black” was a massive success.

Sirius XM’s Alt Nation first began playing Foster the People’s song “Pumped Up Kicks” on June 18, 2010, before the song was officially released or even monitored by Mediabase. “Pumped Up Kicks” did not start playing on terrestrial radio until a few months later in August 2010. In 2011, the song became a major hit. On December 9, 2011, Foster the People will be doing a special performance for Sirius XM listeners. In connection with that performance, a member of the band stated: “Alt Nation played our music before any other radio outlet in the country—this show is a way for us to thank the SiriusXM team and do something special for our fans.” SXM Dir. Ex. 26.

12 29. Terrestrial radio is not exposing listeners to as much new music as does Sirius

XM. To demonstrate this, I looked at Mediabase’s “Current and Recurrent vs. Gold Ratios” for twelve Alternative and Active Rock terrestrial radio stations in the Top 10 radio markets in the nation on a particular date (in this case October 19, 2011). SXM Dir. Ex. 27. As used in

Mediabase’s calculations, “Current and Recurrent vs. Gold Ratios” compare the amount of

Current and Recurrent music to Gold music being played on a particular station. “Current music” is generally music that is currently on a most-played chart for a particular genre. After current music falls off the chart, it is considered “recurrent” for approximately two years.

“Gold” music encompasses all music in a particular genre that has been off the chart for more than two years, and hence represents older music. An examination of the ratios demonstrates that Sirius XM’s Alternative Rock station (Alt Nation) and Active Rock station (Octane) play a significantly higher percentage of Current and Recurrent music than other Alternative and Active

Rock terrestrial radio stations. SXM Dir. Ex. 27. On average, 65.8% of music played on Octane and Alt Nation is Current and Recurrent music. Id. In contrast, the twelve terrestrial radio stations examined only play an average of 35.1% Current and Recurrent music (with the highest

Current and Recurrent percentage being 59.9%, and the lowest being 16.1%). Id. This data set demonstrates that Sirius XM is exposing listeners to newer Alternative and Active Rock music, while terrestrial radio tends to play older music that the audience is more familiar with.

30. Steve Knopper of Rolling Stone has noted the importance of Sirius XM in breaking new bands in the current industry environment: “Some of those [bands] like Mumford and Sons, Vampire Weekend and the Black Keys, thrived by touring relentlessly and breaking their singles via college stations, SiriusXM programs and ad licensing to clients from Cadillac to

Tommy Hilfiger.” SXM Dir. Ex. 21 (emphasis added).

13 31. Well-known artists have attested to the fact that Sirius XM (and not terrestrial radio) is where they turn to hear new music. In an interview with widely respected music journalist and filmmaker Cameron Crowe, dated October 12, 2010, pop icon Elton John made the following comments about satellite radio:

Cameron Crowe: We are here on satellite radio now – SIRIUS – which is kind of a seismic event in the way music is heard now. What is your opinion of satellite radio?

Elton John: Thank God. Thank God, you know. It’s a savior of music, you know. . . . To have people who love music and love promoting people that haven’t been necessarily heard of, but, you know, deserve to be heard. It’s fantastic and that is the only way these people are going to get played on the radio.

Crowe: Yeah.

John: They’re not gonna be played on the normal format of radio. So, Sirius is a Godsend . . . . And nobody I know listens to normal radio. I can’t stand it. ‘Cause you hear the same 10 songs all the time . . . You listen to [satellite radio].

SXM Dir. Ex. 28.

32. Kirk Hammett, a member of the iconic heavy metal band Metallica, had the following to say:

“Satellite radio for me now, is what radio used to mean to me in the ‘70s, which played a lot of hard rock. You know, FM radio is still cool now, but they’re not pushing the envelope, not like satellite radio is.” SXM Dir. Ex. 29.

“It’s great to hear metal bands I wouldn’t otherwise hear, or wont have the time to hear, or I just don’t know about.” Id.

“Satellite radio, for me, it’s like MTV was, you know, in the ‘80s. Basically, it keeps me informed. I love that.” Id.

33. David Draiman, a member of the platinum selling hard rock band Disturbed, said the following about Sirius XM and in particular, the channel Octane, during an interview with

Sirius XM in August 2010:

14 “Look, I – honestly. I mean, you can ask my girl, you can ask anybody – that’s pretty much all I listen to. . . I listen to Octane, I listen to Hair Nation, I listen to Faction, and every once in a while I listen to a little bit of Liquid Metal.” SXM Dir. Ex. 29.

“But um, you know, I – why bother with terrestrial radio when it’s become an ipod? you know, or when it has become shuffle? When it’s this jack format nonsense? I really am much more about when I want to listen to something, that’s what I want to listen to. And, if I want to go ahead and take a break from the hard rock or heavy metal, I’ll go ahead and put on other stuff.” Id.

“But when it comes to just serious back-to-back continuity of the kind of music that we want to listen to – even – I can speak for the whole band. We love it, we’re all about it.” Id.

“But you know, for the newer stuff, I kinda count on you guys. We feel very at home you know, it’s a station you can count on to play the types of music you want to hear. And, it reflects our personal universe. You know, and it gives us kinda the lay of the land. In fact, to be perfectly honest with you, sometimes we’ll get ideas for new bands that we may wanna take out on tour from listening to you guys. . . . In all honesty, we’ve gotten some of our first, initial exposure to bands from hearing them on Octane.” Id.

34. Successful DJ Robbie Rivera has commented on satellite radio’s impact on dance music: “I think the most important thing that’s happening in America that has really brought dance music into the mainstream is probably the satellite radio shows. They have all those DJs playing their mix shows on Sirius XM. I have a mix show there, but you can listen to dance music all the time. I think that helped a lot because they put dance music all over the place, all over America.” Adam Stewart, House Legend Robbie Rivera On The State of Music, Deadmau5

And ‘Jersey Shore’ (Apr. 7, 2010), available at, http://newsroom.mtv.com/2010/04/07/robbie- rivera-deadmau5-jersey-shore/ (attached as SXM Dir. Ex. 34).

35. Newer artists also have recognized the role of Sirius XM in exposing their music to listeners. For example, during an interview in 2010, Elaine Bradley of the alternative band

Neon Trees told Sirius XM: “I never assume that somebody has heard us before. So we play a show and then people are like ‘oh you were so great.’ And I was like, ‘oh is this your first time

15 hearing us?’ And surprisingly, more and more people have said: ‘No.’ And it’s like ‘wow! where’d you hear us?’ And a lot of people say ‘Alt Nation [a Sirius XM channel].’” SXM Dir.

Ex. 31.

36. In a mid-2010 interview, Justin Tranter of the band Semi Precious Weapons commented: “I swear that every single time our song is played on Alt Nation I get – whether it is a Facebook message or a Tweet or a text – it’s been really amazing just to see what Alt Nation has done for us.” SXM Dir. Ex. 32.

37. As is apparent from these examples, not only do Sirius XM’s music channels expose listeners to music they otherwise would not hear; they also give emerging artists a unique opportunity to be heard.

D. Sirius XM Music Channels Efficiently Provide Other Promotional Benefits Not Typically Offered By Terrestrial Radio

38. Sirius XM’s marketing reach provides great value to artists that choose to work with us. Because Sirius XM can efficiently reach millions of listeners on a national scale, we provide marketing and promotional opportunities that are generally unique to satellite radio, as described below.

39. Special Programming: Sirius XM produces several different artist-specific programs that can promote artists and their music. Sirius XM produces a program called “Artist

Confidential,” each episode of which focuses on a particular artist and includes a live performance and interview. A newer initiative involves a “Town Hall” broadcast, which features a particular artist or band and a question-and-answer session with fans. We also work with artists to host Guest DJ sessions that then air on appropriate Sirius XM channels. A list of artists involved in these special programs is attached as SXM Dir. Ex. 35.

16 40. On-air Promotion: Live or recorded announcements will air across multiple

Sirius XM music, talk, and sports channels to promote an artist-specific special program or contest, typically revolving around a new album release.

41. Retail Marketing: Depending on the artist, we may offer retail marketing opportunities. For example, Sirius XM will include an artist’s name and a mutually agreed-upon song title on the retail packaging for one of Sirius XM’s radios, as well as on radio screens featured in marketing/advertising materials created with our retailers (Best Buy, RadioShack, etc.). Such marketing materials would include point-of-purchase materials and newspaper circulars. Further, Sirius XM may offer to use an artist’s image in Sirius XM retail brochures, channel guides and other marketing materials offered by our retailers and OEMs.

42. Direct Marketing: Sirius XM also has the ability to promote an artist by highlighting the artist in the email newsletters sent to Sirius XM subscribers across the country.

Sirius XM may offer to dedicate an email blast to Sirius XM subscribers promoting a particular artist’s event (for example, an album release, tour or exclusive Sirius XM programming).

43. Online Promotion: Sirius XM can promote an artist through its online presence.

Sirius XM can dedicate a page on the Sirius XM website to a particular artist. Sirius XM can also promote an artist through its social media presence (including Facebook and Twitter),

Sirius XM’s YouTube page, and ongoing viral marketing by Sirius XM.

44. Publicity: Sirius XM may issue a press release to announce the commencement of a special program created around a particular artist, such as an artist-hosted album preview, live performance or guest DJ session. These press releases are provided to national and local media outlets by the Sirius XM Corporate Communications group.

17 45. Contests: Sirius XM creates contests in connection with featured artists. If an artist is announcing a new tour or a special live performance exclusively for Sirius XM subscribers, Sirius XM can launch a sweepstakes on its platform and extensively promote it on air and online. SXM Dir. Ex. 39 provides examples of how the above-discussed marketing capabilities are offered to artists. Local terrestrial radio stations cannot provide similar promotional opportunities on a national basis.

IV. Record Labels And Artist Management Companies View Satellite Radio As A Powerful Promotional Vehicle

46. As discussed above in Section II, Sirius XM goes to great lengths to develop channel brands, identify meaningful music selections, and develop relationships between on-air talent and Sirius XM listeners. Because of all these factors, record companies and artists recognize the value of our channel brands and that airplay of a song on a Sirius XM channel connects with our listeners in a more engaging manner than airplay on an Internet-delivered music service with generic channel brands that lack hosts and other production elements or on a local terrestrial station that lacks the passion, credibility, and authenticity of a listening experience on Sirius XM. As a result, record labels and artists view satellite radio as a multi- faceted promotional vehicle.

A. Record Labels Aggressively Seek Airplay on Sirius XM Music Channels

47. While there are numerous ways in which consumers are exposed to music, like stories in newspapers and magazines, television appearances, blog postings, and music videos, the simple and universally recognized fact is that airplay on radio has continually proven to be

18 the biggest driver of record sales.5 The music industry understands this well, as evidenced by the major investments record companies make to secure radio airplay for their music.

48. Record companies have large marketing and promotion operations specifically charged with obtaining radio airplay, typically organized with regional operations under national direction. They also use independent record promoters to encourage radio programmers to play their music. Record labels provide Sirius XM and other radio outlets their recordings for free, often weeks before public release, in hopes of generating pre-release demand. The labels recognize that it is incredibly difficult to have success with an album or song without extensive airplay. This is why many of the most influential executives at the major labels tend to be those who have demonstrated an ability to get music played on the radio. In my career in radio programming, culminating in my position in charge of programming for the 71 Sirius XM music channels, the desire of record labels to receive airplay on radio has been a constant reality.

49. Sirius XM is recognized as a leader in music selection. Oftentimes, music first played on Sirius XM music channels ends up on the playlists of more mainstream terrestrial radio stations. This has resulted in record labels citing airplay of their artists and works on Sirius

XM as a reason why other programming outlets should expose these work and artists. Record companies similarly often mention airplay on Sirius XM in their advertising as a means of promoting individual artists and songs. Examples of these advertisements are attached as SXM

Dir. Ex. 36.

50. Sirius XM also receives support from artist management companies that recognize the promotional value of airplay on Sirius XM. As a result, Sirius XM has built strong

5 See, e.g., James N. Dertouzos, Radio Airplay and the Record Industry: An Economic Analysis, (June 2008), attached as SXM Dir. Ex. 37; Edison Research, The National Record Buyers Survey (2001), attached as SXM Dir. Ex. 38.

19 relationships with well-known management companies such as Red Light Management (Dave

Matthews Band, Tim McGraw), Q Prime (Red Hot Chili Peppers, Muse), and Monotone

Management (Vampire Weekend, Foster the People).

51. In recognition of the promotional benefits of airtime on Sirius XM’s channels, artists and/or their record labels are often willing to grant waivers of certain of the Section 114 statutory license restrictions and in particular, a waiver of the number of times Sirius XM can play a song in a given time period, in order to allow Sirius XM programmers to highlight more of the music from a particular artist or new album on our channels.6 For example:

In April 2011, Warner Brothers agreed to a waiver regarding certain music of Metallica from April 9, 2011 through September 7, 2011 for use on a full-time channel dedicated principally to the music of Metallica. SXM Dir. Ex. 40. The waiver permitted Sirius XM to play certain songs “without restriction as to the number of songs from the same phonorecord or by the same artist that may be played in a certain time period.” Id. Additionally, Sirius XM was permitted to “promote and publicize, through the use of any means,” the Metallica songs covered by the waiver. Id.

In November 2009, Warner Brothers agreed to a waiver regarding certain music of Tom Petty. SXM Dir. Ex. 40. The waiver allowed Sirius XM to play music from “The Live Anthology” without restriction as to airplay from November 18, 2009 through December 31, 2009.

In September 2010, Universal Music Group granted Sirius XM the right to play certain of Elton John’s songs from October 14, 2010 through October 22, 2010 in connection with a feature to be broadcast on Sirius XM. SXM Dir. Ex. 40. The waiver permitted Sirius XM to play certain songs “without restriction as to the number of songs . . . in a certain time period.” Id. Additionally, Sirius XM was permitted to promote the music, “through the use of any means,” including “announcing the dates and times of transmissions.” Id.

Sirius XM has received hundreds of such waivers in connection with recordings by artists including the following:

6 These waivers relate to the “sound recording performance complement,” which limits the number of tracks that can be played from a particular artist or album in a three-hour period.

20 Katy Perry (Capitol Records)

Eminem (Interscope Records)

Bon Jovi (Island Def Jam Music Group)

Vampire Weekend (XL Recordings)

Janet Jackson (EMI Music North Americas)

Rihanna (Island Def Jam)

SXM Dir. Ex. 40. Record companies obviously grant these waivers because they recognize the value in airplay on Sirius XM’s music channels, and the ability of such airplay to generate sales.

52. Artists are often interviewed by Sirius XM personalities or perform live in our studios. These performances and interviews provide another promotional opportunity for artists.

In one recent example, on June 16, 2011, Aretha Franklin recorded a track-by-track preview of her new album and sat down for an interview with Sirius XM. In connection with that appearance, Franklin granted Sirius XM the right to use her name, image, and likeness and any transcripts, audio, video, photographs, and other recordings of the interview, performance, and/or appearance. SXM Dir. Ex. 41 We have received hundreds of similar such releases from artists including Patti Smith, Carrie Underwood, Colbie Callait, Queen Latifah, Jack Johnson, Lady

Antebellum, and Kings of Leon. Id.

B. Record Labels Recognize That Beyond Airplay, Sirius XM Music Channels Offer Unique Promotional Opportunities Through Artist-Specific Channels And “Pop-up” Channels

53. Beyond regular airplay on our music platform, there are numerous other promotional benefits that artists and record labels enjoy through arrangements with Sirius XM.

54. Record companies typically focus their promotional efforts on new music, rather than their existing catalog. As a result, a vast amount of older music remains unexposed, and thus effectively unavailable, to many listeners. As discussed above in Section II, we have

21 RESTRICTED – Subject to Protective Order in Docket No. 2011-1 CRB PSS/Satellite II developed channels and programs in most of the main music categories in conjunction with well- known artists such as Elvis Presley and Bruce Springsteen, which focus exclusively on the music of Presley (Elvis Radio) and Springsteen (E Street Radio). Similarly, Jimmy Buffett’s Radio

Margaritaville plays the music of several artists with older catalogs of music, including Jimmy

Buffett, Bob Marley, Paul Simon, Little Feat, The Beach Boys, and Santana. These artist- themed channels provide a unique opportunity to record labels to promote music that otherwise would have very little opportunity for new exposure to listeners.

55. Notably, and evidencing the promotional value artists see in this experience,

Sirius XM does not need to pay significant amounts to these artists to use their names and likeness in conjunction with our artist-themed channels. While we made some payments at the start of some of these relationships, artist have, by agreement, in most cases been reduced significantly or eliminated.7

56. Record labels and artists often approach Sirius XM with an expressed interest in creating an artist-themed “pop-up” channel (these temporary channels are distinct from our long- term artist-themed channels such as Elvis Radio and Siriusly Sinatra). These limited- engagement artist channels focus exclusively on the named artist’s catalog of music, and provide a record label with the opportunity to generate sales in an older catalog of music that may not otherwise receive a lot of promotional exposure. Temporary artist channels have included arrangements with George Strait and the estate of Miles Davis, among others.

22 57. Similarly, there is also significant interest from record labels and artists to work with Sirius XM to develop artist-themed “pop-up” channels that do not exclusively play the named artist’s music. While these short-term channels do not exclusively play a particular artist’s music, they do have a significant focus on a particular artist’s catalog of music.

Examples of these types of artist-branded “pop-up” channels include:

Coldplay (EMI/Capitol)

Metallica (Warner Bros. Records)

Linkin Park (Warner Bros. Records)

R.E.M. (Warner Bros. Music)

Depeche Mode (Warner Bros. Music)

Sheryl Crow (Universal Music Group)

Weezer (Universal Music Group)

Duran Duran (Sony/BMG)

Dave Matthews Band (RCA)

Michael Jackson (Epic)

AC/DC (Columbia)

58. Additionally, record labels with artist catalogs from the 60s and 70s have approached Sirius XM seeking ways to generate sales in these older catalogs of music, which receive play on classic rock and oldies radio stations but no focused promotional attention. In response, we have created temporary “pop-up” channels with artists like Elton John and Simon and Garfunkel.

V. Airplay On Sirius XM Leads To Increased Record Sales

59. Going back to the birth of satellite radio, artists, their managers, and record labels have always recognized that increased record sales can be attributed to airplay on Sirius and XM

23 RESTRICTED – Subject to Protective Order in Docket No. 2011-1 CRB PSS/Satellite II channels (and now Sirius XM channels). Time and again during my tenure at Sirius and now

Sirius XM, airplay of a particular song on Sirius XM’s music channels has led directly – and demonstrably – to increased sales volumes of those songs and the albums on which they appear.

A. Record Labels Frequently Acknowledge Sirius XM’s Positive Impact On Sales

60. Over the past several years, artists and record labels have increasingly become vocal in acknowledging the impact that exposure of their works on Sirius XM has had on digital sales of those works. A sampling of these communications is provided below.

Bob Schneider’s Album “A Perfect Day”

61. On May 2, 2011, emailed programming executives at Sirius XM to report on the success an artist named Bob Schneider had attained by virtue of exposure on Sirius XM. While Schneider is popular in Austin, Texas, he had received only limited national exposure until we began playing his music. Reporting on Schneider’s

“BEST scan week EVER,” commented: “[t]he sales are pretty spread out around the nation which I am attributing to your airplay.” SXM Dir. Ex. 42 (emphasis added).

“Bounce” By Emphatic

62. On May 20, 2011, emailed Sirius XM regarding increased sales for a song titled “Bounce” by the band Emphatic. reported a

“4 week trend on ‘Bounce’ = 285, 319, 328, 854! Up 160% this week. Seems to coincide[] pretty well with your airplay. Other stations just starting to play. Just wanted to say thanks for the HUGE support on ‘bounce by Emphatic!!!” SXM Dir. Ex. 43 (emphasis in original).

“I Am Doing Alright” By Jacob Lyda

63. On June 8, 2011, emailed Mark Sebastian, Senior Director of Country Music Programming at Sirius XM (known

24 RESTRICTED – Subject to Protective Order in Docket No. 2011-1 CRB PSS/Satellite II as “John Marks” in the industry), regarding the artist Jacob Lyda’s song “I’m Doing Alright.”

He stated: “You guys re-engaged and my single sales doubled this week! It’s rare that I can directly correlate my sales like that. You’re the best!” SXM Dir. Ex. 44.

“Ultragigantor” By Red Line Chemistry

64. During the summer of 2011, the Octane and Faction channels began playing the band Red Line Chemistry’s song “Ultragigantor.” On August 3, 2011,

emailed Jose Mangin (Music Programmer for Octane) and reported that since Sirius XM began playing “Ultragigantor,” “we have seen an explosion of sales we know it’s attributed to you since we see the reaction coming from non radio markets (album sales up

25%, single sales up 60% this week alone!) So thank you!” SXM Dir. Ex. 45.

Daniel Glass, President and CEO of Glassnote Records

65. Daniel Glass, President and CEO of Glassnote Records, has recognized the promotional aspect of Sirius XM’s music channels. In the May 2011 issue of industry trade magazine Hits, Glass stated “Sirius XM continues to grow and support meaningful music reaching more people than ever and we feel the impact on sales.” SXM Dir. Ex. 46.

B. Industry Sales Data Establishes That Sirius XM Promotes Sales

66. In addition to testimonials of the sort set forth above, since the Satellite I proceeding, we have tracked industry data from sources such as SoundScan (for sales data) and

Mediabase (for airplay data) which lend further support for the proposition that airplay on Sirius

XM promotes sales. I again provide several examples of this phenomenon.

“Colours” By Grouplove

67. During the first half of 2011, Alt Nation began playing the song “Colours” by the new artist Grouplove. On May 12, 2011,

25 RESTRICTED – Subject to Protective Order in Docket No. 2011-1 CRB PSS/Satellite II

emailed Jeff Regan, Music Programmer for Alt Nation, to report on digital single sales for Coulors. wrote: “[D]igital single sales on ‘Colours’ were up again from 1,049 to 1,195. Pretty great and because of Alt Nation. That is the only explanation.” SXM Dir. Ex. 47 (emphasis added). wrote again to various members of the

Sirius XM Music Programming team on June 13, 2011: “Since you started playing Grouplove

‘Colours’ on Alt Nation we have seen our digital single sales go way up, it can [be] directly attributed to your airplay. Our sales went from 200 singles a week to 800 to around 1,000-1,200 a week since you started playing it.” Id. (emphasis added). Asking the Sirius XM team to attend the band’s launch in New York City continued that “I think [it] would [be] important for you to attend, your early airplay to sales is certainly our focal point in our planning meetings on the band and very inspiring to us in our launch. We are so excited to have you leading the charge.” Id.

68. Industry sales data corroborates Alt Nation’s positive effect on sales of Colours.

Mediabase data establishes that Alt Nation first played Colours on April 18, 2011. SXM Dir. Ex.

48. Alt Nation was not the first radio station to play the song, but the only radio station to give it more than ten spins per month prior to that time was WSUN-FM in Tampa, Florida. Id. For the week ending April 17, 2011 (the week before Alt Nation started playing the song), SoundScan data shows that there were a total of only 221 total digital sales of the work. SXM Dir. Ex. 49.

During the second half of April (beginning on April 18, 2011), Alt Nation played Colours 69 times. SXM Dir. Ex. 48. During the first week Alt Nation played Colours (the week ending

April 24, 2011), digital sales jumped to 773 – a 250% increase from the week prior. SXM Dir.

Ex. 49. The following week (ending May 1, 2011), digital sales jumped again to 1,049 – a 36% increase from the prior week. Id. During the third week of airplay on Alt Nation (the week

26 ending May 8, 2011), digital sales jumped from 1,049 to 1,195 (a 14% increase). Id. While

WSUN-FM in Tampa continued to play the song and one other station in Little Rock, Arkansas gave the record some spins, no other radio stations were giving significant airplay to Colours during April and May 2011. SXM Dir. Ex. 48. This surge in record sales can only be attributed to Alt Nation’s airplay.

“Trojans” By Atlas Genius

69. In September 2011, Sirius XM’s Alt Nation was the only music station in

America playing the music of Australian band Atlas Genius. The lead singer of the band, Keith

Jeffery, wrote to Jeff Regan, Music Programmer for Alt Nation, “[p]eople from MTV and other management groups have contacted us after hearing Trojans on Alt Nation, and we’ve picked up a lot of fans.” SXM Dir. Ex. 50. Jeffery wrote to Regan again on September 29, 2011: “Just want to thank you for the continued support. We’ve been getting a crazy amount of traffic.” Id.

70. SoundScan and Mediabase data again demonstrate that Alt Nation’s airplay of this work had a significant impact on sales. According to Mediabase, Alt Nation first played

Trojans on September 13, 2011, at which time it was the only radio station in the country playing the song. SXM Dir. Ex. 51. As of September 11, 2011, there were 405 digital downloads of the song. SXM Dir. Ex. 52. From September 13, 2011 through the end of September, Sirius XM’s

Alt Nation played the record 88 times. SXM Dir. Ex. 51. For the week ending September 18,

2011 – representing the first week of airplay on Alt Nation – digital sales of “Trojans” jumped

88% (to 761). SXM Dir. Ex. 52. By October 2, 2011, as Alt Nation continued to feature the song, digital sales had risen even further – to a total of 829. Id. For the week ending October 9,

2011, digital sales rose 12% to 930. For the week ending October 16, 2011, sales continued to

27 rise another 29% (to 1199). Id. The table below illustrates the dramatic increase in digital sales of Trojans following Alt Nation’s addition of the song to its playlist.

The correlation between Alt Nation’s airplay of the song during September and October 2011 and the increase in sales of Trojans during this time period seems indisputable.

“Dancing Shoes” By Green River Ordinance

71. In November 2011, Green River Ordinance sent Sirius XM a “thank you” video message. The band stated:

Thank you so much for playing “Dancing Shoes.” It’s been crazy, just the response and the people have been hearing our music for the first time . . . thanks so much for playing the song and just taking a chance on new music . . . that’s kinda one of the coolest things that has ever happened to our band and so people are lovin’ it, a bunch of our fans are listening and just new people who have never heard of the Green River Ordinance . . . so you guys are the reason for that and we just wanted to say thanks. . .

SXM Dir. Ex. 33.

72. Industry data corroborate that The Highway’s airplay of this work had a considerable impact on sales. On October 24, 2011, Sirius XM’s The Highway became the first and only radio station to play Green River Ordinance’s song “Dancing Shoes.” SXM Dir. Ex.

53. Prior to the song’s addition to The Highway’s playlist, sales of the song were very weak.

28 SXM Dir. Ex. 54 The week after the song began receiving airplay (the week ending October 30,

2011) on The Highway, sales skyrocketed from 58 to 1,105 digital sales – the SoundScan data notes this increase as a 999% change from the previous week. Id. This trend continued during its second week of airplay on Sirius XM, as sales increased 29%. Id. During its third week of airplay on The Highway (the week ending November 13, 2011), sales increased again by 13%.

Id. The dramatic increase in sales that occurred when The Highway added the song to its playlist is demonstrated in the chart below.

The direct correlation between airplay on The Highway and the dramatic increase in digital sales of Dancing Shoes seems apparent.

“Shuffle” By Bombay Bicycle Club

73. Another recent band that had breakout success after getting its song played on

Sirius XM’s music channels is the Bombay Bicycle Club. Sirius XM’s Alt Nation began playing

Bombay Bicycle Club’s song “Shuffle” on October 6, 2011. SXM Dir. Ex. 55. In the weeks prior to getting played on Alt Nation, sales of Shuffle were demonstrating a declining trend.

SXM Dir. Ex. 56. During the first week of airplay on Alt Nation (week ending October 9, 2011), sales of the song improved by 25%. Id. During the second week of airplay on Alt Nation (week

29 ending October 16, 2011), sales of the song increased even more – this time by 122%. Id. The table below illustrates this trend and the impact of Alt Nation’s addition of it to its playlist.

VI. Conclusion

74. Sirius XM offers 71 commercial-free music channels in a wide variety of music genres. Each of these originally-produced radio channels is hand-crafted by our team of programming and production experts and are presented to more than 21 million subscribers a week. The combination of our nationally-branded music channels, our stable of talented DJs and artist partners, along with the vast array of specialty programming and unique promotional opportunities offered, creates tremendous value for the artists we play and their respective record labels and management companies. This is evidenced by the direct feedback we regularly receive from artists, their record companies, and artist managers, as well as by the airplay data and highly correlated music sales data illustrated in my testimony.

Capital Reporting Company RESTRICTED Subject to Protective Order in Docket No. 2011-1 CRB PSS-Satellite II

957 RESTRICTED: Subject to Protective Order in Docket No.

2011-1 CRB PSS/SATELLITE II

UNITED STATES COPYRIGHT ROYALTY JUDGES

WASHINGTON, D.C.

------X ) In the Matter of: ) ) Docket No. 2011-1 Determination of Rates and Terms ) CRB PSS/Satellite II for Preexisting Subscription ) Services and Satellite Digital ) Audio Radio Services ) Volume IV ) Pgs. 957 - 1304 ------X

Washington, D.C.

Friday, June 8, 2012

The following pages constitute the continued proceedings held in the above-captioned matter, held at the Library of Congress, Madison

Building, 101 Independence Avenue, Southeast,

Washington, D.C., before Cindy L. Sebo,

RMR/CRR/CSR/RPR/CCR/RSA of Capital Reporting Company, a Notary Public in and for the District of Columbia, beginning at approximately 9:39 a.m.

958 960

1 A P P E A R A N C E S 1 A P P E A R A N C E S (Continued): 2 Copyright Royalty Tribunal: 2 On behalf of Music Choice: 3 CHIEF JUDGE SUZANNE M. BARNETT 3 PAUL M. FAKLER, ESQUIRE JUDGE WILLIAM ROBERTS 4 MATTHEW TROKENHEIM, ESQUIRE 4 JUDGE STANLEY C. WISNIEWSKI 5 Arent Fox LLP 5 6 1675 Broadway 6 On behalf of SiriusXM: 7 New York, New York 10019 7 R. BRUCE RICH, ESQUIRE 8 212.457.5445 8 TODD LARSON, ESQUIRE 9 [email protected] 9 MIRANDA S. SCHILLER, ESQUIRE 10 [email protected] 10 RANDI W. SINGER, ESQUIRE 11 11 SABRINA A. PERELMAN, ESQUIRE 12 12 Weil, Gotshal & Manges LLP 13 13 757 Fifth Avenue 14 14 New York, New York 10153-0119 15 212.310.8170 15 16 [email protected] 16 17 [email protected] 17 18 [email protected] 18 19 [email protected] 19 20 [email protected] 20 21 21 22 22

959 961

1 A P P E A R A N C E S (Continued): 2 1 C O N T E N T S On behalf of SoundExchange: 2 WITNESSES: 3 DAVID A. HANDZO, ESQUIRE 3 STEVEN G. BLATTER DIRECT CROSS 4 REDIRECT MICHAEL B. DESANCTIS, ESQUIRE 5 4 By Ms. Perelman 965 ---- 1034 JARED O. FREEDMAN, ESQUIRE 6 5 By Mr. Moskowitz ---- 1004 ---- GARRETT A. LEVIN, ESQUIRE 6 7 DAVID Z. MOSKOWITZ, ESQUIRE WILLIAM ROSENBLATT EXAMINATION DIRECT 8 CROSS REDIRECT Jenner & Block 9 7 ON QUALIFICATIONS 1099 New York Avenue, Northwest 8 By Mr. Larson 1038 1050 ---- 1189 10 Suite 900p 9 By Mr. Levin ------1099 ---- 11 10 Washington, D.C. 20001-4412 12 DAVID PAUL STOWELL EXAMINATION DIRECT 202.639.6085 CROSS REDIRECT 13 [email protected] 11 ON QUALIFICATIONS 14 [email protected] 12 By Ms. Schiller 1198 1202 ------15 13 By Mr. Freedman ------1245 ---- [email protected] 16 14 [email protected] 15 17 [email protected] 16 SIRIUSXM TRIAL EXHIBITS: * MARKED 18 ADMITTED 19 20 17 15 966 ---- 21 18 16 976 ---- 22 19 17 1038 ---- 20 18 1197 1219 (866) 448 - DEPO www.CapitalReportingCompany.com © 2012 Capital Reporting Company RESTRICTED Subject to Protective Order in Docket No. 2011-1 CRB PSS-Satellite II

962 964

1 C O N T E N T S (Continued) 1 P R O C E E D I N G S 2 SOUNDEXCHANGE TRIAL EXHIBITS: * MARKED 2 CHIEF JUDGE BARNETT: Good morning. ADMITTED 3 Please be seated. Thank you for your 3 25 1005 ---- 4 patience. 4 26 1007 ---- 5 I want to give everyone a heads up 5 27 1010 ---- 6 about the schedule on Monday. We will be recessing 6 28 1010 ---- 7 at 3:45 on Monday. And we've checked the schedule, 7 29 1023 ---- 8 and we believe that that time is still available on 8 30 1105 1107 9 the 19th. 9 31 1109 1110 10 But if you have witness scheduling 10 32 1112 1113 11 issues, I wanted you to be aware we will finish up 11 33 1117 1118 12 at 3:45 on Monday, and we're not going to adjust 12 34 1123 1124 13 the daily schedule any other day to accommodate for 13 35 1128 1129 14 that. 14 36 1132 ---- 15 Okay? 15 37 1137 1139 16 MR. RICH: Are you ready for our next 16 38 1138 1139 17 witness? 17 39 1140 1142 18 CHIEF JUDGE BARNETT: Ready if you are. 18 40 1144 1144 19 MR. RICH: SiriusXM calls Steve Blatter 19 41 1155 1156 20 to the stand. And Mr. Blatter will be examined by 20 42 1159 1160 21 my colleague, Ms. Perelman. 21 43 1162 1163 22 22

963 965

1 C O N T E N T S (Continued) 1 WHEREUPON, 2 SOUNDEXCHANGE TRIAL EXHIBITS: * MARKED 2 STEVEN G. BLATTER ADMITTED 3 called as a witness, and having been first duly 3 44 1164 1165 4 sworn, was examined and testified as follows: 4 45 1169 ---- 5 CHIEF JUDGE BARNETT: Ms. Perelman. 5 46 1251 ---- 6 MS. PERELMAN: Good morning, 6 7 Your Honors. 7 8 8 9 DIRECT EXAMINATION 9 (* Exhibits Retained by Counsel.) 10 10 11 BY MS. PERELMAN: 11 12 Q. Good morning, Mr. Blatter. 12 13 A. Good morning. 13 14 Q. Mr. Blatter, could you please tell us 14 15 by whom are you currently employed? 15 16 A. SiriusXM Radio. 16 17 Q. How long have you worked there? 17 18 A. Approximately nine years now. 18 19 Q. And since -- have you worked there 19 20 since the merger and also for some time before the 20 21 merger between Sirius and XM? 21 22 A. Yes. I joined Sirius Radio in 2003, 22

966 968

1 the merger was 2008, and I've been with the company 1 November. It's on Page -- the top -- the top of 2 since. 2 Page 15. 3 Q. Prior to join Sirius, how long have you 3 There's a reference to an exhibit that 4 been involved in the radio business? 4 should have been Exhibit 30. It was Exhibit 29. 5 A. I've been in radio for 25 years now. 5 So it's been corrected in the version 6 Q. What is your current position at 6 that Your Honors have, and both parties consent to 7 SiriusXM? 7 that correction. 8 A. I'm the senior vice president and 8 So with that correction included, I 9 general manager of music programming. 9 move that the written direct testimony of 10 Q. And what are your main responsibilities 10 Steven Blatter, together with the accompanying 11 as senior vice president and general manager of 11 exhibits, which are Exhibits 1 and then 17 12 music programming? 12 through 56, be moved into evidence subject to the 13 A. I oversee all of our commercial-free 13 protective order. 14 music programming, including the 285 employees that 14 MR. MOSKOWITZ: Good morning, 15 are responsible for the curation and presentation 15 Your Honors. My name is David Moskowitz 16 of our channels to our 22 million subscribers. 16 representing SoundExchange. 17 Q. Mr. Blatter, you should have in front 17 We have no objection to the -- to the 18 of you and the Judges also have what has been 18 majority of these exhibits, but we do object to 19 marked for identification as SiriusXM 19 Exhibits 25, 37 and 38 on the grounds that these 20 Trial Exhibit 15. 20 are -- essentially, this witness is a fact witness. 21 (SiriusXM Trial Exhibit Number 15 was 21 These are academic studies and surveys. He has no 22 marked for identification purposes.) 22 knowledge or experience in these areas to talk

967 969

1 BY MS. PERELMAN: 1 about these studies. 2 Q. Do you see that? 2 Moreover, his testimony is not really 3 A. Yes. 3 being used to even analyze or illustrate. If 4 Q. Do you recognize this document as your 4 you -- for example, the Slide 16 on Exhibit 25 is 5 written direct testimony in this proceeding? 5 cited in the testimony as explaining -- if you look 6 A. Yes, I do. 6 at Slide 16 on that, it shows, How much do you 7 Q. If you could turn to the last page 7 enjoy using satellite radio in your car? And it's 8 behind the first tab. 8 cited in the testimony as saying that SiriusXM 9 Do you recognize that as your 9 users love commercial-free music. 10 signature? 10 In any event, with these studies, we 11 A. Yes, I do. 11 have no ability to cross-examine the witness other 12 Q. And are the attached exhibits the 12 than to -- as what is stated specifically in these 13 documents and other materials that you reference in 13 studies. But we can't examine the methodologies 14 your written direct testimony as being 14 with the witness or anything along those lines. We 15 corroborative or illustrative of the material you 15 think they should not be admitted. 16 discuss in that document? 16 JUDGE WISNIEWSKI: Mr. Moskowitz, could 17 A. Yes. 17 you repeat which ones? 18 MS. PERELMAN: Your Honors, for the 18 MR. MOSKOWITZ: Sure. 19 record, the version of SiriusXM Exhibit -- 19 These are 25, 37 and 38. 20 Trial Exhibit 15 that we've handed up contains one 20 JUDGE WISNIEWSKI: Thank you. 21 small correction that we've exchanged with counsel 21 MR. MOSKOWITZ: And we have a few 22 as -- as compared to the version that was filed in 22 additional objections as well.

970 972

1 CHIEF JUDGE BARNETT: Let's deal with 1 all affected by these academic studies and surveys 2 the exhibits. 2 and how we would effectively be able to 3 Ms. Perelman, with regard to 25, 37 and 3 cross-examine him on those surveys and why they 4 38, are those being offered for the truth of the 4 should be in evidence. 5 matter? 5 If SiriusXM wants these to be in 6 MS. PERELMAN: They're not being 6 evidence, they should provide an expert to give us 7 offered for the truth of the matter, Your Honor. 7 a study so that we can actually examine those 8 As we discussed the other day with 8 studies rather than a fact witness who doesn't know 9 Mr. Meyer, these exhibits could have just been 9 anything about them. 10 cited instead of attached. They're added for -- 10 JUDGE ROBERTS: Counsel said they're 11 for the Judges' convenience, and they're merely 11 not being offered for the truth of the assertion; 12 illustrative of the kinds of things that 12 they're being offered as backing up what he has to 13 Mr. Blatter testifies to from his own personal 13 say about them. 14 knowledge based on his 25 years of experience. 14 MR. MOSKOWITZ: It does not make any 15 And, you know, the rules of evidence in 15 sense that they are not being offered for the truth 16 this proceeding is slightly more flexible than in 16 of the matter asserted. If the whole point is this 17 the -- in the -- under the Federal Rules of 17 is backing up what he's saying, they are being 18 Evidence. 18 completely offered for the truth of the matter 19 Under 37 CFR 351.10(a), anything that's 19 asserted. 20 relevant and not unduly repetitious or privileged 20 His assertion here is that it is 21 shall be admissible, and hearsay should be 21 universally recognized -- the exact statement -- 22 admissible when it's appropriate. 22 JUDGE ROBERTS: The simple and

971 973

1 And we think it's appropriate in this 1 universally recognized fact is that airplay on 2 instance not for the truth, but simply just to 2 radio has continually proven to be the biggest 3 supplement Mr. Blatter's personal knowledge. 3 driver of record sales. 4 JUDGE ROBERTS: Ms. Perelman, where in 4 MR. MOSKOWITZ: If a witness has 5 Mr. Blatter's testimony are these exhibits 5 personal knowledge of this, he can testify to that, 6 referenced? 6 but he's just using these as factually accurate to 7 MS. PERELMAN: Sure. 7 say this is the facts and that's what I'm saying to 8 Exhibit 25, which is the Arbitron 8 be the fact. 9 research, is referenced in Paragraph 26 on Page 11. 9 CHIEF JUDGE BARNETT: Well, I think 10 The cite is at the bottom of Page 11 and goes on to 10 what he's saying is that I took certain actions 11 the top of Page 12. 11 based upon my belief and my belief is based upon 12 JUDGE ROBERTS: Next. 12 this study, which -- 13 MS. PERELMAN: Exhibit 37 is cited at 13 MR. MOSKOWITZ: If his belief is based 14 Paragraph 47 on Page -- in Footnote 5 on Page 19. 14 on this study, then that's being offered for the 15 And 38 is in the same footnote. 15 truth. 16 JUDGE ROBERTS: What about that, 16 CHIEF JUDGE BARNETT: No, I don't think 17 Mr. Moskowitz? They're providing these articles so 17 so. 18 we don't have to go out and hunt them down. 18 JUDGE ROBERTS: He's saying -- I read 19 MR. MOSKOWITZ: If this were an expert 19 this somewhere -- here's where I read it -- 20 witness, I think that there's no problem here, but 20 CHIEF JUDGE BARNETT: I'll tell you 21 he is testifying based on his personal knowledge, 21 what. The objection is overruled to those. 22 and it's not clear how his personal knowledge is at 22 Go ahead. What's your other objection?

974 976

1 MR. MOSKOWITZ: We also object to 1 that, we'll make sure not to use it on the open 2 Exhibit 36, which are purported to be a bunch of 2 record. 3 advertisements that we don't know where they came 3 CHIEF JUDGE BARNETT: The version that 4 from. 4 you gave us this morning, does that include the 5 They're clearly altered advertisements. 5 gray lined -- 6 The things have been highlighted by SiriusXM. 6 MS. PERELMAN: It's grayed out. 7 They're not the originals. They're just altered 7 CHIEF JUDGE BARNETT: Okay. Thank you. 8 advertisements from unknown sources. 8 So 15 is admitted subject to the 9 CHIEF JUDGE BARNETT: I'm sorry. 9 restrictions. 10 That's 36? 10 MS. PERELMAN: Thank you. 11 Ms. Perelman? 11 Also, just for the record, we have 12 MS. PERELMAN: So to the extent that 12 marked but we won't distribute today what's been 13 the -- that the objection is as to authentication, 13 marked as Exhibit 16, which is Mr. Blatter's 14 we can discuss with the witness where these were 14 previously designated testimony from Satellite I, 15 pulled from and by whom. 15 which we will proffer on Monday along with 16 The alterations that Mr. Moskowitz 16 everything else pursuant to your order from 17 points are simply yellow highlights to point out in 17 yesterday. 18 sort of a cluttered advertisement where 18 CHIEF JUDGE BARNETT: Thank you. 19 specifically we're pointing -- the Judges and the 19 (SiriusXM Trial Exhibit Number 16 was 20 parties to what we're citing here. 20 marked for identification purposes.) 21 And these -- again, to the extent it's 21 BY MS. PERELMAN: 22 a substantive evidentiary objection, these are 22 Q. Mr. Blatter, now that all those

975 977

1 certainly not offered for their truth, but just for 1 technicalities are out of the way, can you describe 2 the facts that these advertisements are out there 2 for us how many full-time, commercial-free music 3 and to show the state of mind of the labels who put 3 channels does SiriusXM offer? 4 them out. 4 A. So we currently offer approximately 70 5 JUDGE ROBERTS: Where does Mr. Blatter 5 commercial-free music channels. 6 reference this in his testimony? 6 Q. How does SiriusXM's music lineup 7 MS. PERELMAN: This is referenced in 7 compare to what's available on terrestrial radio? 8 Paragraph 49 on Page 19. 8 A. So compared to terrestrial radio, there 9 CHIEF JUDGE BARNETT: The objection to 9 are a few differences. First is we offer more 10 that one is also overruled. 10 channels. So in a typical terrestrial market 11 So Exhibit 15 is admitted. 11 across America, there's anywhere from approximately 12 (SiriusXM Trial Exhibit Number 15 was 12 five to maybe 15 music channels that are made 13 admitted into evidence.) 13 available in that local terrestrial market. 14 MS. PERELMAN: Thank you, Your Honor. 14 SiriusXM, as I just mentioned, offers 15 Just a quick note. When we filed 15 far more than that with its 70 or so 16 what's now Exhibit 15 in evidence in November, 16 commercial-free music channels. 17 there were certain portions of it that were viewed 17 Another difference is the -- the number 18 as restricted and sensitive information that were 18 of niche channels that SiriusXM offers. There 19 grayed out and that a motion as to those portions 19 are -- you know, in most terrestrial markets, the 20 under the protective order has already been 20 formats that are offered are very mainstream 21 granted. 21 formats, but niche formats are not really heard 22 To the extent we need to address any of 22 today on terrestrial radio on a full-time basis.

978 980

1 And the third is there are actually 1 full-time basis in most, if not all, local radio 2 certain mainstream formats in big, big cities 2 markets across America. 3 across America today that are no longer served in 3 Q. Have there been, to your knowledge, any 4 those markets. 4 changes in the number or nature of formats 5 Q. Just -- just to illustrate a little 5 available on terrestrial radio over the last 6 bit, what are some examples of the kinds of formats 6 five years? 7 that you described that are not generally available 7 A. Well, terrestrial radio has, over the 8 on terrestrial radio anywhere? 8 last five years, become increasingly conservative 9 A. Some examples of that -- and there are 9 in a couple of ways. You know, I mentioned earlier 10 many -- are music styles like reggae, bluegrass 10 that there are now, you know, very mainstream 11 music, jazz and -- and the many subgenres of jazz. 11 formats that are no longer made available on 12 Q. And you mentioned that there might be 12 terrestrial radio as they might have been five 13 certain geographic regions that might not have some 13 years ago. 14 kind of a mainstream format that's available 14 And for the stations that do continue 15 elsewhere. 15 to play music on terrestrial radio, they have 16 Can you give us an example of that? 16 become increasingly conservative over the last 17 A. Yes. 17 five years, and their ability to expose new music 18 One such example is country music, one 18 or deeper playlists of music has -- has declined 19 of the most popular music styles there is in -- in 19 pretty dramatically. 20 America today. But country music, despite its 20 Q. Mr. Blatter, switching topics for a 21 popularity, is not of -- it's not made available on 21 moment, are you familiar with SiriusXM's direct 22 radio in markets -- major markets, like New York 22 licensing initiative?

979 981

1 and San Francisco. 1 A. Yes, I am. 2 Q. In your experience, why does 2 Q. Have you had any role in that direct 3 terrestrial radio not typically make those kinds of 3 licensing initiative? 4 formats available? 4 A. I have not been involved in that 5 A. Terrestrial radio, you know, unlike how 5 process directly. I've been kept abreast of the -- 6 we make decisions in -- in satellite radio, 6 of the -- the progress of the initiative, but I'm 7 decisions are often made based on the desires of 7 not -- not involved firsthand. 8 advertisers; while on satellite radio, our 8 Q. Okay. Now, you speak in your written 9 decisions are made and are completely 9 direct testimony at Pages 4 and 5 about how 10 listener-driven. 10 programming is developed and curated by your 11 Q. Okay. And just to illustrate, if you 11 programmers and how music is selected. 12 could flip to what's marked as Direct Exhibit 24 12 And we won't go into detail on that, 13 appended to your testimony. 13 but I just wanted to flag that for the record. 14 A. Okay. 14 But turning to Paragraph 11 after that 15 Q. Do you recognize this document? 15 section, which is on Pages 5 and 6 of your 16 A. Yes, I do. 16 testimony, there, you state that Going forward, 17 Q. Can you tell us what Exhibit 24 17 SiriusXM will take into account whether there's a 18 illustrates? 18 direct licensing relationship between the company 19 A. This exhibit lists all our 19 and the record label when it makes those 20 commercial-free music channels, and highlighted in 20 programming decisions. 21 yellow are the channels that we offer that have 21 What did you mean by that? 22 formats that are generally not available on a 22 A. So while the artist and song will --

982 984

1 will -- the merits of the artist and song will 1 benefits"? 2 always come first, the -- whether a song is 2 A. Well, typically, what they -- what the 3 directly licensed or not will be yet just another 3 artists that we work with find is by having their 4 factor that programmers will use in helping 4 own radio stations, they're able to generate more 5 determine what songs to play and how often. 5 interest in -- whether it be their concert tours 6 Q. Would you ever allow any of the 6 or, in other cases, their -- recorded music and the 7 programmers who work with you to sacrifice the 7 sale of it. 8 quality or the integrity of the channel to benefit 8 Q. Does SiriusXM also have limited run or 9 a direct license just for that reason? 9 what's called pop-up channels? 10 A. Absolutely not. The quality of our 10 A. Yes, we do. 11 channels and the integrity of them will always come 11 Q. What are pop-up channels and how did 12 first. 12 they come to be? 13 Q. Now your written direct testimony 13 A. So pop-up channels are also typically, 14 discusses what you call artist's channels. 14 but not always, artist-based channels that we make 15 Can you describe for us what an 15 available on a more limited run. So while our -- 16 artist's channel is? 16 our existing full-time artist channels, you know, 17 A. Sure. 17 have been on the air on SiriusXM in most cases for 18 An artist's channel is a channel that 18 multiple years, the pop-up channels are similarly 19 we produce or coproduce in certain cases that 19 programmed, but they typically last anywhere from 20 either plays music from one artist and only their 20 three to, maybe on the high end, 30 days long. 21 music, such as our Elvis radio channel plays only 21 Q. And why does SiriusXM create pop-up 22 music from Elvis Presley. 22 channels for artists?

983 985

1 And then there are other artist's 1 A. We typically create them because we -- 2 channels that we offer that we coproduce with these 2 again, it's more content that is unique to 3 artists that plays a lot of that artist's music, 3 satellite radio. In most cases, they're actually 4 but also plays other music from similar type 4 created -- the creation of them comes out of 5 artists. 5 discussions that we have with the artist's manager 6 Q. Why does SiriusXM have artist-specific 6 and their record company who are typically seeking 7 channels? 7 this sort of exposure to generate more promotional 8 A. We have artist-specific channels 8 awareness for that artist during a given time 9 because it's just something else that we do that we 9 and -- and -- and doing that with the intention of 10 feel provides more exclusive content to our 10 hopefully either selling more concert tickets or 11 listeners that you can't get elsewhere. 11 selling more music. 12 Q. And in your experience, why would an 12 Q. Looking at Paragraph 51 of your written 13 artist want to participate in an artist's channel? 13 direct testimony, which is on Page 20, you discuss 14 A. Our artists are willing to participate 14 there the waivers that some artists have granted to 15 in coproducing these types of channels for two 15 SiriusXM. 16 reasons: one is they see it as another creative 16 Can you explain what those waivers are 17 vehicle for themselves to express themselves; and 17 for? 18 the other is they see the promotional value of 18 A. Yeah. Those waivers are -- waive the 19 having their own radio station on a national 19 performance complement of the Digital Millennium 20 satellite network and the -- the promotional 20 Copyright Act, which -- 21 benefits that come along with that. 21 JUDGE WISNIEWSKI: I'm sorry. Where 22 Q. And what do you mean by "promotional 22 are we at?

986 988

1 MS. PERELMAN: We're on Paragraph 51, 1 channels and present them to our subscribers, that 2 which is on Page 20. 2 we produce these channels not just with 3 JUDGE WISNIEWSKI: Fifty-one, thank 3 contemporary artists, but artists who have been 4 you. 4 around for a long time whose careers are probably 5 THE WITNESS: So those waivers waive 5 rooted back in the '60s and '70s. But they, just 6 the performance complement of the DMCA, and -- 6 as well as the contemporary artists, see the 7 which allows us to play more music from a -- a 7 promotional value from having these sorts of pop-up 8 particular artist within a given period of time. 8 channels on our platform. 9 BY MS. PERELMAN: 9 Q. You state in your written direct 10 Q. Just to illustrate, if you turn to the 10 testimony and you've stated here today that record 11 tab that says Direct Exhibit 40 attached to your 11 companies acknowledge that SiriusXM is a positive 12 testimony, are these a sampling of those kinds of 12 impact on sales and that they aggressively seek 13 waivers that have been executed? 13 airplay. 14 A. Yes. 14 What's the basis for those statements? 15 Q. In your experience, why do artists -- 15 A. The basis for those statements is, you 16 JUDGE WISNIEWSKI: Can I stop you one 16 know, a few things, but, you know, primarily, just 17 second, Counselor? 17 the day-to-day discussions that we have with 18 MS. PERELMAN: Oh, excuse me, I'm 18 artists' managers, record companies whose employees 19 sorry. 19 whose job it is to -- to -- to seek airplay on 20 JUDGE WISNIEWSKI: So you're -- 20 radio, as well as the artists themselves and in our 21 basically, the artist is granting a waiver from the 21 communication with them, they tell us verbally and 22 performance complement, right? 22 sometimes on e-mail that, you know, they value the

987 989

1 THE WITNESS: Yes. 1 airplay on SiriusXM and often cite the subsequent 2 JUDGE WISNIEWSKI: But are they 2 record sales that have come from the airplay on 3 granting a waiver for the payments of royalties for 3 SiriusXM. 4 the extra play? 4 Q. Is there anything beyond the day-to-day 5 THE WITNESS: No. No. 5 interactions that give you a sense of what the 6 JUDGE WISNIEWSKI: Thank you. 6 artists' or record labels' view is on that? 7 BY MS. PERELMAN: 7 A. I'm sorry. Can you repeat that? 8 Q. In your experience, why do artists and 8 Q. Sure. 9 recording companies grant these waivers? 9 Is there anything other than the 10 A. They granted these waivers because they 10 day-to-day interaction and direct feedback that 11 recognize, by our ability to play more music from 11 you get from them that gives you an impression of 12 that particular artist, it's going to subsequently 12 what the record labels' view is of airplay on 13 lead to more record sales. 13 SiriusXM? 14 Q. And just turning now to Paragraph 58 of 14 A. Well, there's -- there's also -- it was 15 your testimony, which is on Page 23, you discuss 15 discussed a little earlier -- promotional 16 there a couple of pop-up channels that were created 16 advertisements that record labels place in industry 17 for artists from the 1960s and '70s. 17 trade publications that often cite SiriusXM in a 18 Why do you focus here specifically on 18 prominent position and cite the airplay on SiriusXM 19 these more legacy artists' requests for pop-up 19 and often will even cite the subsequent sales that 20 channels? 20 came from that airplay. 21 A. I just thought it would be best to 21 Q. We were just looking at it a little 22 fully represent that when we do these pop-up 22 while ago, but if you flip again to Exhibit 36, are

990 992

1 those a sampling of those kinds of advertisements? 1 Q. Has there been any change over time in 2 A. Yes, they are. 2 the level or number of promotional contacts that 3 Q. And what's your understanding as to the 3 SiriusXM receives from these record companies? 4 reason why the record companies advertise the fact 4 A. Yeah. 5 that a particular track has been played on 5 I mean, we've always had relationships 6 SiriusXM? 6 with the record labels across all genres of music, 7 A. Well, SiriusXM's music channels have 7 you know, since as long as I've been at the 8 become rather influential over the years, which I 8 company. But over the last several years, those 9 think is a result of our programming. But as we've 9 relationships have strengthened, and the amount of 10 grown our subscriber base, the channels have become 10 contact that we have with record company employees 11 much more influential. 11 and particularly the radio promotional executives 12 And the record companies recognize how 12 that work in these companies has increased rather 13 influential the stations are that we offer, and 13 dramatically. 14 they will then include them in these advertisements 14 And they watch what we do very 15 to potentially influence other programmers in 15 carefully, literally sometimes on an hour-by-hour 16 America to play the songs that they're marketing. 16 basis. So if we make a change in a particular 17 Q. And if you just turn to the very last 17 hour, there are times I will get a phone call and 18 page of Exhibit 36. 18 say, hey, Steve, what happened to this song? We 19 Is that another example of a record 19 were just playing on Hits 1. We noticed it's not 20 label describing the plays on SiriusXM? 20 being played anymore. What happened? 21 A. Yeah. This is for an artist called 21 So they watch us really, really 22 Plug In Stereo who are on Atlantic Records, which 22 closely.

991 993

1 is part of the Warner Music Group. And you can see 1 Q. Mr. Blatter, do you have an 2 highlighted in yellow in the exhibit that Atlantic 2 understanding as to what effect, if any, SiriusXM 3 was communicating to other programmers that the 3 airplay has on the sale of sound recordings? 4 single had already sold over 100,000 songs or 4 A. Yes, I do. 5 singles after just 600 plays on Sirius Hits 1, 5 Q. What is that view? 6 which is one of our contemporary pop channels. 6 A. The -- airplay received on SiriusXM is 7 Q. And you mentioned that you receive 7 highly correlated with the sale of the music that 8 feedback directly from labels and artists. We're 8 we play. 9 not going to go through them in the interest of 9 Q. And what supports that view? 10 time. 10 A. Well, besides being told that by the 11 But just if you flip towards the back 11 record industry employees, artists' managers and 12 of your book, if you look at Exhibits 42 through 12 the artists themselves, we also do subscribe to 13 45, 47 and 50, are those a couple of examples of 13 SoundScan, which provides us with sales data. It's 14 that kind of feedback that you've received? 14 the, you know, industry resource that both -- 15 A. Yeah, this is just some of the e-mail 15 primarily record companies use, and we use it at 16 communication that we've received through record 16 SiriusXM as well, to gauge the impact that our 17 label executives that are, in most cases, you know, 17 airplay is having on the sale of music. 18 very clearly praising the -- SiriusXM or the 18 Q. And did you go to any source to track 19 programmer who they sent the e-mail to, and then 19 the airplay of certain songs on SiriusXM and 20 citing the, in most cases, pretty dramatic increase 20 elsewhere? 21 in record sales that came as a direct result of the 21 A. Yeah. I mean, at SiriusXM, we use a 22 airplay on that particular channel. 22 service called Mediabase. There's another one

994 996

1 called BDS. Those are the two primary firms that 1 this. We just cited a handful of them in my 2 monitor airplay of radio stations in America. 2 testimony. 3 Q. Okay. And just going back to 3 Q. We're not going to go through each of 4 SoundScan, what data specifically does SoundScan 4 these examples, but just for illustration and to 5 show? 5 help us learn how to read the data that's attached 6 A. SoundScan specifically shows sales 6 to it, if you could turn to Paragraph 71, which is 7 data, both digital track sales and album sales 7 on Page 28. 8 for -- for artists. 8 Here, you talk about the song Dancing 9 Q. And can you track that data on a weekly 9 Shoes by the band Green River Ordinance. 10 basis? 10 A. Yes. 11 A. We have the ability at SiriusXM to 11 Q. Which of SiriusXM's music channels 12 track it weekly. They might offer it -- services 12 first played Dancing Shoes? 13 where you can do it more frequently than that. But 13 A. That would be The Highway. 14 we -- we monitor or receive data from SoundScan on 14 Q. What kind of music does The Highway 15 a weekly basis. 15 play? 16 Q. Does SiriusXM use SoundScan and 16 A. Country music, primarily. 17 Mediabase data in the ordinary course? 17 Q. If we wanted to see the date on which 18 A. Yes, we do. 18 The Highway first played Dancing Shoes, what source 19 Q. For what -- for what purposes? 19 would we go to? 20 A. Well, we use sales data now as another 20 A. Well, the industry source we would use 21 means to give us an indication whether a song is 21 for that would be Mediabase. 22 connecting with an audience. So sales data -- we 22 Q. For this particular song, is the

995 997

1 also do some other proprietary research, but 1 Mediabase data attached as Exhibit 53 to your 2 increasingly so over the last several years, sales 2 testimony? 3 data is now being used as a programming tool for 3 A. Yes. 4 myself and the programmers that I employ to help 4 Q. And can you tell from this data the 5 gauge the success or sometimes, you know, the 5 first date on which The Highway started playing 6 nonsuccess of a particular song. 6 Dancing Shoes? 7 Q. So, Mr. Blatter, if you would turn to 7 A. Yeah. If you look all the way to the 8 Paragraphs -- or Pages 25 through 30, generally. 8 bottom right, you will see that the song was first 9 You discuss a couple of examples of certain songs. 9 played, according to Mediabase, on October 24th, 10 What are these examples intended to 10 2011. 11 illustrate? 11 Q. And was -- according to this data, was 12 A. These are examples of songs that were 12 any other station in the country playing that song 13 being played exclusively on SiriusXM. And we then 13 at that time? 14 tracked the -- our airplay data, matched it up 14 A. No. If there were other stations 15 against the SoundScan sales data, and were then 15 playing it, they would be listed below The Highway 16 able to demonstrate the high correlation between 16 in this document. 17 the airplay and the records sales. 17 Q. And just to confirm, is the rightmost 18 Q. Are these couple of examples that are 18 column in this -- in this page, does that show the 19 discussed on 25 to 30, are those the only such 19 number of spins on that -- or -- what is a spin? 20 examples that you can think of? 20 A. A spin is the play of a song. Each 21 A. No. These are just actually a small 21 play would be a spin. 22 handful. There are numerous other stories like 22 Q. Does that indicate that The Highway

998 1000

1 had -- had played that song 64 times during the 1 says Total, and you track it across to the week 2 week that the data reflects? 2 ending 23rd? 3 A. That's actually 64 times to date. 3 A. Yes. 4 Since the record was first played, it played 64 4 Q. Okay. And then the week during which 5 times. 5 SiriusXM picked up Dancing Shoes and was the only 6 Q. Thank you. 6 radio station playing it in the country, that would 7 If you just turn to the second page of 7 have been the week ending October 30th; is that 8 Exhibit 53. 8 right? 9 Can you tell us what this page 9 A. That's correct. 10 reflects? 10 Q. Because it picked it up on the 24th, 11 A. This is a similar report that was just 11 right? 12 run a few days later. So it looks like the initial 12 A. Yes. 13 report was run on November 3rd, and this report was 13 Q. So can you tell us how many digital 14 started on November 7th. 14 tracks of Dancing Shoes were sold in that first 15 Q. And it shows an increased number? 15 week that SiriusXM picked up the song? 16 A. So you can see the spins had gone up 16 A. So the sales for that digital track 17 from 64 times at the -- at the time the previous 17 after The Highway started playing the song leaped 18 report was done to 112 spins a few days later. 18 from 58 tracks the previous week to 1,105 digital 19 Q. Was any other radio station in the 19 tracks for the week ending October 30th. 20 country playing Dancing Shoes at that time? 20 Q. And just to confirm, no other radio 21 A. No. 21 station was playing that at that time, right? 22 Q. Now, if we wanted to track the digital 22 A. No, there was not.

999 1001

1 track sales for that -- for these weeks, where 1 Q. Just to carry it out to the end, the 2 would we go next? 2 following week in the second week of airplay of 3 A. We would then look at the corresponding 3 that song, how many digital tracks were sold? 4 SoundScan data. 4 A. 1,426. 5 Q. Which is attached to your testimony as 5 And I think it's important to just show 6 Exhibit 54; is that correct? 6 here as well that you can see that the sales of 7 A. Yes, it is. 7 this track are actually spread out across the 8 Q. If you could help us learn how to read 8 country. It's not, you know, really designated to 9 this, which is very small and I apologize. 9 any one particular market, which I think shows the 10 If we wanted to look at the digital 10 national reach that SiriusXM has. 11 track sales data for Dancing Shoes in the week 11 Q. If you flip back to the text of your 12 before SiriusXM picked it up, what column would we 12 written direct testimony on Page 29, you'll see a 13 look at? 13 graph. 14 A. That would be the week ending 14 I'll let you get there. 15 October 23rd, 2011. 15 Page 29. 16 Q. Okay. And can you tell how many total 16 A. Okay. 17 digital tracks of Dancing Shoes were sold during 17 Q. What does that graph illustrate? 18 the week ending October 23rd, 2011? 18 A. This is just a graphical depiction of 19 A. Sure. 19 the sales data for the song Dancing Shoes from 20 On a national basis, 58 digital tracks 20 Green River Ordinance. And you can see on the 21 were sold of that song. 21 bottom there -- you can see, previous to the week 22 Q. And you can see that from the row that 22 of the 10/24/2011, the sales were, you know, in

1002 1004

1 the -- probably, you know, 50- to maybe 1 manner hurting their sales. In fact, it's always 2 200-units-a-week range. 2 actually the exact opposite of that. 3 And then, when The Highway added it on 3 MS. PERELMAN: Thank you, Mr. Blatter. 4 October 24th, you can see the hockey sticklike 4 No further questions. 5 reaction where the sales, you know, skyrocket in 5 CHIEF JUDGE BARNETT: Mr. Moskowitz. 6 the weeks to follow based on the airplay from 6 7 The Highway. 7 CROSS-EXAMINATION 8 Q. Has the band Green River Ordinance had 8 9 anything to say about its experience on SiriusXM? 9 BY MR. MOSKOWITZ: 10 A. They have. 10 Q. Good morning, Mr. Blatter. 11 Q. What have they had to say about it? 11 A. Good morning. 12 A. They've been actually very vocal. They 12 Q. My name is David Moskowitz. I 13 even took to recording a video, which they sent to 13 represent SoundExchange. I have a few questions to 14 our programmers after they saw this amazing 14 ask you this morning. 15 reaction that this song was having after the 15 Could you please turn to Paragraph 12 16 airplay on The Highway. 16 of your written direct testimony? 17 In the video, they praise SiriusXM and 17 In Paragraph 12, you discuss SiriusXM's 18 had commented on what a remarkable thing this has 18 direct licenses, right? 19 been for their career. 19 A. Yes. 20 Q. We're not going to play it in the 20 Q. And you state that you've reviewed a 21 interest of time, but is that SiriusXM Exhibit 33, 21 list of direct license labels; is that correct? 22 and it's transcribed here in Paragraph 71 of your 22 A. That is correct.

1003 1005

1 testimony? 1 (SoundExchange Trial Exhibit Number 25 2 A. Yes. 2 was marked for identification 3 Q. Is this kind of feedback out of the 3 purposes.) 4 ordinary in your experience? 4 BY MR. MOSKOWITZ: 5 A. It's not out of the ordinary when 5 Q. Mr. Blatter, do you have in front of 6 you're having the sort of impact that SiriusXM is 6 you SoundExchange -- SoundExchange Trial 7 now having on -- on artists' careers and our 7 Exhibit 25? 8 ability to sell music for them and further -- 8 A. Yes, I do. 9 further their careers. 9 Q. Is that listed also as SiriusXM Direct 10 Q. Okay. Mr. Blatter, last question. 10 Exhibit 14? 11 Your testimony talks a lot about how 11 A. Yes. 12 people in the industry, artists and labels and 12 Q. Does this document contain the list of 13 their managers, have acknowledge the airplay is 13 direct license labels that you reviewed? 14 good for exposure, good for record sales. 14 A. This might not have been the same list 15 How about the opposite? 15 I reviewed when my testimony was written, but this 16 Has anyone ever come to you and said, 16 list does look familiar. 17 please stop playing my music, you're killing my 17 Q. When you reviewed the list of direct 18 sales? 18 license labels, were you familiar with all of the 19 A. No. That's -- in my 25 years, I've 19 labels? 20 never come across anybody in the industry, whether 20 A. Can you repeat that question? 21 it be record label executives, artists' managers or 21 Q. When you reviewed -- you stated earlier 22 the artists themselves that felt we were in some 22 that you reviewed a list of direct license labels.

1006 1008

1 When you reviewed that list, were you 1 Q. And in your testimony in Paragraph 12, 2 familiar with all of those labels? 2 you discuss L.A. Guns as having -- featuring 13 3 A. It's hard for me to recollect going 3 tracks in Hair Nation; is that correct? 4 back to November when I was reviewing the list. 4 A. Yes. 5 I did -- I'm sorry -- I did -- I was 5 Q. Does this document show the list of 6 familiar with many of the independent labels that 6 L.A. Guns' songs that you were discussing in your 7 were listed on the sheet, but not all. 7 testimony? 8 Q. Based on your review of the list, you 8 A. It appears to, yes. 9 determined that certain songs within the catalogs 9 Q. Do you know roughly how frequently 10 of some of the direct license labels were already 10 SiriusXM plays L.A. Guns each month? 11 featured on SiriusXM; is that correct? 11 A. I wouldn't know off the top of my head, 12 A. Correct. 12 no. They're an important artist to the channel. 13 Q. How did you determine those direct 13 Q. Are you aware that SiriusXM played 14 license label songs were already featured on 14 L.A. Guns over 200 times in August 2011? 15 SiriusXM? 15 A. I'm not aware of the specific number, 16 A. That information was not actually -- 16 but that wouldn't surprise me. 17 the information as to what the songs were that were 17 Q. Are you aware that during the fourth 18 directly licensed was -- was provided to us -- was 18 quarter of 2011, SiriusXM did not play a single 19 provided to me through the agent that we're working 19 directly licensed song from L.A. Guns? 20 with to work with us on these direct license deals. 20 A. I was not. 21 And then that information was then cross-referenced 21 Q. You were not? 22 with the artists and the songs that we play on our 22 A. I am not.

1007 1009

1 music channels. 1 Q. You are -- you still are not? 2 (SoundExchange Trial Exhibit Number 26 2 A. No. 3 was marked for identification 3 Q. Could you explain? You said you were 4 purposes.) 4 not. 5 BY MR. MOSKOWITZ: 5 Are you saying you were not at the 6 Q. Mr. Blatter, do you have in front of 6 time? 7 you SoundExchange Trial Exhibit 26, which is Bates 7 A. I'm not aware that the -- that we were 8 stamped SXM_CRB_DIR_00000142? 8 not playing directly licensed songs at that time. 9 A. Yes. 9 Q. But are you confident that SiriusXM did 10 Q. Do you know what MusicMaster is? 10 play L.A. Guns during the fourth quarter of 2011? 11 A. Yeah, MusicMaster is a tool that my 11 A. Well, I'm -- I'm -- I'm -- I'm basing 12 programmers use to assist them in scheduling the 12 what's in my testimony based on the data that was 13 music on the channels that they're responsible for. 13 provided to me by the firm that we're working with 14 Q. Is this document a printout from 14 to determine what songs are directly licensed. 15 SiriusXM's MusicMaster database -- or software? 15 Q. On this printout, you see a year 16 A. It appears to be. 16 listed, right? 17 Q. Can you explain what this printout 17 A. Um-hum. 18 shows? 18 Q. And that is the year that the song was 19 A. This printout appears to be a list of 19 released; is that correct? 20 songs from the band L.A. Guns that are in the 20 A. Typically, it's a song that is released 21 library for the channel that we have called 21 or, in certain cases, it's the year that the song 22 Hair Nation. 22 peaked on a chart at that time.

1010 1012

1 Q. Are you aware that the masters for each 1 Q. And are you aware that during the 2 of the original recording tracks listed in SiriusXM 2 fourth quarter of 2011, SiriusXM did not play any 3 Exhibit -- Trial Exhibit 26 are not, in fact, owned 3 versions of those songs that were owned by 4 by Cleopatra Records? 4 Cleopatra Records? 5 A. I am not. 5 A. That might very well be. Like I said, 6 (SoundExchange Trial Exhibit Number 27 6 when -- when the information was provided to us, we 7 was marked for identification 7 were -- we had the understanding that the songs 8 purposes.) 8 that are listed in this MusicMaster document were 9 (SoundExchange Trial Exhibit Number 28 9 directly licensed through Cleopatra. 10 was marked for identification 10 Q. In your role as general manager of -- 11 purposes.) 11 as a general manager of music programming, is part 12 BY MR. MOSKOWITZ: 12 of your responsibility to implement the increase of 13 Q. Mr. Blatter, do you have in front of 13 plays of direct license labels? 14 you SiriusXM Trial Exhibits 27 and 28, which are 14 A. I don't think we said we would increase 15 Bates stamped -- SiriusXM [sic] Trial Exhibit 27 15 plays. We said there's the potential for that if 16 has Bates Stamp SXM_CRB_DIR_000088882, and 16 all things are treated equal. 17 SiriusXM -- or SoundExchange Trial Exhibit 28, 17 Q. How do your programmers know that a 18 which is Bates stamped SXM_CRB_DIR_000088668? 18 song is directly licensed? 19 A. Yes. 19 A. So we're actually in that process right 20 Q. Are you familiar with these documents? 20 now of working with our third party to match the 21 A. No. This is the first time I've seen 21 data that's in MusicMaster against the information 22 these documents. 22 that's going to be provided to us that shows what

1011 1013

1 Q. Did you request anything from your 1 songs that we have in our systems that are directly 2 agent regarding the plays of L.A. Guns from 2 licensed. That process has been underway for -- 3 Cleopatra Records? 3 for several weeks and we're hoping, actually, in 4 A. No, no. The airplay data was procured 4 the next few weeks that we'll -- we'll -- we'll 5 by my team at SiriusXM, not by an outside party. 5 finish that process. 6 Q. So did your team examine which label 6 Q. In Paragraph 12, you highlighted 7 owned the recordings that you researched? 7 four -- four particular labels that are -- their 8 A. No. Like I said, we were informed by 8 songs are featured on SiriusXM; is that correct? 9 our agent that those songs were directly licensed 9 A. Yes. 10 through Cleopatra. 10 Q. Are you aware that for the four labels 11 Q. So it was your opinion -- or your -- 11 you highlighted, SiriusXM played a combined total 12 strike that, please. 12 of roughly 1300 transmissions in the fourth quarter 13 It was your understanding that 13 of 2011? 14 Cleopatra Records owned all of the L.A. Guns songs? 14 A. I wouldn't -- I wouldn't know that 15 A. That's correct. 15 number offhand and have never, you know, calculated 16 Q. And was it -- in Paragraph 12, you 16 that number myself. 17 reference two additional songs from Cleopatra, Turn 17 Q. Are you aware that SiriusXM transmitted 18 Up the Radio by Autograph and Seventeen by Winger. 18 about 4.2 million plays during the fourth quarter 19 Was it also your understanding that 19 of 2011? 20 Cleopatra owned all of the recordings of those 20 A. I would have no reason to really be 21 songs? 21 calculating a number like that myself for what I'm 22 A. Yes, it was. 22 responsible for.

1014 1016

1 Q. So you're not aware that for those four 1 A. I'm sorry. Can you say that again? 2 labels, their plays amounted to 0.03 percent of 2 Q. Paragraph 61 to 65. 3 your transmissions during the fourth quarter of 3 -- you offer what you call testimonials 4 2011? 4 from five labels as to the promotional benefits of 5 A. I -- I don't -- I don't really know 5 SiriusXM, right? 6 what relevancy that number is. I mean, we program 6 A. I wouldn't refer to them as 7 SiriusXM not as a -- as an entire platform; we 7 testimonials, no. 8 program our channels on a channel-by-channel 8 Q. Communications from the five labels? 9 basis. 9 A. Sure. 10 So there are artists that are listed 10 Q. Sure. 11 here (indicating) that are key artists for the 11 Isn't it true that none of these labels 12 channels that play them. And -- and -- and I think 12 have signed direct license deals with SiriusXM? 13 that's -- that's, you know -- you know, worth 13 A. That could be. We're in the process 14 noting. 14 of -- you know, from what I understand, we're in 15 I mean, there's a channel we have 15 the direct licensing process. So I'm not involved 16 called Jam_ON, where a band that's cited here, 16 in it directly, so it could very well be, between 17 String Cheese Incident -- I mean, String Cheese 17 the time this testimony was written and today, a 18 Incident is a core artist to that channel and one 18 direct license was signed with one of these labels. 19 of the most important artists that we can play on a 19 I might not know that as of today. 20 channel like Jam_ON. 20 Q. As of the time of your testimony, had 21 Q. Are you aware that for this String 21 SiriusXM signed any of those labels to direct 22 Cheese Incident label, that that label was played a 22 licenses?

1015 1017

1 total of less than 400 times during the entire 1 A. No, but nor would I expect them to be. 2 fourth quarter of 2011? 2 When my testimony was done, we had just 3 A. I don't know what relevancy that is or 3 started the direct licensing process. And from 4 what context, you know -- that -- I can tell you 4 what I understand about the process, it's a long 5 that String Cheese Incident is -- is a major act 5 race and -- and we have maybe gone around the track 6 for us on Jam_ON, one that we actually use in 6 one time in a 500-lap race. 7 describing the channel. 7 So it's wouldn't be all that surprising 8 We describe all of our channels by 8 if these -- if these examples cited here were not 9 five, six, seven artists. We may play hundreds of 9 directly licensed at that time. 10 them on that channel. String Cheese Incident is so 10 Q. In Paragraphs 66 to 73, you provide 11 important to us that it's one of the five or six 11 four examples that you say purport to show 12 artists we use to define Jam_ON. That's how 12 empirical evidence of promotional benefits, 13 important they are to us. 13 correct? 14 Q. But you just don't play them? 14 A. Correct. 15 A. No. We do play them. 400 times is a 15 Q. Isn't it true that as of the time of 16 lot of space. 16 your testimony, SiriusXM has not directly licensed 17 Is that what you just said? I wouldn't 17 any of the four tracks that you discussed? 18 know offhand, but if you say 400 times, that's a 18 A. To the best of my knowledge, these have 19 lot of play for one artist. 19 not been directly licensed, but I'm really not sure 20 Q. That's for the entire label. 20 what relevancy that would be. 21 But anyway, moving to -- in 21 It's just showing -- this was put here 22 Paragraph 61 to 65 -- 22 to demonstrate the -- the promotional value of

1018 1020

1 SiriusXM and how it leads to more record sales. 1 Q. How did you find this study? 2 Q. In your view, your primary objective is 2 A. You know, I don't actually recall where 3 to satisfy the listener; is that correct? 3 we came across this study offhand. 4 A. Yes. 4 Q. Who provided you this study? 5 Q. If a direct license label has a 5 A. I don't recall. 6 recording that doesn't fit on SiriusXM or that 6 Q. You used the term "we." 7 would reduce listening satisfaction, you wouldn't 7 Who are you referring to with "we"? 8 play that label, right? 8 A. I'm sorry. Which -- how did I use the 9 A. We would not. Not that label, but 9 word "we"? 10 maybe that particular song on that label. 10 Q. You said, I don't know how we found 11 Q. Just briefly in Paragraph 51, you 11 this study. 12 mentioned that labels sometimes grant performance 12 A. So -- well, in preparing my testimony, 13 complement waivers to SiriusXM? 13 when there's research -- and I did have the 14 A. Yes. 14 assistance of people who work with me on my staff 15 Q. And you testified that those waivers 15 to identify, you know, documents that I thought 16 don't include waivers of the royalties; is that 16 could be useful for my testimony. 17 correct? 17 Q. Did you search for other studies? 18 A. Correct. 18 A. Did I search for other studies? 19 Q. And so when these record labels agree 19 Q. Um-hum. 20 to allow you to play more of their songs, they 20 A. Meaning outside of this one here? 21 receive more royalties? 21 Q. Yes. 22 A. I would think that would be the case, 22 A. Yes.

1019 1021

1 yeah. 1 Look, as part of my job, I'm constantly 2 Q. Could you turn to Paragraph 47? 2 looking at -- at, you know, whatever sources, you 3 A. Forty-seven, you said? 3 know, I think would be useful for me to be 4 Q. Yes. 4 reviewing and learning from to do the job to the 5 In Paragraph 47, you assert that the 5 best of my abilities. 6 simple and universally recognized fact is that 6 Q. Isn't it true the NAB commissioned this 7 airplay on radio has continually proven to be the 7 study to fight against the enactment of a 8 biggest driver of record sales, right? 8 performance fee on terrestrial radio? 9 A. Correct. 9 A. I -- I couldn't tell you what the 10 Q. And you cite two studies for this point 10 background was on this study. I don't know. 11 in Footnote 5? 11 Q. Could you turn to Page 5, please? 12 A. Yep. Yes. 12 And the last paragraph reads, As a 13 Q. Would you turn to SiriusXM Direct 13 record industry advocates for the direct payment 14 Exhibit 37 that was attached to your testimony? 14 from radio stations to music labels and artists 15 A. Okay. 15 through a new performance fee, it should be noted 16 Q. Is this a copy of the first study 16 that disturbing the current symbiotic relationship 17 listed in Footnote 5? 17 that is found to exist between radio and the record 18 A. Yes. 18 industry could actually harm, not help, all 19 Q. Have you read this study previously? 19 parties. 20 A. I have. 20 A. Okay. 21 Q. Who commissioned the study? 21 Q. But you have no idea if this is -- 22 A. The study was commissioned by the NAB. 22 A. This is just a statement. It doesn't

1022 1024

1 state that's why the study was conducted. 1 you're saying? 2 Q. Could you turn -- actually -- does this 2 Q. Page 74? 3 study explicitly acknowledge that there is actually 3 A. Page 74. 4 conflicting evidence on whether airplay increases 4 Q. And do you see that there is one 5 music sales? 5 Liebowitz 2007 article listed in the references, 6 A. I'm sorry. Can you restate that? 6 correct? 7 Q. Does this study explicitly acknowledge 7 A. Actually, it looks like there's two 8 that there is conflicting evidence on whether 8 Liebowitz articles. 9 airplay increases music sales? 9 Q. From 2007? 10 A. I don't specifically remember that in 10 A. Yes. 11 reading through the study. 11 Q. Yes, there's one. 12 Q. Could you turn to Page 11, please? 12 And does that -- is the -- 13 Reading a couple of sentences in, it 13 SoundExchange Trial Exhibit 29, does that appear to 14 says, Again, conventional wisdom is that radio 14 be the article referenced -- the Liebowitz 2007 15 airplay stimulates record sales. This belief is 15 article referenced in this study? 16 consistent anecdotal evidence, including the fact 16 A. It appears to be based on the cover 17 that record labels pay large sums to promote their 17 page. 18 releases and garner radio airplay. Unfortunately, 18 Q. Have you reviewed this study 19 there has been previous -- precious little 19 previously? 20 scholarly research on this topic, and the two most 20 A. This Liebowitz study? 21 recent contributions to the literature have 21 Q. Yes. 22 contrasting conclusions. 22 A. I have not.

1023 1025

1 And at the bottom, it goes on and says, 1 Q. Did you do anything to search for 2 A second more recent study by Liebowitz examines 2 studies against what your belief was? 3 aggregate sales of albums in the top 100 designated 3 A. I've -- I've been searching for studies 4 market areas. He examines changes between 1998 and 4 my entire career that -- that show the correlation 5 2003 in album sales and estimates the impact of 5 between radio airplay and record sales. I've never 6 changes in Arbitron ratings for the stations with 6 come across a study that showed the negative impact 7 music formats. 7 that you're stating this study evidently shows. 8 In contrast to the Montgomery and Moe 8 Q. But it is actually referenced in the 9 findings, Liebowitz finds a large negative effect 9 article that you said you read and saw, and you 10 at the industry level. 10 couldn't find it? 11 Mr. Blatter -- 11 A. So I reviewed this research study and 12 (SoundExchange Trial Exhibit Number 29 12 saw those conclusions. I did not go through every 13 was marked for identification 13 reference and every study that's mentioned as a 14 purposes.) 14 reference in this study. 15 BY MR. MOSKOWITZ: 15 Q. Doesn't -- if you just look at the 16 Q. Mr. Blatter, do you have before you 16 abstract of Mr. Liebowitz' article, he says, This 17 SoundExchange Trial Exhibit 29? 17 paper undertakes an econometric investigation of 18 A. Yes, I do. 18 the impact of radio play on sales of sound 19 Q. And if you turn to -- I'm sorry to go 19 recordings using a sample of American -- 20 back -- Page 74 of the study that you referenced in 20 MS. PERELMAN: Objection, Your Honor. 21 SXM Direct Exhibit 37? 21 This document is not in evidence, and 22 A. I'm sorry. What section was that 22 counsel is reading it directly into the record.

1026 1028

1 MR. MOSKOWITZ: I'm not reading into 1 Q. The studies that you cited in your 2 the evidence for the truth; I'm reading it into the 2 testimony are concerning terrestrial radio; is that 3 evidence to impeach Mr. Blatter's statement that it 3 correct? 4 is universally recognized fact, based on his 4 A. It depends what studies. 5 research, that radio airplay increases ad sales -- 5 I mean, you know, how are you defining 6 or music sales. 6 a study. I mean, there's a study that's very 7 CHIEF JUDGE BARNETT: Overruled. 7 specific not just to satellite radio, but specific 8 BY MR. MOSKOWITZ: 8 channels on satellite radio. So be more specific 9 Q. The results indicate that radio play 9 if you can, please. 10 does not have the positive impact on record sales 10 Q. I'm specifically referring to the two 11 normally attributed to it and, instead, appears to 11 studies you cited in Footnote 5, SiriusXM Direct 12 have an economically important negative impact 12 Exhibit 37 and SiriusXM Direct Exhibit 38. 13 implying that overall radio listening is more of a 13 Are these not studies about terrestrial 14 substitute for the purchase of sound recordings 14 radio? 15 than it is a complement. 15 A. Those two studies might not have -- 16 It goes on at the bottom. It says, 16 might not have been specific as to the type of 17 This research also exposes a fallacy of composition 17 radio in which they were researching. 18 in applying to an entire market a generally 18 Q. They might not have been? 19 accepted positive relationship that holds for 19 A. I mean, I can go back through and be 20 individual units. 20 more specific for you, but they might not have 21 Is it still your view that it is a 21 mentioned specifically satellite radio. 22 universally accepted fact that radio airplay 22 I think what I said was that radio

1027 1029

1 increases music sales? 1 airplay is -- it's universally accepted that radio 2 A. Absolutely. 2 airplay leads to more record sales. 3 And I will tell you since we started 3 Q. Could you turn to Paragraphs 66 to 73? 4 subscribing to SoundScan at SiriusXM a year ago, I 4 CHIEF JUDGE BARNETT: Are you referring 5 believe that more than I ever have in my 25 years 5 to paragraphs in the testimony? 6 in radio. 6 BY MR. MOSKOWITZ: 7 Q. But apparently, Mr. Liebowitz disagrees 7 Q. I'm sorry. Back in your testimony, 8 with you; is that correct? 8 please. 9 A. So I don't know who Mr. Liebowitz is. 9 Here, you purport to show anecdotal 10 I don't know this study. I've never seen this 10 empirical evidence of the impact of SiriusXM on the 11 study before. So I can't really speak to it. 11 sale of four songs; is that right? 12 I can speak to my experience of 12 A. You're starting at Paragraph 66? 13 25 years in the radio industry working very, very 13 Q. Yes. 14 closely, intimately, with the artists and the songs 14 A. Yes. 15 that we play and the representatives of those 15 Q. In your analyses, you examined only 16 artists at the record companies and their managers 16 terrestrial and satellite radio airplay, right? 17 and the artists themselves. And they all tell me 17 A. The analysis might have been inclusive 18 the same thing: radio airplay leads to record 18 in some other channels that are not satellite and 19 sales. 19 terrestrial airplay, such as, you know, the -- 20 Q. Am I right that these studies you have 20 Mediabase also monitors video play on MTV and some 21 cited do not concern satellite radio? 21 other video sources. 22 A. What studies are you referring to? 22 And I think it actually includes some

1030 1032

1 Music Choice channels as well. 1 communication -- e-mail communication that's in 2 Q. But, for example, you never looked at 2 here (indicating) where our airplay is cited and -- 3 plays on noninteractive Internet radio? 3 and we're getting communication back from the 4 A. We follow closely what happens on -- in 4 record company executive, it's pretty clear that 5 Internet radio, as do the -- you know, our industry 5 what they're seeing is the airplay on SiriusXM, and 6 contacts at record labels and managers. And based 6 the subsequent record sales is a direct relation to 7 on their desire and others, those Internet radio 7 our airplay. 8 outlets are not factored into Mediabase airplay or 8 Q. Mr. Blatter, you're not aware of any 9 BDS airplay for that matter as well. 9 empirical evidence as the promotional or 10 Q. So the answer is you did not; is that 10 substitutional effect of SiriusXM on overall music 11 correct? 11 sales, right? 12 A. We did not, but we did not because the 12 A. I've never -- I've never heard of -- of 13 reason they're not included in Mediabase or BDS is 13 radio airplay being a substitute for record sales 14 because the industry as a whole has determined that 14 anywhere. 15 the airplay on those Internet outlets at this point 15 Q. And you're not aware of any empirical 16 has not been demonstrated to have any impact -- any 16 evidence as to whether SiriusXM is more or less 17 meaningful impact on record sales. 17 promotional or substitutional than other music 18 Q. And you've also never looked at plays 18 services, right? 19 on interactive on-demand services; is that correct? 19 A. Can you repeat that? 20 A. I would give you the same response. 20 Q. You're not aware of any empirical 21 And I asked this question rather pointedly in my 21 evidence as to whether SiriusXM is more or less 22 discussions with record label executives and 22 promotional or substitutional than other music

1031 1033

1 artists' managers. 1 services? 2 And I ask them regularly: Are you 2 A. I think it depends how you want to 3 seeing any impact from some of these interactive 3 define empirical. I think the exhibits that I show 4 services on your artists' careers? Are you seeing 4 here with the examples of the Mediabase airplay 5 any record sales come from these interactive 5 data and the subsequent SoundScan airplay data 6 services? And I get the same answer every time I 6 is -- to me, that's empirical data, and it shows 7 ask that question. It's no, we're not seeing any 7 that correlation. 8 impact whatsoever today. 8 And it's proven out by the record 9 Q. So your analyses just assume there's no 9 company executives who sent us this communication 10 impact based on your anecdotal information? 10 saying, hey, appreciate the airplay, look at these 11 A. I wouldn't -- I think that goes beyond 11 record sales as a result of it. 12 anecdotal. So when I'm asking some of the most 12 MR. MOSKOWITZ: Thank you. No further 13 respected figures in the music industry and the 13 questions. 14 artists themselves, are you seeing any resulting 14 CHIEF JUDGE BARNETT: Thank you, 15 airplay? 15 Mr. Moskowitz. 16 Are you seeing any -- any -- any impact 16 Pardon me, Counsel for Music Choice -- 17 of the exposure you might have gotten on an 17 I'm sorry -- introduce yourself again today. 18 interactive service? Are you seeing more record 18 MR. FAKLER: Sure, Your Honor. 19 sales as a result of that? And the response I 19 Paul Fakler of Arent Fox. 20 always get is no, not yet, at least not today. 20 CHIEF JUDGE BARNETT: Mr. Fakler, any 21 I just want to mention if -- if that 21 questions for this witness? 22 was the case, I think if you look at some of the 22 MR. FAKLER: No, Your Honor.

1034 1036

1 CHIEF JUDGE BARNETT: Ms. Perelman, 1 studies -- referring to the Montgomery and Moe 2 redirect? 2 study and the Liebowitz study -- Although both 3 MS. PERELMAN: I'll be very brief. 3 studies suffer from certain flaws, the Montgomery 4 4 and Moe study is the more reliable of the two. The 5 REDIRECT EXAMINATION 5 data errors and methodological choices made in the 6 6 Liebowitz study are problematic. He does not 7 BY MS. PERELMAN: 7 adequately account for population and/or audience 8 Q. Mr. Blatter, Mr. Moskowitz was 8 distributions across DMAs or station coverage 9 directing you to the portion of your testimony and 9 areas. Further, the Liebowitz approach is inferior 10 the attached exhibit at Exhibit 37, which is, for 10 under the wide range of conditions likely to 11 the record, an article called Radio Airplay and 11 prevail with the results dubious because of some 12 Record Industry: An Economic Analysis. 12 unfounded assumptions about the pattern of 13 Do you recall that? 13 regression errors. 14 A. Yes. 14 Did you read that portion of the 15 Q. And Mr. Moskowitz pointed you to the 15 article when you attached it to your testimony? 16 portion of that exhibit that references a study by 16 A. Yes. 17 someone by the name of Liebowitz; is that right? 17 MS. PERELMAN: Nothing further. 18 A. Yep. 18 MR. MOSKOWITZ: No further questions, 19 Q. And that study examined changes between 19 Your Honor. 20 1998 and 2003 in album sales, according to the 20 CHIEF JUDGE BARNETT: Judge? 21 article, right? 21 JUDGE WISNIEWSKI: No. 22 A. Yes. 22 CHIEF JUDGE BARNETT: May this witness

1035 1037

1 Q. In your view, is the impact of record 1 be excused? 2 sales now in any way different from the impact 2 MS. PERELMAN: Yes, Your Honor. 3 of -- on record sales during that period as a 3 THE WITNESS: Thank you. 4 result of airplay on SiriusXM? 4 CHIEF JUDGE BARNETT: Thank you, 5 A. Well, if this study was citing 1998 to 5 Mr. Blatter. 6 2003, neither Sirius or XM were really on the air 6 I think this is a good time for us to 7 at that point. We might have just come to market 7 take our morning recess, and we will do so. 8 in 2003, but it was at the very earliest stages of 8 (Whereupon, a brief recess was taken 9 our -- of our existence. 9 from 10:55 a.m. to 11:07 a.m.) 10 Q. And Mr. Moskowitz was directing you to 10 CHIEF JUDGE BARNETT: Good morning. 11 portions of the Liebowitz study, but in reviewing 11 Please be seated. 12 what's attached to your testimony as Exhibit 37 and 12 Mr. Larson. 13 deciding to attach it to your testimony, did you 13 MR. LARSON: Thank you. 14 also -- and Mr. Moskowitz was referring you to 14 SiriusXM calls as its next witness 15 portions on Pages 11 and 12 of Exhibit 37 -- did 15 William Rosenblatt. 16 you also continue to read into the following 16 WHEREUPON, 17 paragraph that starts, Although both studies? 17 WILLIAM REISS ROSENBLATT 18 A. I had read this study a while back. I 18 called as a witness, and having been first duly 19 did not reread it while we were sitting here just 19 sworn, was examined and testified as follows: 20 now. 20 21 Q. I'll just read that portion for the 21 22 record. And it says, quote, Although both 22

Before the UNITED STATES COPYRIGHT ROYALTY JUDGES THE LIBRARY OF CONGRESS Washington, D.C.

) In re ) ) DETERMINATION OF ROYALTY RATES ) Docket No. 16-CRB-0001-SR/PSSR AND TERMS FOR TRANSMISSION OF ) (2018-2022) SOUND RECORDINGS BY SATELLITE ) RADIO AND “PREEXISTING” ) SUBSCRIPTION SERVICES (SDARS III) )

WRITTEN DIRECT TESTIMONY OF STEVEN BLATTER (On Behalf of Sirius XM Radio Inc.)

I. Introduction and Summary of Prior CRB Testimony

1. I am the Senior Vice President and General Manager of Music Programming at

Sirius XM Radio Inc. (“Sirius XM”). I have worked in the radio industry, including terrestrial

radio, for 30 years. In my current role, which I have filled for 13 years, I am responsible for all

music programming at Sirius XM and supervise more than 275 full- and part-time employees.

My previous experience includes programming local terrestrial radio stations in New York and

Los Angeles as well as creating and overseeing content produced for hundreds of local radio

stations across North America.

2. I testified in the two prior satellite radio proceedings before the Copyright Royalty

Board (“CRB”), SDARS I (on behalf of Sirius) and SDARS II (on behalf of the merged Sirius

XM). My testimony in those proceedings – which has been designated for inclusion here and is available to the Judges in its entirety – addressed a number of topics regarding our music channels that remain important to this day:

 The Creation and Curation of Sirius XM’s Music Channels: My prior testimony

described the components that make Sirius XM’s music channels unique, engaging, and

PUBLIC VERSION

memorable to our subscribers. Each Sirius XM-produced music channel offers not just a

playlist in a particular genre or from a particular decade but a cohesive experience built

around expert programming, song sequencing, skilled and charismatic DJs and on-air

hosts, and interstitial elements that support the theme and “feel” of the channel, all in a

commercial-free environment. This remains true of our current music channel lineup,

which is identified in Section II below.

 Sirius XM’s Music Channels vs. Terrestrial Radio: Sirius XM is, at heart, a form of

radio; given our prominence in the car, we compete most directly with traditional

terrestrial radio. My prior testimony highlighted the many ways Sirius XM improves

upon traditional radio and offers a range of programming not found among our terrestrial

radio competitors. Specifically, because terrestrial radio stations succeed by delivering

the largest possible audience to their advertisers, the options in any given market tend to

be dominated by stations offering limited playlists of popular music that will attract many

and alienate few. By contrast, because Sirius XM works on a subscription model with a

national listener base, no particular channel needs to attract enough listeners on its own to

appeal to local advertisers. We therefore are able to offer a range of channel offerings

that is far broader than what is found on local terrestrial radio, including a wide range of

niche channels, across every genre, that are not typically found on local radio (e.g.,

bluegrass, heavy metal, classic country, deep house, new age jazz, and 90s and 2000s

R&B).

Within any given channel, we also are able to offer a much deeper playlist that plumbs

the depths of the given subgenre. And on those stations that offer new music, we are able

to feature a broader array of such new music than one would encounter on more risk-

2 PUBLIC VERSION

averse local stations. Because of these traits, we are widely recognized as an industry

leader in new music discovery. Terrestrial radio stations often follow us in adding tracks

that we have selected to play on our new music channels to their own playlists. As a

result of this approach – still present in our music offerings today – Sirius XM attracts

knowledgeable and sophisticated music fans who appreciate, and who are willing to pay

for, our more tailored channel offerings and deeper playlists.

 Record Company Efforts To Get Airplay on Sirius XM Music Channels: My prior

testimony also highlighted the degree to which record companies, recording artists, and

the artists’ management companies work every day to get their music played on Sirius

XM. Whether it is providing Sirius XM with free copies of new releases, making their

artists available for in-studio interviews and performances, or waiving the restrictions of

the “sound recording performance complement,” our music programmers continue to

receive daily outreach from the record labels (typically from their radio promotions staff)

imploring us to play their music on Sirius XM and, if it already is being played, to

increase the frequency and extend the plays to additional Sirius XM channels. The

message is unmistakable: airplay on Sirius XM – targeted to our sophisticated, educated,

music-buying listeners – helps the labels sell more of their music. I discuss below how

this dynamic is reflected in our relationship with our direct licensors, and has extended to

the promotion of on-demand music services like Spotify, Apple Music, and YouTube.

 The Promotional Impact of Airplay on Sirius XM: The fourth main topic covered by my

SDARS I and SDARS II testimony was the strong promotional impact of Sirius XM

airplay on record sales. My prior testimony described the scores of emails,

correspondence, and other positive feedback from artists and their labels highlighting and

3 PUBLIC VERSION

thanking us for the impact that airplay (including exclusive airplay) on Sirius XM had on

their sales. My SDARS II testimony supported these testimonials with sales data showing

the immediate sales spikes that occurred after Sirius XM added a new track to its

playlists, with no other radio support. I also highlighted the many ancillary promotional

benefits that Sirius XM offers artists, including artist-specific programs like “Artist

Confidential” and Town Hall interviews, guest DJ sessions, email blasts and other direct

marketing mentions, artist pages on our website, and contests for tickets to concerts

hosted by Sirius XM, among others.

3. The remainder of my testimony, after summarizing our current music channel lineup (Section II below), expands upon several of the above points and updates them to reflect developments during the recent statutory license term, including:

 recent artist and label recognition of Sirius XM’s ability to break new artists and hit songs;

 increasing media recognition of Sirius XM’s ability to launch artists’ careers and promote music sales;

 data concerning Sirius XM’s promotional impact on music purchasing and consumption; and

 Sirius XM’s increasing number of live in-studio events and live coverage of music festivals from around the country on our music channels.

II. Sirius XM’s Current Music Channel Lineup

4. Sirius XM offers its subscribers 73 full-time music channels, which, as explained in my prior testimony, are carefully curated by Sirius XM’s skilled programming hosts and DJs.

As shown in SXM Ex. Dir. 1, Sirius XM’s channels fall into the following nine genres:

4 PUBLIC VERSION

 Pop: 15 stations, including Sirius XM Hits 1 (today’s pop hits), our decade stations (‘80s

on 8, ‘90s on 9, etc.), The Pulse (adult pop hits), and Pitbull’s Globalization (worldwide

rhythmic hits).

 Rock: 22 stations ranging from dedicated artist channels (Elvis, Bruce Springsteen, Tom

Petty, Pearl Jam, Grateful Dead, Jimmy Buffett’s “Margaritaville”) to classic rock,

alternative and indie rock, hard rock and heavy metal, punk, and reggae.

 Hip-Hop: 4 stations, including Hip-Hop Nation (hip-hop hits), Backspin (classic hip-

hop), and Shade 45 (Eminem’s dedicated hip-hop channel).

 R&B: 5 stations, consisting of R&B hits (The Heat and Heart & Soul) and throwback

channels (Sirius XM FLY, Soul Town, The Groove).

 Dance/Electronic: 4 stations, ranging from BPM (electronic dance hits) to Sirius XM

Chill (deep house music) to Studio 54 Radio (‘70s to ‘00s dance hits).

 Country: 8 stations, including country hits from the ‘80s to today (Prime Country, The

Highway), dedicated artist channels (Willie’s Roadhouse, Kenny Chesney’s No Shoes

Radio, and the recently launched Garth Brooks channel), and specialty channels (Outlaw

Country, Bluegrass Junction).

 Jazz/Standards: 8 stations, ranging from artist-focused stations (Siriusly Sinatra, BB

King’s Bluesville) to show tunes, easy listening, classic jazz, smooth jazz, and new age.

 Classical: 2 stations, including our exclusive Metropolitan Opera channel.

 Christian: 3 stations, including gospel, Christian pop, and Southern gospel.

5. In addition, for those subscribers with radios that have the XMH chipset, which offer higher-bandwidth to the subscriber (as further described in the testimony of Terrence

5 PUBLIC VERSION

Smith), we offer an additional 12 music channels, including channels in the Pop, Rock, Country,

Dance/Electronic, and R&B genres. See id. (denoting XMH channels as “+XTRA Channels”).

III. Sirius XM’s Promotional Value Remains Unmatched

6. My prior testimony explained the degree to which record company promotions personnel, band/artist managers, and artists themselves reach out to me and my programming staff – on a daily basis – to try to persuade us to play their music on Sirius XM. Such outreach is more common today than ever before. As demonstrated by correspondence I and others at Sirius

XM have received in the past five years, as well as articles and trade press I have reviewed during that time, the desire for Sirius XM airplay is well-established: artists, their managers, and record labels not only understand and appreciate the tremendous promotional value of airplay on

Sirius XM, but frequently thank us for helping boost the sales of their records. These first- person testimonials are buttressed by data showing that airplay on Sirius XM stimulates digital track sales through retailers like iTunes and increases streaming activity on services such as

Spotify and Apple Music.

A. Bands, Label Staff, and Artist Managers Routinely Recognize the Promotional Impact of Sirius XM on Sales

7. Since my SDARS II testimony, Sirius XM has continued to receive countless emails from artists, labels, and managers thanking Sirius XM for playing their music and recognizing the direct impact that Sirius XM airplay has had in driving sales of their releases, including on iTunes. A sample of these emails is provided below.

“Brand New” by Ben Rector

8. On June 20, 2016, thanked Sirius XM for promoting Ben Rector’s “Brand New” on The Pulse (our adult pop hits station): “Back in

September . . . The Pulse started playing Ben Rector’s ‘Brand New’ . . . [t]he first station in

6 PUBLIC VERSION

America . . . [f]ast forward and today we hit Top 10 on both Adult Top 40 charts, with no end in sight . . . . Thank you does not begin to express how much we appreciate you believing in Ben . .

. .” SXM Dir. Ex. 2.

“Lose It” By Oh Wonder

9. On January 11, 2016, emailed Sirius XM to share data showing a direct correlation between the number of spins of Oh Wonder’s “Lose It” on Sirius

XM Alt Nation and an increase in iTunes sales. correspondence showed that only three days after Alt Nation first began playing “Lose It,” iTunes daily sales for the track almost doubled. SXM Dir. Ex. 3.

“Shades of Grey” By Oliver Heldens

10. On November 12, 2015, wrote to Geronimo at

Sirius XM describing the dramatic increase in sales of Oliver Heldens’ single “Shades of Grey” after the song was played exclusively on BPM, an electronic/dance music station on Sirius XM:

Heldens’ sales on iTunes increased by 95% over the previous week, while his streams increased by 20% on Spotify and by 101% on Apple Music. SXM Dir. Ex. 4.

“Body Talk” By Dimitri Vegas

11. On May 6, 2015, informed Sirius XM that

Dimitri Vegas’s dance single “Body Talk” rose “+218%” on iTunes, and that “100%” of that rise was attributable to airplay on Sirius XM’s BPM dance/electronic channel. SXM Dir. Ex. 5

(emphasis in original).

Alina Baraz

12. On April 21, 2015, attributed artist Alina

Baraz’s success to Sirius XM airplay, writing: “[Y]ou guys have had a MASSIVE impact. We

7 PUBLIC VERSION

are over 80,000 sold and now over 10 million streams and you are the station that has really

raised its profile to get it there. No way would be here without you guys.” SXM Dir. Ex. 6

(emphasis in original).

Sons of Texas

13. On March 11, 2015, told Sirius XM that Sons of

Texas, “a baby band that really hasn’t ventured far out of state yet,” experienced “big upticks” in

sales after being played on Sirius XM. As concluded, “[w]hen I see across the

board lifts like that I know it’s your airplay impacting. Thanks for everything!!!!” SXM Dir.

Ex. 7 (emphasis added).

“500 Miles” By Haley & Michaels

14. On November 5, 2014, contacted Sirius XM to credit our

support for the increase in downloads of Haley & Michaels’ newest single, “500 Miles,” which

outpaced the country duo’s first single. “[w]anted to make sure to let you know that

thanks to your support, ‘500 Miles’ is up 12% on SoundScan from the previous week, and

has sold just shy of 5,000 downloads in 2 weeks.” SXM Dir. Ex. 8 (emphasis added).

“Thank God I Got Her” By Jonny Diaz

15. On July 10, 2014, contacted Sirius XM to express

appreciation for the success of Jonny Diaz’s song “Thank God I Got Her.” wrote:

“The power of [T]he [H]ighway is HUGE. Thank God I Got Her is top ten on the i-tunes

chart solely on the strength of The Highway. A million thank yous sir.” SXM Dir. Ex. 9

(emphasis added).

8 PUBLIC VERSION

“I’m in a Hurry” By Florida Georgia Line

16. On June 5, 2014, contacted John

Marks, programmer of The Highway, with the following message: “THANK YOU VERY

MUCH for spinning I’m In a Hurry by [Florida Georgia Line] . . . . Just this week alone we’ve sold over 2,200 downloads in only 3 days! The track has jumped up on the top 75 Country songs on iTunes. FGL is hot and this song obviously has immediate impact – a summer hit!”

SXM Dir. Ex. 10 (emphasis added).

Young The Giant’s Album “Mind Over Matter”

17. On January 29, 2014, credited airplay and

promotion on Sirius XM for the early success of Young The Giant’s new album, “Mind Over

Matter.” He thanked Sirius XM executives, telling them the label is “thrilled with the start of the

YTG Album. And I feel you guys had a real big part in our strong 1st week . . . . [T]he show we

did with you guys really helped and obviously worked . . . . Our sales dept. had this at early

projections of 22k. Well we blew that away selling 33,250. Debut 7* Single is about to go Top

5 at the format. Thanks for all the support.” SXM Dir. Ex. 11.

Brandy Clark

18. On October 22, 2013, thanked the

programmer for Sirius XM’s The Highway for Brandy Clark’s rise on the iTunes Country charts:

“Brandy [Clark] is at #4 on iTunes Country chart this morning. Thank you so much for

promoting her like you have. This would have been impossible without your help.” SXM

Dir. Ex. 12 (emphasis added).

9 PUBLIC VERSION

19. On October 16, 2013, award-winning country music artist reached out personally to Sirius XM’s country music programming executive to thank him for promoting his single “ ,” writing: “We did 2600+ singles 1st week without any [terrestrial]

radio push. That’s all word of mouth and The Highway spins. Thank you very much!”

SXM Dir. Ex. 13 (emphasis added).

AWOLNATION

20. On April 24, 2013, and

informed Sirius XM of the explosive impact its airplay had had on AWOLNATION’s record sales: “AWOLNATION . . . SELLS OVER 50,000 – 60,000 TRACKS EVERY

WEEK SINCE SIRIUS HITs1 STARTED PLAYING.” SXM Dir. Ex. 14 (emphasis in

original). The email continued: “This is the biggest PROOF of SiriusXM power I have ever

seen.” In words, “WE ARE LIVING PROOF OF THE POWER OF SiriusXM

HITS 1! . . . I literally felt it from the moment KID decided to give us spins on 20 on 20. Then when we crossed to Hits 1, we exploded into 50k per week status . . . we will cross 3 Million single soon!” Id.

“Changes” By Chubby Checker

21. On April 22, 2013, emailed Sirius XM programmer Kid

Kelly, thanking him and attributing the increase in sales of Chubby Checker’s song “Changes”

on iTunes to airplay on Sirius XM, even though the song had already been played on terrestrial

radio. As put it, “Chubby’s sales of ‘Changes’ has to be due to the Sirius XM

play, because I’ve had terrestrial play at some Gospel AM stations for a month. He’s finally

getting heard because of Sirius XM – Thank You!” SXM Dir. Ex. 15. (emphasis added).

10 PUBLIC VERSION

“Bad Ass” By Kid Ink

22. On February 20, 2013, contacted Sirius XM Senior

Producer of Specialty Programming at Shade 45 (Eminem’s Uncut Hip-Hop Channel) to thank him for playing Kid Ink’s record and for the sales that immediately followed: “[Y]ou guys were of course the 1st station to add the record and we sold 47k 1st week! Top 10 on iTunes

HipHop.” SXM Dir. Ex. 16 (emphasis added).

Alt-J

23. On January 3, 2013, emailed Sirius XM programming executives, crediting airplay on Sirius XM’s music channel, Alt Nation, for the band’s increased sales and sold out tour dates: “[S]ince you started playing ALT J on Alt

Nation our sales have gone way up. Check it out[,] album selling 6,000 a week and single

4,499. The tour dates are selling out 4 months in advance in NYC, Seattle, San Fran, KC,

Boston, and more.” SXM Dir. Ex. 17 (emphasis added).

24. These testimonials, including from

come from sales and marketing executives whose jobs it is to promote their artists’ recordings and who know firsthand the media outlets that drive such sales. The volume and persistence of requests over the years from such sources to get their artists’ songs played by Sirius XM, together with the kinds of testimonials excerpted above as to the direct correlation between exposure of a song on Sirius

XM and spikes in its sales, demonstrate the strong promotional impact performances of recorded music on Sirius XM has on record industry sales.

25. Sirius XM’s promotional impact is not limited to download sales. In the five years since my SDARS II testimony, we have also observed that airplay on Sirius XM stations

11 PUBLIC VERSION

also leads to a marked increase in on-demand play on streaming services such as Spotify and

YouTube. , for example, emailed us to thank

Sirius XM for promoting Old Dominion’s single, noting the impact on Spotify activity: “Just wanted to let you know that the guys[’] single Shut Me Up has crossed the one million streams mark on Spotify, all thanks to you guys and The Highway airplay, and in only 6 weeks! Thank you all so damn much! We love you guys and appreciate the support.” SXM

Dir. Ex. 18 (emphasis added).

B. Record Labels and Managers Frequently Credit Sirius XM with Helping to Break New Artists

26. Sirius XM continues to be identified by record labels and managers as the first major outlet to break bands on the national stage. The following are examples of emails from labels and managers thanking Sirius XM for championing their bands and introducing the rest of the industry to them.

“Strangers” By The Rebel Light

27. On June 14, 2016, updated Sirius XM on the release of The Rebel Light’s single “Strangers” on Dualtone Records and credited Sirius

XM for much of the band’s early success. wrote: “Your support has now officially helped us sign with 2 record labels in 6 months!! The power of Regan & Alt Nation!

Honestly we talk about you and what you have done for us every [] day and how grateful we are for your continued support. None of this would have been possible without [you] and we can[’]t thank you enough for literally being the person th[at] launched The Rebel Light’s career.”

SXM Dir. Ex. 19 (emphasis added).

12 PUBLIC VERSION

FKA Twigs

28. On December 10, 2014, reached out to Sirius XM to thank us for promoting FKA Twigs: “ALL YOUR SPINS got her [onto the] MTV artist to watch program for 2015 ...... now she’s going mainstream . . . thank you, for the crazy support from both of you . . . it’s truly wonderful!!!!” SXM Dir. Ex. 20 (emphasis in original).

“Pompeii” By Bastille

29. On October 14, 2013, thanked Sirius XM’s

Alt Nation for Bastille’s tremendous success. told Sirius XM that our airplay

“blazed the trail for us to have a [] massive hit and build the foundation for an excellent band . . . . 6 months later[,] ‘Pompeii’ is #1 on the Alternative radio chart and Bastille has a bright future in the States. It started at Alt Nation and, on behalf of the band, management and all of us at the label, thank you very much. We appreciate everything you do for our artists and here’s to many more years of great music and fun!” SXM Dir. Ex. 21 (emphasis added).

C. Artists Provide Exclusive and Live Content to Sirius XM Because of Sirius XM’s Promotional Impact

30. Artists are eager to provide exclusive content to Sirius XM because of the significant promotional impact of our airplay. For example, artists often perform live at our studio and interact with Sirius XM listeners. Artists assist us to create pop-up channels, which are short-term, artist-hosted channels that showcase the artist’s music catalog.

31. Exclusive In-Studio Interviews/Performances. Sirius XM often hosts exclusive promotional events on its music channels, which give artists a unique opportunity to connect with fans on-air. The following are just a few examples:

 Hamilton Town Hall hosted by Anderson Cooper: In May 2016, Sirius XM presented an exclusive special featuring the Tony Award-nominated cast of the Broadway hit musical Hamilton. The Town Hall, hosted by Anderson Cooper, offered subscribers inside access to the creators and stars of the show.

13 PUBLIC VERSION

 Duran Duran Town Hall: In April 2016, Duran Duran held a Sirius XM Town Hall to promote their new album Paper Gods, during which the band took questions from fans.

 Adele Town Hall: In November 2015, Sirius XM hosted a Town Hall with Adele, who sat down with a select group of Sirius XM listeners for an intimate Q&A session about her career and her then-soon-to-be-released, highly anticipated new album, 25.

 Andra Day In-Studio Performance: In November 2015, rising soul artist Andra Day promoted her new album on Sirius XM, performing several live songs and chatting on the air with Sirius XM’s host.

 Time for Three In-Studio Performance: In June 2015, classical string trio Time for Three promoted their latest album at Sirius XM studios, where they played live for Sirius XM fans.

 Kelly Clarkson In-Studio Performance: In March 2015, Kelly Clarkson gave a live, in-studio performance at Sirius XM, covering Tracy Chapman’s “Give Me One Reason.”

 Ingrid Michaelson World Premiere: In February 2014, Ingrid Michaelson appeared live on Sirius XM’s The Pulse to world premiere her new hit single “Girls Chase Boys.”

32. Pop-Up Channels. Artists routinely seek to be featured on Sirius XM short-term or “pop-up” channels that feature the recordings of a single artist. This phenomenon powerfully demonstrates their belief, expressed to me time and again, that a Sirius XM channel dedicated to their music represents a powerful promotional opportunity. Their willingness—and strong desire—to have their music featured around the clock also refutes any contention that Sirius XM airplay substitutes for sales or other forms of paid consumption. Some examples of prominent artist pop-up channels follow.

33. In 2014 Sirius XM hosted an exclusive music station devoted to the songs of Billy

Joel. The “Billy Joel Channel” was so successful at driving sales for Joel’s concerts that Sirius

XM brought the channel back in January 2016 and again in October 2016. SXM Dir. Ex. 22. In addition to playing songs from all of Joel’s live and studio albums, Joel occasionally introduced

14 PUBLIC VERSION

his songs on-air, offering unique insight into their meaning. Id. Over multiple recording sessions, Joel spent many hours of his time recording these song introductions and relating these stories exclusively to Sirius XM subscribers. The Billy Joel Channel also featured interviews and performances from Joel, as well as special guest DJ sessions from Joel and other artists. Id.

34. Sirius XM featured a James Taylor pop-up channel in May and June of 2015 to promote Taylor’s new album Before This World. Taylor also participated in a Sirius XM Town

Hall, where he talked with Sirius XM listeners about his music. As Rolling Stone reported on

June 24, 2015, Taylor’s new album achieved the number one record on the Billboard 200 for the first time in his career—47 years after his first album debuted. The magazine credited Sirius XM for helping Taylor reach number one. The magazine observed that “the surprise Number One was also aided by a vigorous promotional campaign by SiriusXM, who dedicated a channel to the singer and hosted a Town Hall with Taylor prior to Before This World’s arrival.” SXM

Dir. Ex. 23 (emphasis added).

35. The 24 pop-up channels created by Sirius XM during the recent rate period featured the music of such diverse and iconic artists as David Bowie, Bon Jovi, Tom Petty,

Metallica, Neil Diamond, Pink Floyd, Michael Jackson, Iron Maiden, Prince, and Barbra

Streisand. A complete list of these and other pop-up channels is attached hereto as SXM Dir. Ex.

24.

36. Section 114 Waivers. Many musicians and record labels, in recognition of the promotional value of airplay on Sirius XM, have granted Sirius XM waivers of certain of the

Section 114 statutory license restrictions that would otherwise limit the number of plays an artist’s music can receive in a given time period on Sirius XM. These waivers allow Sirius XM to play that artist’s music more frequently and thus enable rights-holders who have signed the

15 PUBLIC VERSION

waivers to receive even greater promotion from Sirius XM airplay. Section 114 statutory waivers also make possible the album previews and many of the pop-up channels discussed above as well as Sirius XM’s popular artist-dedicated channels, such as Bruce Springsteen’s E

Street Radio, Pearl Jam Radio, Tommy Petty Radio, the Grateful Dead Channel, and Elvis

Radio.

37. Most recently, Sirius XM launched an exclusive channel devoted to the music of

Garth Brooks, which marked the first time Brooks’ music has been presented 24/7 in one place, on radio or online. As Mr. Brooks said: “It’s rare to be handed an opportunity that can grow as big as you can dream. That is exactly what SiriusXM has just done . . . . I speak for myself and the entire team when I say we are very grateful for this opportunity and very excited about its potential.” SXM Dir. Ex. 25. In celebration of the launch of The Garth Channel, Mr. Brooks performed a special invitation-only concert for Sirius XM listeners on September 8, 2016, playing a concert at the historic Ryman Auditorium in Nashville for the first time in his career.

The Section 114 waiver makes The Garth Channel possible.

D. Bands’ Press Releases Tout the Promotional Impact of Sirius XM

Artists, labels, and their managers often credit Sirius XM in press releases they issue to promote their music. Below are just a few examples.

John Newman’s “Love Me Again” Achieves Gold Status

38. On January 31, 2014, Universal Music’s Republic Records issued a press release that cited the role of Hits 1, Sirius XM’s pop hits channel, in breaking John Newman’s gold single “Love Me Again.” According to the press release, the song sold over 500,000 digital copies, but before it could reach that milestone, “Sirius XM Hits 1, which played the song before any major radio outlet, [] officially spun the song over 1,000 times.” SXM Dir. Ex.

16 PUBLIC VERSION

26 (emphasis added). The release further stated that “[t]he value and reach of the satellite radio juggernaut has proven a major catalyst for breaking the UK singer, songwriter, and producer, consistently exposing him to a massive American audience before anyone. They also continue playing the single non-stop, fueling his fire even further.” Id. (emphasis added).

JT Hodges is #1 Two Weeks in a Row

39. The artist JT Hodges issued a press release on March 14, 2013, announcing that

his single “Sleepy Little Town” was number one on Sirius XM’s The Highway for two

consecutive weeks. Mr. Hodges thanked Sirius XM for “giving this song a life and outlet for it

to connect with people like I knew it could.” SXM Dir. Ex. 27. Executives at Mr. Hodges’ label, Universal Music, likewise observed that “Sirius XM listeners and young country fans have embraced the music of JT Hodges. He continues to have a strong download pattern of sales driven by their active listeners. With the unwavering support of Sirius XM and social media,

JT Hodges is carving out a new model of success.” Id. (emphasis added).

E. The Media Is Tuned Into the Promotional Impact of Sirius XM Airplay

40. The media also has recognized Sirius XM’s promotional value over the past five

years. This has taken the form of articles about Sirius XM programming in industry magazines

as well as interviews with artists who have credited Sirius XM airplay and promotion with

elevating their success, expanding their fan base, encouraging terrestrial radio stations to add

their music to their playlists, and leading them to the top of the Billboard and Mediabase charts.

Additionally, and as I discuss further below, a recent Billboard Magazine survey of record label

executives named three of Sirius XM’s programmers as the most influential rock programmers in

the industry, responsible for “shaping the future of rock.”

17 PUBLIC VERSION

Rolling Stone: “How Satellite Radio Is Breaking Country’s Next Big Stars”

41. Rolling Stone captured this phenomenon in a January 5, 2015, article titled “How

Satellite Radio Is Breaking Country’s Next Big Stars,” chronicling the tremendous promotional impact Sirius XM has had on the careers of new country music stars. The article described how

Sirius XM’s then-Senior Director of Music Programming, John Marks, carefully curated its contemporary country music channel, The Highway, to play songs from unknown but promising country music artists alongside hits from country’s proven stars. The strategy is to give unfamiliar artists national exposure, in many cases before they have signed with a label. As the article explained:

Thanks to Marks, acts with no label backing whatsoever have gotten a taste of national airplay. The Highway’s audience of Sirius XM subscribers may be dwarfed by the number of listeners who tune in to FM country, but its reach stretches from Albuquerque to Anchorage. Kevin Neal, a booking agent whose clients Florida Georgia Line benefited mightily from satellite play, puts it this way: “You get on the national stage immediately, as opposed to getting your song tested in Bossier City.”

SXM Dir. Ex. 28 (emphasis added). Indeed, The Highway’s mission is “to find acts to break nationally.” As Mr. Marks explained, subscribers in large measure tune into Sirius XM for “the curation of new music.” Part of the station’s role is to expose listeners to new music that they are likely to enjoy based on their channel preferences. By contrast, “[t]he opposite of novelty reigns at FM, the idea being that listeners are less likely to start channel surfing when songs they already know come on.” Id.

The New York Times: “Star-Making Machinery’s New Player”

42. In an article detailing the profound impact Sirius XM has had on breaking new artists in the alternative rock space, The New York Times profiled Jeff Regan (the program director of Sirius XM’s alternative rock music channel Alt Nation) as an alternative rock music tastemaker. SXM Dir. Ex. 29. As the article reported, Mr. Regan “jump-started the careers of

18 PUBLIC VERSION

several groups, among them Atlas Genius, Blondfire and Capital Cities . . . . plucking [these]

artists from obscurity,” and also helped create alt-rock success stories like Foster the People and

Grouplove. Id. At the time Mr. Regan broke Atlas Genius after discovering their music on a blog, for example, “[n]ot only did the band have no record cont[r]act, but its members were also going to college and had never even toured.” Id. Nonetheless, Regan promoted Atlas Genius’s song “Trojans,” which led to “[a] bidding war between several major labels” (the band ultimately signed with Warner), and the song went to number five on Billboard’s 2012 alternative music chart. Id. The Times also noted that many emerging bands are crediting airplay on Sirius XM’s

Alt Nation and XMU (Sirius XM’s new indie rock station) for “increased sales” and for

“attract[ing] interest from major labels.” Id. The article further observed that Sirius XM’s The

Highway was the first to play Florida George Line’s “Cruise,” a hit that topped the Billboard

Country Songs chart for four weeks, and that after this early airplay on Sirius, the duo signed a recording contract with Republic Nashville. Id.

AllAccess: Interview with Logan Mize

43. AllAccess’s April 12, 2015 interview with Logan Mize, a country singer who had

recently signed with Sony Music’s Arista Nashville label, attributed Mr. Mize’s success, in

significant part, to airplay on Sirius XM’s The Highway. AllAccess noted that Mr. Mize’s

“current single, ‘Can’t Get Away From A Good Time,’ . . . earned sales in excess of 100,000

copies and had been streamed more than 1,000,000 times before ever hitting terrestrial

radio thanks in part to a boost from SiriusXM’s ‘The Highway.’” SXM Dir. Ex. 30

(emphasis added).

Rolling Stone Country: Interview with Ray Scott

44. On September 29, 2014, Rolling Stone Country published an interview with

19 PUBLIC VERSION

newly emerging country artist Ray Scott, who “credits a lot of his success to the support of

Sirius XM, who consistently played his songs and spurred his digital resurgence,” both when he was signed with Warner Brothers and more recently as an independent. SXM Dir. Ex.

31 (emphasis added).

Sirius XM Programmers Top Billboard Magazine’s List of Most Influential Programmers

45. In May 2016, Billboard Magazine published the results of a survey conducted of record label executives who were asked to identify those rock programmers in the industry “most influential in getting exposure for new rock repertoire.” Three of our programmers – Mr. Regan of Alt Nation, Jess Besack of The Spectrum, and Vincent Usuriello of Octane – collectively took first place. As the article announcing the results of the survey noted, Ms. Besack offered the support of The Spectrum to The Record Company and Barns Courtney, among other acts; Mr.

Regan “got behind Twenty One Pilots’ ‘Ride’ before its release as a single, helping the track’s rise to No. 1 on the Alternative airplay chart”; and Mr. Usuriello “has seen his support for the band From Ashes to New propel its album Day One to No. 2 on Hard Rock Albums.” SXM Dir.

Ex. 32.

F. Industry Data Confirms That Airplay on Sirius XM Leads to Increased Sales

46. Industry data from sources such as SoundScan (tracks sales data) and Mediabase

(tracks airplay data) further support that airplay on Sirius XM leads to greater music sales.

“S.O.B.” By Nathaniel Rateliff & The Night Sweats

47. Once Sirius XM began playing Nathaniel Rateliff & The Night Sweats’ single

“S.O.B.” sales jumped 418% that first week:

20 PUBLIC VERSION

The viral effect of Sirius XM’s airplay on the single is also demonstrated by a July 13, 2015

email to Sirius XM from (one of Sirius XM’s direct

licensors). thanked Sirius XM for the viral effect Sirius XM’s airplay had on

“S.O.B.” She reported that “due to the amazing airplay from Spectrum channel, ‘S.O.B.’ single

sales have completely blown up . . . we went from 7 downloads to over 1,600 this week. This

song is performing like a hit – instant and undeniable – and it’s ALL SIRIUSXM.” SXM Dir.

Ex. 33 (emphasis in original). The band also took out an advertisement in Billboard which

credited Sirius XM for launching the band’s hit single, noting that “SiriusXM launched ‘S.O.B.,’ and it became a foot-stompin’ classic.” SXM Dir. Ex. 34.

“Break Up With Him” By Old Dominion

48. Sirius XM’s The Highway added Old Dominion’s song “Break Up with Him” to its playlist on October 13, 2014. In the first week after being added to The Highway, sales of the song increased by 134%. By December, as The Highway continued to feature the song, sales

21 PUBLIC VERSION

continued to grow. The graph below illustrates the tremendous sales increase that followed

Sirius XM’s airplay:

49. The band also credited airplay on Sirius XM as the catalyst for spurring interest in

the band on terrestrial radio stations across the country. As part of an interview with MusicRow,

the author observed that “[t]hanks to early airplay on SiriusXM, [the band’s hit single]

‘Break Up With Him’ piqued the interest of radio programmers across the nation, and

several stations began airing the album cut late last year. Responding to demand for the

song, the band’s management is rushing out the radio single, with an official impact date of Feb.

9.” SXM Dir. Ex. 35 (emphasis added).

“Dangerous” By Big Data

50. On December 10, 2013, Sirius XM added Big Data’s song “Dangerous” to its rotation. One week prior to Sirius XM adding the song to its playlists, Big Data sold 89 copies of the song. Immediately after Sirius XM began playing it, sales skyrocketed to 1,400. The graph below reflects the dramatic increase in sales following the Sirius XM airplay:

22 PUBLIC VERSION

51. On April 1, 2015, Cleveland.com, a news website serving northeastern Ohio, published an interview with Alan Wilkis, a member of Big Data. Asked when he first realized his song “Dangerous” was a hit, Mr. Wilkis cited its airplay on Sirius XM: “The moment I knew my life was really changing was . . . when Sirius XM started putting the song into rotation.

That sort of blew the lid off of everything.” SXM Dir. Ex. 36 (emphasis added).

“Chillin It” By Cole Swindell

52. On May 7, 2013, Sirius XM’s The Highway debuted country artist Cole

Swindell’s single “Chillin It.” This was the first time Swindell’s music was ever played on the radio, and Swindell, at the time an unsigned artist, documented his listening experience on

YouTube. Swindell remarked that hearing his song on The Highway “[was] huge for me as an unsigned artist . . . XM is so powerful . . . . A huge thanks to John Marks over at XM radio, buddy, thank you for believing in my music and taking a chance on my new song, ‘Chillin It.’”

23 PUBLIC VERSION

Mr. Swindell’s remarks can be viewed at the following address:

https://www.youtube.com/watch?v=d1emNSngqrc (last visited October 13, 2016).

53. In a later interview with AllAccess, Swindell attributed his big break to Sirius

XM: “[My manager] got the song to John Marks with SiriusXM, and he gave me a shot. After being #1 for several weeks and selling a noticeable amount of downloads, labels began to take interest, and I signed with Warner Bros. Records. They immediately took it to country radio, who really embraced me, and the rest is history.” SXM Dir. Ex. 37.

54. SoundScan and Mediabase data also indicate that downloads of Swindell’s music increased after his Sirius XM debut. Prior to May 2013, Swindell had zero direct sales. The week ending May 5, 2013, digital sales were at 804. The next week ending May 12, 2013, downloads grew to 4,711, with downloads the following week jumping 44% to 6,795. After

Memorial Day weekend, when Sirius XM gave Summer-themed music like Swindell’s extra spins, weekly sales jumped to 13,900, with sales remaining strong thereafter:

24 PUBLIC VERSION

The correlation between airplay on The Highway and the increase in Swindell’s sales during this

period is apparent. Indeed, in an August 2013 article entitled “Cole Swindell’s ‘Chillin It’

Cruises With Satellite-Enabled Guidance,” Billboard Magazine recognized that the Highway

introduced “Chillin’ It” to the public, and, in turn, helped “land [Swindell] a major-label

recording contract.” SXM Dir. Ex. 38. Citing the above sales data, Billboard noted that

Swindell’s single “picked up steam quickly” as a result of Sirius XM’s “early support,” which

was “key in the song’s cruise to prominence.” Id. As Swindell put it, “[n]ot only did [Sirius

XM] help me get a record deal, they helped put me on the fast track.” Id.

IV. Sirius XM’s Direct Licensing Initiative

55. As discussed in greater detail in the testimony of George White, Sirius XM has

signed direct licenses with 498 record companies and artists. A key benefit of those direct

licenses is the mutually beneficial relationships that are fostered between Sirius XM

programmers and the artist and label of the directly licensed material. As Mr. White describes,

the license gives artists and label promotion staff access to the Sirius XM programmers who select songs to air on Sirius XM channels – something they find incredibly valuable given the promotional value of Sirius XM airplay and the fierce competition to place songs on our service.

56. Our programmers value the relationships we have developed with our direct licensors as well. We receive a steady flood of emails, calls, social-media messages, CDs, and mp3 files from record labels seeking airplay – so much so that it is impossible to devote significant attention (or even listen) to all of them. When we receive a submission from a direct licensor, though, we do pay special attention to the tracks and/or pitches that we receive, and seek out opportunities for them when we have open slots in our programming schedule. We know that our direct licensors (along with Mr. White’s team) know our channels well, and know

25 PUBLIC VERSION

what sort of tracks will potentially fit on those channels. That is valuable to us – as are the

DMCA waivers contained in the direct licenses (which allow us to feature several tracks from a new album in a short period) and the willingness of directly licensed artists to come in for interviews and performances.

57. An example that demonstrates our relationship with our direct licensors is the band “Ghost,” whose label Loma Vista is covered by our direct license with Concord. The band’s track “Cirice” was a bit outside the norm of what we typically play on Octane, our contemporary hard rock channel. We nevertheless took a chance on the track because of our direct license relationship with Concord, and we were the first major outlet in the U.S. to do so.

Not only was the band appreciative, but the track took off when we started playing it – and ended up winning the 2016 Grammy for Best Metal Performance. Since that time, we have played the band’s second single heavily, and more recently were provided the exclusive world premiere of its newest single, “Square Hammer.”

58. Another great example is Andy Grammer. For several years, Sirius XM has had a direct license with Grammer’s label, S-Curve Records. S-Curve renewed their license in January

2015, when the term of the original license was set to expire. As a result of that license, Steve

Greenberg, president of the label, has formed a close relationship with our programming team, including Kid Kelly, who oversees our pop and “hits” stations.

59. Around October 2014, Mr. Greenberg sent us a copy of Andy Grammer’s newest single, “Honey I’m Good.” The track, a hand-clapping, singalong stomp, was something of an outlier as compared to the smoother, R&B-oriented hits popular at the time – more hootenanny than dance club. Had S-Curve not been a direct licensor, we almost certainly would not have taken a chance on a long-shot like this. But Kid Kelly thought the quirky song might resonate

26 PUBLIC VERSION

with listeners and worked directly with Mr. Greenberg (who also produced the track) to craft a

plan to feature the track on Hits 1 (our main pop hits station) just before Christmas, a period when the major labels largely take a break from promoting new tracks and “Honey I’m Good” wouldn’t be competing as heavily with other new releases. What happened is that sales exploded

(without airplay on any other radio stations), as evidenced by the following industry data:

As The New York Times reported, following Sirius XM’s airplay, “Honey I’m Good” caught on with FM stations as well, and “eventually sold about 2.5 million copies, making it one of the biggest hits of 2015.” SXM Dir. Ex. 39. The song also reached number 9 on the Billboard Hot

100, was certified triple-platinum by the RIAA, and was ranked one of the ten best-selling songs of 2015 according to SoundScan. Mr. Greenberg told The New York Times that “[t]here is absolutely no chance in the world that this record could have launched without the backing of

Hits 1.” Id.

27 PUBLIC VERSION

60. I’ll conclude this section with yet one more example, Daya’s “Hide Away,” which is covered by our direct license with Z-Entertainment. Daya was a developing, unpolished 15- year-old from Pittsburgh with no prior releases. Steve Zap at Z-Entertainment approached us with her new track “Hide Away,” and Kid Kelly recognized something special in it. He began playing the track regularly on Venus, our rhythmic pop station, and then Hits 1, rising to well over 100 spins per week – meaning we were playing the track, on average, nearly once every hour. Had we not had a direct license with Z-Entertainment, we likely would never had heard the track in the first place, and there is almost no chance we would have played a completely unknown artist in heavy rotation on our two biggest hits stations. Thanks to our relationship, we did – and the track, which had never reached 2,500 weekly downloads prior to our airplay, soon became a million-selling hit, as indicated by the chart below:

Immediately following airplay on Sirius XM’s pop stations, downloads of “Hide Away” increased. and emailed me and

28 PUBLIC VERSION

Kid Kelly on August 6, 2015, attributing the climb of the song “Hide Away” by Daya on the iTunes charts to airplay on Sirius XM’s Hits 1. wrote, “We move up in iTunes every

time u spin it . . . . Jumped 30 positions due to [H]its 1 weekday play. INSANE guys.” SXM

Dir. Ex. 40 (emphasis added). likewise observed that this is “[t]he power of Sirius XM

Hits 1! But more importantly, I have never had a record this reactive!!!!! Thank you.” Id.

(emphasis added).

V. Sirius XM Continues to Differentiate Itself Through the Production and Broadcast of In-Studio Performances and Live Music Festivals

A. In-Studio Guest Interviews and Performances

61. My prior testimony described Sirius XM’s programming philosophy, including

our significant investments in on-air talent. We recognize that much of the music that is played

on Sirius XM is widely available, often for “free” to the listener, on terrestrial radio. Our on-air

hosts provide one of the ways we differentiate our channels from our competition. In addition to

our regular on-air hosts, Sirius XM continues to make significant investments in identifying

high-profile, talented guests to appear across our music (and non-music) channels. Today, Sirius

XM has nine full-time employees and four part-time employees on its talent acquisition staff focused exclusively on identifying, acquiring and scheduling guests and events that we believe improve the listening experience, motivate “trialers” to subscribe, and reduce our “churn” rate.

62. Our talent acquisition staff receives hundreds of solicitations every week for guest appearances on Sirius XM. In addition, many of our music programmers and on-air hosts have developed deep connections in the music and entertainment industries and also regularly receive such solicitations. Because appearing on Sirius XM is an opportunity for exposure in a commercial-free environment, we receive many more solicitations for appearances than we have

29 PUBLIC VERSION

space available. We therefore have to be selective in whom we invite and on what channels we

schedule appearances.

63. Attached as SXM Dir. Ex. 41 is a sample collection of emails distributed by Liam

Davenport, Sirius XM’s Director of Talent Relations, detailing who will be appearing across the

various Sirius XM channels on the following day. As shown in Mr. Davenport’s emails, Sirius

XM has a very diverse group of daily guests. Obviously, many are recording artists who appear

to perform in our studios. For example, Adele’s first national radio appearance was in the Sirius

XM studios. In Section III.C above, I highlighted several other exclusive in-studio interviews

and performances that Sirius XM has hosted over the last rate period.

64. We also recognize that even subscribers listening to our music channels have

other interests. We therefore book as guests non-musicians to offer unique and compelling commentary that our subscribers are unlikely to hear anywhere else – certainly not on their local

music station or webcasting service. For example, just this past September, Dr. Neil deGrasse

Tyson, the famous American astrophysicist and cosmologist, was interviewed on Hip Hop

Nation and Shade 45, two of our hip hop channels. We also routinely welcome actors and actresses to our studio for appearances on our music channels, including such bold-faced names as Samuel L. Jackson, Kate Hudson, and Anna Kendrick. We have also had talented musicians

and non-musicians serve as “Guest DJs” on our channels, including Kristen Chenoweth, Bette

Midler, Ethan Hawke, Judd Apatow, and Daniel Radcliffe. We will continue to invest in these

sorts of live events because we believe it adds to the compelling mix of music and non-music

content that creates buzz for our service, keeps audiences engaged and drives subscriptions.

65. Such performances and interviews are not always limited to our studios. Sirius

XM produces a number of live events, accessible only to our subscribers, at concert venues

30 PUBLIC VERSION

around the country. By broadcasting exclusive live performances from some of the biggest and hottest names in entertainment, these events provide an excellent opportunity to demonstrate our value to prospective and existing subscribers. Just this year, Sirius XM has produced and broadcast the following subscriber-exclusive events:

 Garth Brooks at Ryman Auditorium (Nashville, TN), broadcast on The Garth Channel;

 Coldplay at Stephen Talkhouse (Amagansett, NY), broadcast on Alt Nation;

 Kenny Chesney at the Stone Pony (Asbury Park, NJ), broadcast on No Shoes Radio;

 CHVRCHES at Rough Trade (New York, NY), broadcast on Sirius XMU;

 Kings of Leon at Poisson Rouge (NYC), broadcast on Alt Nation;

 Korn at the Ace Hotel (Los Angeles), broadcast on Octane; and

 Kaleo at The Box (NYC), broadcast on The Spectrum.

B. Live Music Festival Coverage

66. Spotlighting another closely related development during this current rate period,

Sirius XM has in recent years focused considerable resources on covering live music festivals.

Over the past five years, large-scale music festivals have become an increasingly entrenched part of American culture. Live broadcasts of these music festivals, and other live music events, lead to more passion, excitement, and engagement both for our service and music generally.

67. With input from various members of my team, I am responsible for selecting the music festivals that Sirius XM will broadcast. Rather than occurring on a separate channel, our festival coverage typically “takes over” an existing station during the days the festival is happening. As I describe below, broadcasting a live music event, such as a music festival, is quite complicated and requires significant involvement across a number of our teams, as well as a major expenditure of time and money.

31 PUBLIC VERSION

68. Broadcast Engineering. Sirius XM has a team of three full-time employees in

the Broadcast Engineering group, along with more than twenty independent contractors, to

manage the broadcasting of live events. This team handled some 450 live music and non-music

events in 2015 and expects to broadcast more than 500 in 2016. In nearly every one of these

events, at least one Sirius XM resource is “on the ground” to manage the broadcast engineering

logistics. For large events, such as the Coachella Valley Music & Arts Festival described below,

as many as seven engineers are on-site to manage the capture and transmission of dozens of

artists appearing on multiple stages. This group handles the “backhaul” of the audio (typically

via ISDN or satellite uplink) from the event to our New York or Washington, DC facilities,

where it is then routed to the appropriate studio (or studios) for broadcast on a particular channel

(or channels).

69. Programming. Live broadcasting a music festival also requires a significant

number of production staff and music programmers on-site. Our production team handles the

logistics of capturing and editing live elements from the event for use as interstitials in the

broadcast. It is not uncommon for as many as 15 people from the music programming and

talent acquisition teams to be on-site at a large festival. Our festival coverage also includes backstage interviews with the performing artists conducted by our on-air hosts. At certain

events, our on-air hosts are also the on-stage hosts of the entire event.

70. In 2016, Sirius XM has brought performances from at least 15 festivals and live

music events to our subscribers, and Sirius XM hosts were present for each of these events.

Included in this year’s line-up are:

 South by Southwest (March 16-18) is a multi-day music festival held in Austin, Texas. A number of Sirius XM hosts broadcast shows from Austin during SXSW, including Outlaw Country host Mojo Nixon, Shade 45 host Sway, and SiriusXMU hosts Jenny Eliscu and Julia Cunningham.

32 PUBLIC VERSION

 Ultra Music Festival (March 16-18) is a multi-day event held in Miami, Florida featuring electronic music. This year’s festival featured interviews and live DJ sets from Armin van Buuren, Tiesto, Eric Prydz, Martin Garrix, Afrojack, Carl Cox, Carnage, Cedric Gervais, and Hardwell. SMX broadcasted the Ultra festival on UMF Radio. The coverage was anchored by Sirius XM on-air hosts Geronimo, Liquid Todd, Ben Harvey, Danny Valentino, and Tommy West.

 Coachella Valley Music & Arts Festival (April 15-17) consists of two three-day events occurring over two consecutive weekends. This year’s festival included interviews and live performances from Guns N’ Roses, Ellie Goulding, Jack U, M83, The Kills, Disclosure, Zedd, CHVRCHES, Halsey, Calvin Harris, Sia, and Major Lazer. Sirius XM launched a limited-run, Coachella Radio channel, and also aired performances on Electric Area, Alt Nation, and The Spectrum. The coverage was anchored by Sirius XM on-air hosts Jenny Eliscu, Jeff Regan, Julia Cunningham, Geronimo, Liquid Todd, and Ben Harvey.

 Rock On The Range (May 20-22) is a multi-day event outside Columbus, Ohio billed North America’s biggest hard rock festival. This year’s event featured performances by Disturbed, Five Finger Death Punch, Shinedown, Rob Zombie, Lamb of God, Megadeth, Bring Me The Horizon, Bullet For My Valentine, Sixx:A.M., Hellyeah, Sevendust, Trivium, Pop Evil, Red Sun Rising, Sick Puppies, Ghost, and Highly Suspect. The event was hosted by Octane DJs Kayla Riley, Grant Random, and Jose Mangin. In addition, performances also aired on Liquid Metal and Faction.

 The Bonnaroo Music and Arts Festival (June 9-12) is a multi-day event outside Nashville, Tennessee featuring more than 150 musicians, bands, and comedians performing across 12 separate stages. This year’s festival featured performances by Pearl Jam, Dead & Company, LCD Soundsystem, J. Cole, Ellie Goulding, Macklemore & Ryan Lewis, Death Cab for Cutie, M83, Haim Halsey, CHVRCHES, Chris Stapleton, and The Chainsmokers. Sirius XM broadcasted Bonnaroo performances across a number of channels, including The Spectrum, Alt Nation and Jam On. The coverage was anchored by Sirius XM on-air hosts Jenny Eliscu, Jeff Regan, Julia Cunningham, and Ari Fink.

 The Electric Daisy Carnival (June 17-19) is a multi-day event held in Las Vegas and one of the largest electronic music festivals in the world. Sirius XM broadcasted sets from famous DJs such as Armin van Buuren, Knife Party, Martin Garrix, Afrojack, Axwell / Ingrosso, Carnage, Duke Dumont, Alesso, Above & Beyond, and NGHTMRE on Electric Area. The coverage was anchored by Sirius XM on-air hosts Geronimo, Liquid Todd, Ben Harvey, Danny Valentino, and Tommy West.

 The Firefly Music Festival (June 19-21) is a multi-day event outside Dover, Delaware and the largest music festival on the East Coast. This year’s festival aired on Alt Nation, The Spectrum, SiriusXMU, and Jam On, and featured performances from The Killers, Kings of Leon, Cage the Elephant, Hozier,

33 PUBLIC VERSION

Modest Mouse, Foster the People, Spoon, Bastille, Walk the Moon, Gary Clark Jr., Big Data, Awolnation, Sublime with Rome, and Cold War Kids. The coverage was anchored by Sirius XM on-air hosts Jenny Eliscu, Jeff Regan, and Madison.

 Panorama Festival (July 22-24) is a multi-day music festival on New York City’s Randall’s Island Park. This year’s festival aired live on SiriusXMU, Alt Nation, and The Spectrum, and featured live performances by Alabama Shakes, The National, Nathaniel Rateliff & The Night Sweats, Run the Jewels, Grace Potter and the Nocturnals, and Major Lazer. The coverage was anchored by Sirius XM on-air hosts Jenny Eliscu, Jeff Regan, and Madison.

 Lollapalooza (July 28-31) is a multi-day event featuring live performances held in Chicago’s Grant Park and featured interviews and live performances from The Red Hot Chili Peppers, Jane’s Addiction, Flume, St. Motel, The Arcs, Foals, Third Eye Blind, Two Door Cinema Club, and more. The coverage was anchored by Sirius XM on-air hosts Jenny Eliscu, Jeff Regan, and Julia Cunningham.

 The LOCK’N Festival (Aug. 25-28) is a multi-day event featuring roots, rock, and so-called “jam bands” such as Phish, Ween, Widespread Panic, Gov’t Mule, and The Allman Brothers Band. Sirius XM broadcasted the LOCK’N Festival on Jam On, and it was anchored by Sirius XM host Ari Fink.

 The Electric Zoo (Sept. 2-5) is a multi-day event on New York City’s Randall’s Island featuring some of the biggest names in electronic music. This year’s event included interviews and performances by Sam Feldt, Tiesto, Cash Cash, Big Gigantic, GTA, Felix Jaehn, and Steve Aoki. Sirius XM broadcasted the event on Electric Area, and the coverage was anchored by Sirius XM on-air hosts Geronimo, Liquid Todd, Ben Harvey, Danny Valentino, and Tommy West.

 The Farm Aid Concert (Sept. 17) is an annual fund-raising event to aid local farmers that is co-organized by Willie Nelson. This year’s event featured performances by Willie Nelson, Neil Young, John Mellencamp, Dave Matthews & Tim Reynolds, Alabama Shakes, Nathaniel Rateliff and the Night Sweats, Sturgill Simpson, Jamey Johnson, and Margo Price. The event was broadcast live on Willie Nelson’s own channel, Willie’s Roadhouse, with Sirius XM host Dallas Wayne conducting backstage interviews with artists.

 Hair Nation Festival (Sept. 17) is a one-day festival held in Southern California that featured interviews and performances from popular artists played on Sirius XM’s Hair Nation channel, including Lita Ford, Kix, Vixen, Enuff 2 Nuff, Slaughter, and more. The coverage was anchored by Sirius XM hosts Luc Carl and Dangerous Darren.

 The Route 91 Harvest Country Music Festival (Sept. 30-Oct. 2) is a multi-day event in Las Vegas, Nevada featuring some of country music’s most popular artists as well as up-and-coming talent. This year’s festival featured Toby Keith,

34 PUBLIC VERSION

Chris Young, Brad Paisley, Tyler Farr, Jana Kramer, William Michael Morgan, Aaron Lewis, The Randy Rogers Band, Aubrie Sellers, Billy Currington, and LoCash. In addition to live broadcasting the event on The Highway, Sirius XM hosts Storme Warren, Buzz Brainard, and Al Skop were on-site to host the broadcast and interview artists backstage; and

 The Austin City Limits Music Festival (Sept. 30-Oct. 1) consists of two three-day events occurring over two consecutive weekends. This year’s festival featured Major Lazer, Radiohead, The Chainsmokers, Kendric Lamar, LCD Soundsystem, and Mumford & Sons. Given the eclectic music line-up, Sirius XM broadcasted portions of Austin City Limits on Alt Nation, The Spectrum, SiriusXMU, and Hip Hop Nation. Hosts from each of these channels were on-site at the festival to host the broadcasts and conduct interviews with artists. The coverage was anchored by Sirius XM on-air hosts Jenny Eliscu, Jeff Regan, and Julia Cunningham.

71. Additionally, this Fall we are presenting the second leg of the Alt Nation

Advanced Placement Tour, a multi-city tour featuring up-and-coming alternative rock artists such as The Shelters, Night Riots, and the Hunna. The Tour grew out of Alt Nation’s “Advanced

Placement” program, a weekly showcase of new alternative rock artists described by Billboard as “a place for discovery of new talent.” One of the days of this tour will be broadcast on Alt

Nation.

35 PUBLIC VERSION

Page 3160 Page 3162 UNITED STATES COPYRIGHT ROYALTY JUDGES 1 APPEARANCES (Continued): The Library of Congress 2 On behalf of Sirius XM Radio, Inc.: ------X 3 R. BRUCE RICH, ESQ. IN THE MATTER OF: ) 4 RANDI W. SINGER, ESQ. ) 5 TODD D. LARSON, ESQ. DETERMINATION OF ROYALTY ) Docket No. 6 JACOB EBIN, ESQ. RATES AND TERMS FOR ) 16-CRB-0001-SR/PSSR 7 MEREDITH SANTANA, ESQ. TRANSMISSION OF SOUND ) (2018-2022) 8 ELIZABETH SPERLE, ESQ. RECORDINGS BY SATELLITE ) 9 Weil, Gotshal & Manges LLP RADIO AND "PREEXISTING" ) 10 767 Fifth Avenue SUBSCRIPTION SERVICES ) 11 New York, New York 10153 (SDARS III) ) 12 212-310-8007 ------X 13 BEFORE: THE HONORABLE SUZANNE BARNETT 14 ADAM TOLIN, ESQ. THE HONORABLE JESSE M. FEDER 15 Weil Gotshal & Manges, LLP THE HONORABLE DAVID R. STRICKLER 16 17 Hulfish Street 17 Princeton, New Jersey 08542 Library of Congress 18 809-986-1105 Madison Building 19 101 Independence Avenue, S.E. 20 Washington, D.C. 21 May 10, 2017 22 9:16 a.m. 23 VOLUME XIII 24 Reported by: Karen Brynteson, RMR, CRR, FAPR 25

Page 3161 Page 3163 1 A P P E A R A N C E S: 1 APPEARANCES (Continued): 2 On behalf of SoundExchange, The American Federation 2 On behalf of Music Choice: 3 of Musicians of the United States and Canada, Screen 3 PAUL FAKLER, ESQ. 4 Actors Guild and American Federation of Television 4 JOHN SULLIVAN, ESQ. 5 and Radio Artists, American Association of 5 MARGARET WHEELER-FROTHINGHAM, ESQ. 6 Independent Music, Universal Music Group, Warner 6 Arent Fox LLP 7 Music Group, RIAA: 7 1675 Broadway 8 DAVID A. HANDZO, ESQ. 8 New York, New York 10019-5874 9 STEVEN R. ENGLUND, ESQ. 9 212-484-3990 10 JARED O. FREEDMAN, ESQ. 10 11 JEROME L. EPSTEIN, ESQ. 11 JACKSON TOOF, ESQ. 12 ALEX TREPP, ESQ. 12 Arent Fox LLP 13 KENDALL TURNER, ESQ. 13 1717 K Street, N.W. 14 DEVI RAO, ESQ. 14 Washington, D.C. 20006 15 PREVIN WARREN, ESQ. 15 202-857-6395 16 EMILY CHAPUIS, ESQ. 16 17 Jenner & Block, LLP 17 18 1099 New York Avenue, N.W., Suite 900 18 19 Washington, D.C. 20001 19 20 On behalf of Sony Music Entertainment: 20 21 ROLLIN A. RANSOM, ESQ. 21 22 PETER I. OSTROFF, ESQ. 22 23 Sidley Austin LLP 23 24 555 W. Fifth Street, Suite 4000 24 25 Los Angeles, CA 90013 25

1 (Pages 3160 to 3163) Page 3340 Page 3342 1 JUDGE BARNETT: Thank you. Please raise 1 is like? 2 your right hand. 2 A. Sure. There is probably a lot more 3 Whereupon-- 3 demand for my time than I have time for. As Sirius 4 STEVEN BLATTER, 4 XM has grown over the years and had lots of success 5 having been first duly sworn, was examined and 5 in developing and breaking artists, there is a 6 testified as follows: 6 tremendous amount of demand on my time to talk to 7 JUDGE BARNETT: Please be seated. 7 artist managers, record labels, people who work in 8 DIRECT EXAMINATION 8 marketing roles, and all typically seeking exposure 9 BY MS. SPERLE: 9 or promotions for their artists on Sirius XM. 10 Q. If you could please state your full name 10 Q. How about your direct reports and the 11 for the record and spell your last? 11 programmers at Sirius XM, do they interact with 12 A. Steven Blatter, B-l-a-t-t-e-r. 12 record labels and artists as well? 13 Q. And where do you work? 13 A. They do. 14 A. I work for Sirius XM. 14 Q. Could you tell the Judges about that as 15 Q. What is your position at Sirius XM? 15 well? 16 A. I'm the senior vice president and general 16 A. Similar to me, I mean, there is not 17 manager of music programming. 17 enough time in their day typically to devote to the 18 Q. How long have you held that position? 18 demand on their time from the music community who 19 A. Close to 14 years. 19 are, like myself, they are typically seeking 20 Q. And as senior vice president and general 20 exposure for their artists on Sirius XM or looking 21 manager for music programming, what exactly do you 21 to do further promotions with us to, you know, 22 do? 22 enhance the visibility of their artists and do 23 A. I'm responsible for the programming 23 things that could further impact their artists' 24 development and operations of all our 85 music 24 careers. 25 channels that are commercial free and offered to our 25 Q. How long have you been working at Sirius

Page 3341 Page 3343 1 31 and a half million subscribers. 1 XM? 2 Q. And how many people report to you? 2 A. Close to 14 years. 3 A. Approximately 275 people report under my 3 Q. And what was your job experience before 4 responsibility. 4 that? 5 Q. And what is the structure of the music 5 A. Prior to Sirius XM, most of my career was 6 department at Sirius XM like? 6 spent in terrestrial radio for about 10 or 15 years. 7 A. Reporting in to me I have six or seven 7 Most of my career was spent in the markets of New 8 what are referred to as genre heads. Each of them 8 York and Los Angeles. 9 have a specialty in a particular style or two of 9 And my two primary formats I have worked 10 music. 10 in were country music and rock music, but I have 11 In addition to the genre heads, there is 11 also had positions where I had exposure and 12 another couple of direct reports who have more 12 experience to multiple styles of music in my time 13 purely operational roles within the department. 13 previous to Sirius XM. 14 Q. How about Sirius XM's music programmers, 14 Q. Now, you have a binder in front of you, 15 where do they fit into the structure of the music 15 if you could take a look at what is in there under 16 department? 16 tab number 5. It is Trial Exhibit 5. Do you 17 A. So each of our 85 channels has a 17 recognize that document? 18 programmer that is dedicated to and each of those 18 A. Yes, I do. 19 individual channel programmers report up to the 19 Q. What is it? 20 genre heads that I just referenced. 20 A. This is my written direct testimony 21 Q. In your capacity as head of music at 21 submitted in this hearing. 22 Sirius XM, do you have occasion to interact with 22 Q. If you turn to the last page, is that 23 label executives or band managers, artists? 23 your signature? 24 A. Yes, I do. 24 A. Yes, it is. 25 Q. Could you tell the Judges a bit what that 25 Q. We also have two other documents in your

46 (Pages 3340 to 3343) Page 3344 Page 3346 1 binder labeled as Trial Exhibits 15 and 18. Do you 1 everything from pop, to rock, to hip-hop and R&B, to 2 recognize those as your testimony in the SAT I and 2 dance music, to country music and Christian formats, 3 SAT II proceedings? 3 jazz standards, classical, to kids programming. 4 A. Yes. 4 Q. And how does the Sirius XM music lineup 5 MS. SPERLE: At this time Sirius XM 5 compare with what's available on AM/FM radio in 6 offers into evidence Trial Exhibits 5, 15, and 18. 6 terms of breadth of offering? 7 MR. TREPP: No objection, Your Honor. 7 A. Well, in a typical AM/FM market, a 8 JUDGE BARNETT: 5, 15 and 18 are 8 typical local market in America, there is usually 9 admitted. 9 approximately five to 15 music channels available in 10 (Joint Exhibit Numbers 5, 15 and 18 were 10 most local markets in America. So if you compare 11 marked and received into evidence.) 11 that to Sirius XM, we're offering on the music side 12 BY MS. SPERLE: 12 85 channels. 13 Q. Mr. Blatter, there is a number of other 13 There is a much wider range of channels 14 documents that are exhibits in your binder. Do you 14 that are available on Sirius XM than you are going 15 recognize those as different documents that you 15 to get on local AM/FM radio. 16 reference in your testimony itself? 16 Q. And what type of genres are available on 17 A. Can you repeat that? 17 Sirius XM that might not be available on AM/FM 18 Q. The other trial exhibits in your binder, 18 radio? 19 do you recognize those as exhibits and documents 19 A. Well, some of the types of channels that 20 that you reference in your written direct testimony? 20 we offer that you are not going to get on AM/FM 21 A. Yes, I do. 21 radio would be niche genres like bluegrass music or 22 Q. At this time we're going to go by 22 jazz music. And then on the -- there are also 23 category, but at this time I would like to offer 23 mainstream formats that might have used to be 24 into evidence Exhibits 624, 626, 627, and 641, which 24 exposed on AM/FM radio but aren't any more today. 25 I believe are exhibits to which Mr. Trepp and I have 25 So in the country category, as an

Page 3345 Page 3347 1 discussed previously and SoundExchange does not have 1 example, there is a series of more classic-based 2 an objection. 2 country channels that might play music from the '60s 3 MR. TREPP: That's correct, no objection. 3 or '70s or '80s or '90s and those, you know, as I 4 JUDGE BARNETT: So 624, 626, 627, and 641 4 just stated are country, types of country music that 5 are admitted. 5 aren't exposed much at all any more on FM radio 6 (Sirius XM Exhibit Numbers 624, 626, 627, 6 today. 7 and 641 were marked and received into evidence.) 7 Q. Let's take an example. I live in New 8 BY MS. SPERLE: 8 York City. It's major metropolitan area. Are there 9 Q. Thank you. Mr. Blatter, how many 9 any genres of music that I wouldn't be able to hear 10 full-time music channels does Sirius XM offer? 10 in New York on AM/FM that I could hear on Sirius XM? 11 A. Approximately 85 music channels. 11 A. Yes, there are. 12 Q. If we could take a look at what is 12 Q. Can you give me an example? 13 labeled in your binder as Trial Exhibit 601. This 13 A. I mean one such example, one of the 14 was previously admitted into evidence. 14 bigger ones today at least in New York City, one of 15 Do you recognize this document? 15 the most popular genres in music today is 16 A. Yes, I do. 16 alternative rock music. And as big as New York City 17 Q. What is it? 17 is, it is the biggest market in America, there is 18 A. This is a visual display of all of the 18 actually no FM radio station in New York that plays 19 channels that, which are approximately 150, all the 19 alternative rock music. 20 channels that are available on Sirius XM. 20 Q. And why does Sirius XM devote so many of 21 Q. Can you tell the Court a bit about Sirius 21 its channels to these kind of niche genres of music 22 XM's music channels and the range of offerings that 22 when AM/FM doesn't? 23 it has? 23 A. There is two reasons. One is we have the 24 A. Sure. At a broad level, we offer 24 channel capacity to offer a lot more genres in music 25 channels in a variety of categories of music; 25 than what you are going to have on FM radio. But I

47 (Pages 3344 to 3347) Page 3348 Page 3350 1 think perhaps even more important than that is our 1 examples of what those are but I will give you one 2 businesses is a subscription-based business. And so 2 or two right now, is if you were listening to let's 3 we're able to program our channels and package them 3 say our '80s channel, instead of it just being a 4 and present them in a way that is what we believe 4 continuous stream of songs, pop songs from the '80s, 5 our listeners truly want to hear. 5 mixed in there are a lot of, you know, short form 5, 6 On FM radio it is a little bit different 6 10, 15 second elements that play pop culture pieces 7 where the business of terrestrial radio is one that 7 that remind you and give you a little nostalgic 8 is built on advertising. And when you are an FM 8 value of the kind of music you are hearing on the 9 radio programmer, which I have been, you need to not 9 '80s channel, maybe even get a little chuckle out of 10 just cater to the needs of your listeners but also 10 you. 11 catering to the needs of advertisers. 11 So that's part of the packaging on a 12 So you could have a very large audience 12 channel like the '80s channel. There will also be 13 for a channel on FM radio but if there is not enough 13 features we will create to put in the music that 14 advertisers that want to reach that audience, that's 14 might be, you know, if it is a contemporary station, 15 not going to be a successful format on FM radio. So 15 it could be a music news feature that we update 16 I think a lot of the channels we offer on Sirius XM 16 throughout the course of the day that keeps that 17 fall into that category. 17 audience up-to-date on things going in that 18 Q. Now, on page 2 of your written direct 18 particular genre of music. 19 testimony you state that each Sirius XM produced 19 Q. And where exactly does Sirius XM get its 20 music channel offers not just a playlist in a 20 hosts from for its music channels? 21 particular genre or from a particular decade, but a 21 A. They come from a variety of places. Many 22 cohesive experience built around expert programming, 22 of our hosts do have experience in backgrounds and 23 song sequencing, skilled charismatic DJs, and you go 23 working in terrestrial radio prior to joining Sirius 24 from there. 24 XM, but we have also done a lot of organic 25 When you are saying that Sirius XM is not 25 development and trained hosts over the years at

Page 3349 Page 3351 1 just a playlist, what are you referring to? Can you 1 Sirius XM. 2 explain how that kind of differs from streaming 2 And the third sort of major source of 3 services? 3 talent for us are just artists. There is a lot of 4 A. Sure. One thing that I say to my 4 former or current musicians that host their own 5 programming team on a fairly regular basis is that 5 shows on Sirius XM. 6 we're not just in the curation business, we're in 6 Q. Is there any sort of specialty music 7 the entertainment business. And what I mean by that 7 programming that Sirius XM offers? 8 is, yes, it is important that my programmers select 8 A. Yes, there is. 9 and sequence the right songs for the channel or 9 Q. Can you tell the Court a bit about it? 10 channels they are responsible for, but what is even 10 A. Sure. So in addition to the regular 11 more important to us is to make sure that we're 11 programming that we offer on Sirius XM, we do spend 12 packaging and presenting those songs in a way that 12 a lot of time trying to think of creative ways to 13 is truly emotionally engaging for that audience. 13 further engage with our audience, so we have done 14 Q. And when you say packaging and presenting 14 things like develop a series called Town Hall 15 it in an engaging way, what are you referring to? 15 Series, which is an opportunity for our subscribers 16 A. I am referring to a number of things. 16 to sit in a studio or room with an artist and 17 First is most of our channels have on-air hosts who 17 participate in a Q-and-A session with that artist. 18 present that music to the audience. Those on-air 18 We have another series called The Artist 19 hosts have a lot of knowledge and credibility and 19 Confidential Series, which is a similar type 20 authenticity of the kind of music they play. 20 environment, but in this environment it is usually 21 And so that's an important aspect to what 21 one of our hosts interviewing the artist and the 22 we do in presenting our channels. And there is also 22 artist will actually perform their music for that 23 a lot of focus and time and resources spent on what 23 audience in our studios. 24 we will refer to as the interstitial elements played 24 Q. What about live event coverage, does 25 on our music channels. And there is a number of 25 Sirius XM offer that as well?

48 (Pages 3348 to 3351) Page 3352 Page 3354 1 A. Yes, we do. We have over the years now 1 Q. Now, in your capacity as head of music, 2 really ramped up the amount of live programming 2 do you investigate how air play on Sirius XM, what 3 available on Sirius XM, particularly live 3 happens after a track plays on Sirius XM and how it 4 performances. So whether it is smaller acts that 4 affects record sales? 5 are playing in a small club, some venue around the 5 A. Yes, I do. 6 country, to larger acts performing at a major music 6 Q. How do you go about doing that? 7 festival, over the last three or four years now we 7 A. There is two, there is two sources that 8 have created partnerships with pretty much every 8 we look at to gauge the impact that our air play is 9 major music festival in America. 9 having on record sales. The first source is just 10 So festivals that have sort of become 10 our internal scheduling software or a third-party, 11 household names now like Coachella or Bonnaroo or 11 radio air play monitoring service called Mediabase 12 Lollapalooza are broadcast in their entirety on 12 which we subscribe to as well. 13 Sirius XM. 13 So we will look at the air play 14 Q. And is that the kind of offering that 14 statistics on Mediabase and then also look at the 15 consumers can get somewhere else? 15 sales data through a service provided by Nielsen 16 A. There is nobody else in the radio space 16 Soundscan, which gives us access to the digital 17 that has been broadcasting these types of festivals 17 track sales data on a song-by-song basis. 18 like we have, no. 18 Q. And did you include a figure 19 Q. Let's talk about promotion. In your 19 demonstrating the sales data for Nathaniel Rateliff 20 testimony you state bands, label staff, artist 20 and The Night Sweats in your testimony? 21 managers routinely recognize the promotional impact 21 A. Yes, I did. 22 of Sirius XM on sales. 22 Q. If we could bring up Trial Exhibit 5, 23 How do you know that? 23 that figure that's shown on page 21. The page is 24 A. There is a few different ways. First 24 restricted but if we can just show the figure. 25 through the interactions with the music industry 25 Great.

Page 3353 Page 3355 1 that I referenced earlier in my testimony, so in our 1 What are we looking at here? 2 regular conversations with record label executives, 2 A. This is a snapshot historical trend of 3 artist managers, sometimes the artists themselves, 3 the sales for the song that we added from Nathaniel 4 they will, you know, tell us firsthand about the 4 Rateliff called S.O.B. on the Spectrum. 5 impact they are seeing of our air play on their 5 And you can see prior to the Spectrum 6 artist's record sales and consumption elsewhere. 6 playing the song it was selling approximately 375 or 7 Q. Now, I know you have a number of 7 so digital tracks a week, and then when the Spectrum 8 different examples in your testimony and that's 8 added the song, for the week of June 29th, 2015, you 9 already in evidence, so we don't want to go through 9 can see the very large spike in sales from the 375 10 all of it for the Court, but could you tell us about 10 range to north of 1800 tracks a week. 11 just one example that you discuss how about 11 Q. Now, is Sirius XM taking full 12 Nathaniel Rateliff and The Night Sweats? 12 responsibility for the complete commercial success 13 A. Sure. 13 of Nathaniel Rateliff and The Night Sweats? 14 Q. So how exactly did Sirius XM become aware 14 A. No, we're not. 15 of Nathaniel Rateliff and The Night Sweats? 15 Q. Why not? 16 A. So one of our programmers was invited to 16 A. Well, we certainly played a major part in 17 a showcase for this artist in New York City to come 17 sort of putting Nathaniel Rateliff on the map at a 18 see him perform live. This was in New York City. 18 very, very early stage before anybody else had done 19 And she was very impressed with the artist, and not 19 so, but there were certainly other ways in which his 20 just with the performance, with their music, and 20 music got exposed after the success that he had on 21 decided to give this artist a chance on our Spectrum 21 Sirius XM. 22 channel. 22 JUDGE STRICKLER: What are those other 23 And The Spectrum became the first radio 23 ways? 24 station in America to play Nathaniel Rateliff and 24 THE WITNESS: Well, one of the ways in 25 The Night Sweats in a regular rotation. 25 particular for Nathaniel Rateliff is he did have an

49 (Pages 3352 to 3355) Page 3356 Page 3358 1 appearance I think a few months after The Spectrum 1 problem we had and speculate. 2 started playing him on the Jimmy Fallon Show. 2 Is there anything in your testimony about 3 And then they were, you know, his record 3 that? That's a simple yes-or-no question. 4 label was able to sort of sell the success that they 4 THE WITNESS: I don't believe there is. 5 were having with this song on Sirius XM's The 5 JUDGE STRICKLER: Thank you. 6 Spectrum Channel to other radio programmers which 6 BY MS. SPERLE: 7 then further, you know, drove up his sales with all 7 Q. Now, do you ever receive outreach from 8 the additional promotional exposure they gained. 8 record labels after Sirius XM plays a song or puts 9 MR. TREPP: Objection, Your Honor. 9 it into rotation? 10 That's hearsay. 10 A. Yes, we do. 11 JUDGE BARNETT: Ms. Sperle, why would it 11 Q. And have you seen advertisements that are 12 be appropriate for us to admit this as hearsay? 12 taken out that advertise air play on Sirius XM on 13 MS. SPERLE: Well, Mr. Blatter is talking 13 different tracks? 14 about the statements that he has heard and what he 14 A. Yes. 15 observed in the marketplace in terms of other things 15 Q. Now, if we could take a look at -- and 16 that have gone on and his impression about why 16 only put this up on our restricted screen for just 17 certain advertisements or things were put into 17 the witness to identify the document. If we can 18 evidence. 18 take a look at Trial Exhibit Number 634. 19 JUDGE BARNETT: Sustained. 19 Now, I know that this is restricted so 20 JUDGE STRICKLER: Let me come at it a 20 without revealing any restricted information, could 21 different way so we don't have that problem. 21 you tell us -- could you identify what this document 22 Is there anything else in your testimony, 22 is? 23 in your written testimony, that indicates whether or 23 A. This is an advertisement that was taken 24 not the Nathaniel Rateliff music was being played on 24 out by the Concord Music Group about Nathaniel 25 interactive services or non-interactive services or 25 Rateliff. This was many months after he had now

Page 3357 Page 3359 1 through any other medium at the same time it was 1 gone on to have a lot of success. 2 playing on Sirius XM? 2 In fact, I believe this ad was one that 3 THE WITNESS: I mean, the songs might 3 appeared in Billboard. I believe it was actually on 4 have been in the systems of an interactive service. 4 the back page of Billboard. 5 JUDGE STRICKLER: My question is really, 5 Q. And is this something that was provided 6 just so we don't get into the same problem as 6 to you? 7 before, whether there is anything about the playing 7 A. Yeah, the ad was sent to me also 8 of his music on any other service at the same time 8 separately by the head of marketing of the Concord 9 as Sirius XM, he was being played on Sirius XM? 9 Music Group, Nathaniel Rateliff's label. 10 THE WITNESS: I mean, to the best of my 10 JUDGE BARNETT: Can someone explain to me 11 knowledge there was no other, you know, in that 11 why an ad that was published is marked restricted? 12 early stage, there was no other service that was 12 MS. SPERLE: It is because of the 13 playing his music in any meaningful way. 13 particular note that is shown with it, and the 14 JUDGE STRICKLER: Whether there was or 14 identity of the exact individual and their personal 15 there was not any other meaningful playing of his 15 communication to Mr. Blatter. 16 music, that issue is not addressed in your direct 16 JUDGE BARNETT: Is that confidential or 17 testimony? 17 sensitive business information that this individual 18 In other words, if this music was being 18 works for Concord Music Group? 19 played on Spotify or not, or was being accessed on 19 MS. SPERLE: That isn't, Your Honor, but 20 Spotify, you don't have any data about that in your 20 the fact that it was a personal communication, that 21 testimony? 21 is something that Mr. Blatter receives in the course 22 THE WITNESS: Well, the Spotify one is a 22 of business, that would not be known to the public, 23 little, I mean -- 23 the contents of this we believe is. 24 JUDGE STRICKLER: My question is really 24 JUDGE BARNETT: I give. 25 simple because I don't want you to get into the 25 MS. SPERLE: Your Honor, at this time

50 (Pages 3356 to 3359) Page 3360 Page 3362 1 Sirius XM offers Trial Exhibit Number 634 into 1 services, has any artist or label ever reached out 2 evidence. 2 to you and said play less of our music, we would 3 MR. TREPP: Your Honors, SoundExchange 3 rather have people listen to this on Spotify? 4 objects for several reasons. Taking the note and 4 A. No, that has never happened either. 5 the advertisement separately, the note is plainly 5 Q. Are you familiar with Sirius XM's direct 6 hearsay. You know, this particular note happens to 6 licensing initiative? 7 be from a witness who will be appearing in this 7 A. Yes, I am. 8 proceeding. 8 Q. What has been your role, if any, with the 9 However, notes of this type the Judges 9 direct license initiative? 10 have recognized during the course of Web IV to be 10 A. My role has been an arms-length role. I 11 particularly unreliable. The Judges have found in 11 typically will become involved after a direct 12 past proceedings that this type of anecdotal 12 license deal is struck with a new partner. I will 13 evidence may symbolize nothing other than common 13 be made aware that we now have a new direct license 14 courtesy. 14 deal in place. 15 And without the author of the note here 15 Q. And as head of music programming at 16 to help the Judges make a determination about which 16 Sirius XM, are you aware that one of the selling 17 of the available interpretations of this document is 17 points for direct licenses is that direct licensors 18 appropriate, we don't think it is reliable. 18 could get increased access to your programming staff 19 Putting aside the hearsay question, there 19 and the potential of increased plays on Sirius XM as 20 are, of course, foundation questions as well, and 20 a result? 21 there are also relevance questions. Simply because 21 A. Yes. 22 this type of evidence is anecdotal. 22 Q. Have you seen that happen? 23 JUDGE BARNETT: The objection is 23 A. Yes, I have. 24 overruled. All of your objections go to the weight 24 Q. Do direct licensors receive any sort of 25 the Judges give this particular exhibit. 25 priority in Sirius XM's music programming as a

Page 3361 Page 3363 1 (Sirius XM Exhibit Number 634 was marked 1 result of signing a direct license? 2 and received into evidence.) 2 A. Well, while the merits of the artist and 3 BY MS. SPERLE: 3 song will always come first, we will favor a direct 4 Q. And just looking at the advertisement in 4 license track when the programmer feels that the 5 particular, could you read for the Court just the 5 direct license track would fit just as well as other 6 sentence or point out for the Court where it 6 non-directly-licensed tracks. 7 mentions Sirius XM? 7 JUDGE STRICKLER: So a tie goes to the 8 A. Sure. You know, in this ad it says 8 direct license? 9 Sirius XM launched S.O.B., which was the single that 9 THE WITNESS: In essence, yes. 10 we had added, had launched S.O.B. and it became a 10 BY MS. SPERLE: 11 foot-stompin' classic. 11 Q. Could you give us an example of an 12 Q. And in the course of your career have you 12 instance where, to your knowledge, Sirius XM might 13 seen other record labels advertise air play on 13 have taken a chance on a track because it was from a 14 Sirius XM? 14 direct licensor, where maybe Sirius XM otherwise 15 A. Yes, I have. 15 wouldn't have at that time? 16 Q. Now, your testimony talks a lot about 16 A. Sure, one such example of that is Andy 17 people in the music industry and their 17 Grammer. 18 communications to you regarding air play, and 18 Q. And what happened with Andy Grammer? 19 promotion on Sirius XM. 19 A. Andy Grammer is on a label called S-Curve 20 Have you ever heard the opposite? And 20 Records, who we do have a direct license with. One 21 what I mean by that is has an artist or a label ever 21 of the principals of S-Curve Records had a 22 reached out to you and said stop playing our music, 22 relationship with our head of pop programming, Kid 23 it is killing our sales? 23 Kelly. He played the record for Kid Kelly. 24 A. No, that has never happened. 24 And while the song was, at least at the 25 Q. How about with regard to on-demand 25 time, perceived to be kind of quirky and maybe not

51 (Pages 3360 to 3363) Page 3364 Page 3366 1 all that mainstream-sounding for a pop channel, Kid 1 increased plays, could Sirius XM play more music 2 Kelly felt strongly enough about it that he decided 2 from direct licensors? 3 to put that song into rotation on our flagship top 3 A. Could we? Yes. 4 40 station, Hits One. 4 Q. Now, and why could you? 5 Q. And did you look at what happened to 5 A. We have the ability to, you know, control 6 sales after Sirius XM added that particular track 6 the playlist that are presented across our 85 music 7 with air play? 7 channels. 8 A. Yes, I did. 8 Q. Now, Mr. White, Mr. George White will be 9 Q. Is that figure included in your written 9 testifying next week with regard to the direct 10 direct testimony? 10 licenses that Sirius XM has, but could you give us a 11 A. Yes, it is. 11 sense of how your programming staff interacts with 12 Q. If we could put that up on the screen. 12 Mr. White's direct licensing staff at Sirius XM? 13 It's Trial Exhibit 5. It's the figure on page 27. 13 A. Sure. So George White and his staff 14 Can you explain for the Judges what we're looking at 14 have, you know, regular, you know, in-person or 15 here? 15 phone conversations with the programmers on my team 16 A. So this is also a historical sales trend 16 talking about which labels they have direct licenses 17 for the song Honey, I'm Good from Andy Grammer. 17 with. 18 Q. And when did Sirius XM approximately put 18 Sometimes they will make introductions, 19 that track into air play? 19 you know, to the programmers from these direct 20 A. Late in December of 2014. 20 license labels. 21 Q. Now, we just talked about Andy Grammer. 21 There is also communication that goes out 22 Are there other examples of instances where Sirius 22 via e-mail from the direct license group that lists 23 XM has given a chance to a track that it otherwise 23 all the record labels that have direct license with 24 wouldn't have if this wasn't from a direct licensor? 24 us, and that is a communication that goes out on a 25 A. Yes, we have. 25 very regular basis.

Page 3365 Page 3367 1 Q. Could you give us a couple other 1 And then the third way is there is, there 2 examples? 2 is in our music scheduling software systems, there 3 A. Sure. One such example is the artist 3 is a flag in the software that indicates all songs 4 Daya, D-a-y-a, who is on a label named 4 that are directly licensed to the company. 5 Z-Entertainment who we have a direct license with. 5 Q. Going back at the very beginning of your 6 Another is a band called All Time Low who 6 testimony, you talked about the limited time that 7 are on Hopeless Records, who is also another direct 7 you and your staff have and the volume of contacts 8 license partner. 8 that you get from record labels. 9 Q. And are there other examples even beyond 9 As head of music, have you seen direct 10 what you have mentioned right now? 10 licensors get increased access to your programming 11 A. Yes, there are. 11 staff and more of their ear as a result of signing 12 Q. Does Sirius XM guarantee direct licensors 12 direct licenses? 13 increased plays on Sirius XM? 13 A. Yes. 14 A. No, we don't. 14 Q. Now, I want to take a moment and do some 15 Q. Why not? 15 kind of housekeeping and cleanup. If you could take 16 A. So the approach we take at Sirius XM to 16 a look at some of the documents that you have in 17 our music programming as a whole is a very human 17 your binder. 18 approach. And we always feel like, you know, we 18 In particular, if you could look at 19 have to believe in the particular artist and their 19 Exhibit 602 to 606, 610 to 611, 614, 616 to 617, 621 20 song first before anything. 20 and 633. 21 And that's always been sort of a 21 Would you agree with me that those are 22 philosophy that we have programmed all our channels 22 communications that you have received from different 23 under. 23 record labels and individuals affiliated with 24 Q. If Sirius XM changed its approach and 24 certain record labels or bands? 25 decided to guarantee certain direct licensors 25 A. Yes.

52 (Pages 3364 to 3367) Page 3368 Page 3370 1 Q. And those are communications that you 1 JUDGE BARNETT: Then we don't have the 2 have received in the course of business? 2 top of the chain in our book. Oh, I see. I'm 3 A. That's correct. 3 sorry. This one is here. I apologize. My mistake. 4 Q. And are these communications all ones 4 You are absolutely right. 5 that you reference and refer to in your testimony? 5 Then you are offering them to establish 6 A. Correct. 6 the fact that there were contacts from these people? 7 MS. SPERLE: Your Honor, at this time 7 MS. SPERLE: Yes, and what they -- 8 Sirius XM offers Exhibits 602 to 606, 610 to 611, 8 JUDGE BARNETT: Not for anything that is 9 614, 616 to 617, 621 and 633 into evidence. 9 stated in these e-mails? 10 MR. TREPP: Your Honor, if I may just for 10 MS. SPERLE: Right. Just for the fact 11 the record, we make the same hearsay and relevance 11 that these were what the record labels, the people 12 objections. 12 from these record labels expressed to Mr. Blatter 13 We would also like to note that Mr. 13 and his programming staff. 14 Blatter is not the proper sponsoring witness for the 14 JUDGE BARNETT: That's the contents of 15 record company communications upon which he is 15 the e-mails and that's what's not admissible. But 16 relying. In fact, these arguments are very similar 16 the exhibits that you enumerated can be admitted but 17 to arguments that were raised during the examination 17 not for the truth of the matter asserted therein, 18 of Professor Lys and it would seem odd to have it 18 which -- never mind. 19 both ways. 19 (Sirius XM Exhibit Numbers 602 to 606, 20 And additionally with respect, in 20 610 to 611, 614, 616 to 617, 621 and 633 were marked 21 particular, to document 603, 604 and 605, 21 and received into evidence.) 22 SoundExchange notes that Mr. Blatter is not in an 22 MS. SPERLE: Yes, Your Honor. Then we 23 appropriate position to cure a critical foundational 23 have a few others. I will try to rattle through 24 issue, which is that several of the e-mails contain 24 them quickly. 25 unsourced information that SoundExchange will not 25 BY MS. SPERLE:

Page 3369 Page 3371 1 have any opportunity to probe. 1 Q. For numbers 607 to 609, 612 to 613, 615, 2 MS. SPERLE: If I may respond. 2 618 to 620, Mr. Blatter, are these also e-mail 3 JUDGE BARNETT: You may. 3 communications that you have received but at this 4 MS. SPERLE: We're not offering any of 4 time not from participants or witnesses in this 5 these particular documents for the truth of the 5 particular proceeding? 6 matter asserted that whatever the factual assertions 6 A. That's correct. 7 are in these particular e-mail communications are 7 MS. SPERLE: Your Honor, we would offer 8 true. 8 those exhibits into evidence as well, not for the 9 We're offering these as just evidence of 9 truth of the matter asserted, based on Your Honor's 10 the fact that Mr. Blatter receives communications 10 prior ruling. 11 from record labels about promotion and also to go to 11 MR. TREPP: Same objections, Your Honor. 12 the state of mind of the particular -- state of mind 12 Also just noting for the record this particular 13 evidence for the particular record labels who are 13 suite of exhibits also doesn't involve record 14 sending these to him. Not that the statements 14 company representatives who happen to be 15 themselves are true, but the fact that he receives 15 participants or witnesses in this action. 16 them and what they reflect. 16 And SoundExchange, again, has limited 17 JUDGE BARNETT: That would be easier to 17 ability to probe those. Not offered for truth, it 18 prove, Ms. Sperle, if he were an addressee or 18 is okay. 19 courtesy copy recipient or a sender on any of these, 19 JUDGE BARNETT: These exhibits are 20 which he does not appear to be, and at least the 20 admitted not for the truth of the matter asserted 21 ones I have looked at, 604, 605, 606. 21 but merely to establish the fact that there was 22 MS. SPERLE: In those particular ones he 22 contact with Mr. Blatter. 23 is the person who it was sent to at the top of the 23 (Sirius XM Exhibit Numbers 607 to 609, 24 chain, if we take a look. He should be a recipient 24 612 to 613, 615, 618 to 620 were marked and received 25 on each of these e-mail communications. 25 into evidence.)

53 (Pages 3368 to 3371) Page 3372 Page 3374 1 BY MS. SPERLE: 1 Q. So I believe you testified on direct that 2 Q. And, lastly, if you could take a look at 2 you are familiar with Sirius XM's direct license 3 Exhibits 623, 628 to 630, 632, 635, and 637 to 639. 3 campaign, correct? 4 Do you recognize these as press releases, 4 A. That is correct. 5 ads, and other trade publications that you have 5 Q. And I believe you also testified on 6 cited and referenced in your testimony? 6 direct that your understanding is that Sirius XM 7 A. Yes, I do. 7 does not guarantee plays to any of its direct 8 MS. SPERLE: Your Honor, we offer these 8 licensors, correct? 9 into evidence as well. 9 A. That is correct. 10 MR. TREPP: Your Honor, again, these 10 Q. Mr. Blatter, during your time at Sirius 11 documents are hearsay that are being offered for the 11 XM, you have never instructed a Sirius XM programmer 12 truth of the matter asserted, namely the content of 12 to play a track just because the cost of playing 13 the trade publications and periodicals that are 13 that track would be less than the cost of playing 14 cited in the testimony. 14 some other track, correct? 15 We raise the same relevance concerns 15 A. That is correct. 16 particularly because the opinions of a specific 16 Q. And you are not aware of any Sirius XM 17 member of the press corps or trade publication are 17 programmer who has been instructed to play a track 18 not germane to the Judges' attempts to assess the 18 because the cost of doing so would be lower than the 19 quantitative amount of promotion or substitution in 19 cost of playing some other track, correct? 20 this matter. 20 A. That's correct. 21 And the final, the foundation argument 21 Q. Mr. Blatter, I would like to turn to how 22 that SoundExchange will not have an opportunity to 22 Sirius XM programmers made programming decisions as 23 cross-examine the authors of these articles. 23 you described on direct. 24 MS. SPERLE: Your Honor, we're not 24 The music channels on Sirius XM are 25 offering these for the truth of the matter asserted, 25 curated by human beings, right?

Page 3373 Page 3375 1 simply as a reference that these are the materials 1 A. That's correct. 2 that are cited and referenced in Mr. Blatter's 2 Q. It is people who decide which recordings 3 testimony. 3 to play on Sirius XM's music channels, right? 4 And in Satellite II on June 8th, 2012, 4 A. People -- programmers ultimately make 5 these same types of documents were admitted for the 5 those decisions, yes. 6 truth of the matter by Your Honors in that 6 Q. Sirius XM programmers select songs based 7 proceeding, and that's at around pages 970 to 975. 7 on their own judgment, right? 8 So there is precedent for admitting these types of 8 A. Yes, but there are factors that go into 9 documents not for the truth. 9 their judgment beyond what they personally believe. 10 JUDGE BARNETT: They will be admitted not 10 Q. Sirius XM programmers may consult data 11 for the truth of the matter asserted, but simply to 11 when selecting songs, correct? 12 establish that -- simply admitted as catalysts for 12 A. Correct. 13 the material that Mr. Blatter included in his 13 MR. TREPP: And, Your Honors, I'm afraid 14 written direct testimony. 14 this brief line of questioning will involve some 15 (Sirius XM Exhibit Numbers 623, 628 to 15 Sirius XM restricted information, so I would ask 16 630, 632, 635, and 637 to 639 were marked and 16 that we very briefly close the courtroom, not to 17 received into evidence.) 17 Sirius XM personnel of course. 18 MS. SPERLE: I have nothing further. 18 JUDGE BARNETT: Okay. Anyone in the 19 MR. TREPP: May I proceed, Your Honor? 19 hearing room who is not permitted to hear restricted 20 JUDGE BARNETT: You may. 20 material, please wait outside, unless you work for 21 CROSS-EXAMINATION 21 Sirius XM. 22 BY MR. TREPP: 22 (Whereupon, the trial proceeded in 23 Q. Good afternoon, Mr. Blatter. How are 23 confidential session.) 24 you? 24 25 A. Good afternoon. 25

54 (Pages 3372 to 3375) Page 3383 Page 3385 1 O P E N S E S S I O N 1 would expect they would always be communicating with 2 BY MR. TREPP: 2 Mr. White or members of his group. 3 Q. Mr. Blatter, I believe you testified on 3 Q. With respect to tracks that are submitted 4 direct that you have never negotiated a direct 4 to Mr. White's direct licensing group, you don't 5 license for Sirius XM, correct? 5 know whether Mr. White's group sends along each 6 A. That's correct. 6 track that it receives to your programmers, correct? 7 Q. And your programmers don't negotiate 7 A. I mean, in my role I wouldn't know that 8 direct licenses for Sirius XM, correct? 8 information. 9 A. They do not. 9 Q. In fact, you don't know how many of the 10 Q. There is a separate group of employees 10 tracks that direct licensors submit for Sirius XM's 11 that is responsible for negotiating direct licenses, 11 consideration are actually reviewed by your 12 right? 12 programmers, right? 13 A. That is correct. 13 A. No, that would be out of the scope of my 14 Q. And that's the group that is run by Mr. 14 responsibility to know that. 15 White? 15 Q. So, Mr. Blatter, I would like to turn to 16 A. Correct. 16 how Sirius XM evaluates whether to play songs from 17 Q. Now, Mr. Blatter, it is your testimony 17 directly-licensed labels. 18 that a benefit of direct licenses is that it gives 18 You testified on direct examination that 19 record labels greater access to Sirius XM 19 when Sirius XM is deciding whether to play a 20 programmers, correct? 20 directly-licensed track, the merits of the artist 21 A. Yes. 21 and the song always come first, right? 22 Q. But you are not personally aware of any 22 A. Correct. 23 information on the percentage of direct licensors 23 Q. In Judge Strickler's words, the tie goes 24 who actually submit tracks to Sirius XM for 24 to the direct licensor, correct? 25 consideration for air play on its satellite radio 25 A. Correct.

Page 3384 Page 3386 1 service, right? 1 Q. And you communicate that guidance to your 2 A. I don't recall any report ever being 2 direct reports, correct? 3 generated of such. 3 A. Yes, I do. 4 Q. Now, with respect to accessing 4 Q. And your direct reports communicate that 5 programmers, some direct licensors are put into 5 guidance to Sirius XM programmers, right? 6 direct contact with your programmers, right? 6 A. Yes, they do. 7 A. Yes. 7 MR. TREPP: And, Your Honor, I am at this 8 Q. Other direct licensors are instead in 8 point going to switch gears, so if it would be an 9 contact with Mr. White's group, right? 9 appropriate time for a break. 10 A. I mean, I can't speak to the relationship 10 JUDGE BARNETT: That would be lovely. 11 that every direct licensor has with the company. I 11 Thank you. We will be at recess for 15 minutes. 12 know that when, you know, George White or somebody 12 (A recess was taken from 2:52 p.m. to 13 in his group makes an introduction to us from 13 3:16 p.m.) 14 somebody who is now a direct license partner, we 14 JUDGE BARNETT: Please be seated. 15 will then begin to grant that label more access to a 15 BY MR. TREPP: 16 particular programmer or series of programmers. 16 Q. Welcome back, Mr. Blatter. 17 Q. But, Mr. Blatter, my question is whether 17 A. Thank you. 18 or not it is true, and I think we can agree that it 18 Q. So I'd like to switch gears for a moment. 19 is, that some direct licensors are in contact 19 I believe you testified on direct that you have 20 directly with Mr. White's group, not with your 20 worked at Sirius XM since 2003, correct? 21 programmers, right? 21 A. That is correct. 22 A. Well, I would expect they would be in 22 Q. And prior to that you worked in 23 touch with Mr. White's group. I mean, they are the 23 terrestrial radio, right? 24 ones that had, you know, brokered the direct 24 A. Correct. 25 license, you know, relationship with that label so I 25 Q. You have never worked at a digital music

55 (Pages 3383 to 3386) Page 3387 Page 3389 1 service that provides on-demand streaming, correct? 1 don't know whether record company personnel 2 A. I have not. 2 correspond with personnel at interactive streaming 3 Q. And you have never worked at a digital 3 services, right? 4 music service that provides non-interactive 4 A. I've spoken to record company executives 5 streaming, correct? 5 that have told me they have interacted with 6 A. Correct. 6 on-demand streaming services. 7 Q. Since coming to Sirius XM, you have had 7 Q. Well, you don't have a basis for knowing 8 conversations with record labels about the labels' 8 the total volume of promotional activity that record 9 desire for air play on Sirius XM, right? 9 companies direct to interactive services, correct? 10 A. Yes. 10 A. No, I have never calculated the total 11 Q. But nearly all of your conversations with 11 volume of it, no. 12 record companies are about their desire to seek air 12 Q. And you don't have any basis for 13 play on Sirius XM radio, right? 13 determining the volume of promotional activity that 14 A. That's typically their focus of their 14 record companies direct to any streaming service 15 discussion with me. 15 other than Sirius XM, correct? 16 Q. Your conversations haven't concerned 16 A. I have never had any reason to sort of 17 whether record labels also seek air play on digital 17 calculate the volume of interactions that record 18 services like Pandora, correct? 18 companies are having outside of interactions with 19 A. Not typically. 19 Sirius XM. 20 Q. And your conversations haven't concerned 20 Q. And, Mr. Blatter, you don't have an 21 whether record labels pitch interactive services 21 exhaustive knowledge of everything that record 22 about placement on curated playlists, correct? 22 companies do to promote or market their artists, 23 A. Not typically. 23 right? 24 Q. Placement on curated playlists isn't 24 A. What do you mean by "exhaustive 25 something that you hear about during the course of 25 knowledge"?

Page 3388 Page 3390 1 your conversations with record labels, right? 1 Q. You don't know everything that record 2 A. I do from time to time. 2 companies do to promote or market their artists, do 3 Q. So you can't say definitively how 3 you? 4 frequently record companies pitch interactive 4 A. I mean, I wouldn't profess to know 5 services about placement on curated playlists, 5 everything they do but I think I have a pretty 6 right? 6 strong knowledge and level of experience in knowing 7 A. Can you repeat that, please? 7 how record companies do market their artists outside 8 Q. You can't say definitively how often 8 of what they do with Sirius XM. 9 record companies pitch interactive services about 9 Q. But, Mr. Blatter, sitting here today, you 10 placement on curated playlists, right? 10 wouldn't claim to know that -- strike that. 11 A. That's not something that I'm really, you 11 Sitting here today, Mr. Blatter, you 12 know, keeping tabs of. 12 wouldn't claim to know everything that a record 13 Q. And you don't know whether record company 13 company does to promote any particular artist, 14 personnel correspond with personnel at interactive 14 correct? 15 streaming services either, right? 15 A. There are certain artists where I would 16 A. I would imagine they might but, again, it 16 know pretty well how else a record label is 17 is not something that I'm keeping tabs on. 17 marketing a particular artist. I can't state -- I 18 Q. Just to make sure the record is clear, 18 can't say that I know that for every artist they 19 you don't know whether record company personnel 19 market, but there are certain artists that I do. 20 correspond with personnel at interactive streaming 20 Q. So for the certain artists that you claim 21 services, right? 21 to have some awareness of what record companies are 22 A. I would imagine they do. I just don't -- 22 doing with respect to marketing or promotion, the 23 I don't know what every record company executive 23 basis for that knowledge would be what you have 24 does when they are not talking to me. 24 learned from the relevant record company, correct? 25 Q. So you imagine it might occur but you 25 A. Not just the record company, but often

56 (Pages 3387 to 3390) Page 3391 Page 3393 1 the artist manager as well. 1 A. Is it -- sorry? 2 Q. So the basis for your knowledge about 2 Q. Is it something that you claim to watch 3 what record companies do to market or promote would 3 closely? 4 be a third-party, including a record company, the 4 A. Yes. 5 artist, or the artist manager, correct? 5 Q. Is it something that you could quantify 6 A. Those would typically be the people that 6 for the Court? 7 I would be speaking with, yeah. 7 A. I have never had any reason to quantify 8 Q. You don't attempt to track the number of 8 it. If I were asked to quantify it, I could 9 identity of artists that you believe have been 9 probably go back and do so. 10 broken by interactive streaming services, do you? 10 Q. You have made no effort to quantify it in 11 A. I'm sorry, could you just repeat that? I 11 your written direct testimony, correct? 12 didn't understand the beginning of your sentence. 12 A. I don't think that was covered in my 13 Q. Of course. You don't attempt to track 13 written direct testimony, no. 14 the number and identity of artists that you believe 14 Q. Mr. Blatter, you don't know how many 15 have been broken by interactive services, correct? 15 unique tracks Sirius XM played in 2016, right? 16 A. What do you mean by "track"? 16 A. I do not. 17 Q. You don't keep track of artists who you 17 Q. And you don't know how many unique tracks 18 believe have been broken by interactive services, 18 Sirius XM played in any other year, correct? 19 right? 19 A. That's not something that I track. 20 A. We might not have a sort of written 20 Q. You don't know how many new releases 21 document of that, but it is something that we watch 21 Sirius XM played in 2016, right? 22 very closely and have a very good sense as to what 22 A. I don't know a precise number, no. 23 artists might or might not be breaking through an 23 Q. And you don't have a precise number of 24 interactive service. 24 new releases that Sirius XM has played in any other 25 Q. So, Mr. Blatter, I took your deposition 25 year, right?

Page 3392 Page 3394 1 in this matter. You recall that, correct? 1 A. No, but I should say that, you know, the 2 A. Yes. 2 way we program our channels at Sirius XM is we 3 Q. And you gave truthful answers, correct? 3 program each of our channels on a channel-by-channel 4 A. Yes. 4 basis. 5 Q. So I'd like to direct your attention to 5 So we have no real reason to be 6 page 117, line 12 of that deposition. It should be 6 aggregating, as an example, the number of new songs 7 on the screen for you. 7 added to all of our channels over a given period of 8 A. I want to see the context in which this 8 time. Everything is done on a channel-by-channel 9 is in. I will turn to the page. You said page 117, 9 basis. 10 line 12? 10 Q. Understood. But, Mr. Blatter, my 11 Q. Correct. 11 question is just, do you know the number of new 12 Line 12 I asked: Do you track acts that 12 releases that Sirius XM played in 2016 or in any 13 you believe have been broken by other music 13 other year? 14 services? 14 A. No. In my role I really have no reason 15 Line 15 you answered: No. 15 to be tabulating such a number. 16 Correct? 16 Q. And, Mr. Blatter, you have never seen a 17 A. Yes, I did. 17 survey of Sirius XM subscribers that asks Sirius XM 18 Q. Turning to a different subject, Mr. 18 subscribers whether hearing a song on Sirius XM 19 Blatter -- 19 caused them to go out and purchase an album, 20 A. Wait. Could I just clarify that? I 20 correct? 21 think I said, too, I don't have a tracking document 21 A. I don't recall that being asked. It 22 for it but it is something that we watch closely, 22 might very well have been something maybe I don't 23 just to be clear. 23 recall seeing or just wasn't privy to that research. 24 Q. Is it something that you claim to watch 24 Q. Your testimony doesn't cite any survey of 25 closely? 25 Sirius XM subscribers that asked Sirius XM

57 (Pages 3391 to 3394) Page 3395 Page 3397 1 subscribers whether hearing a song on Sirius XM 1 JUDGE STRICKLER: Where the artist turned 2 caused them to purchase an album, correct? 2 you down. 3 A. No, but I should say we really, don't 3 THE WITNESS: That might happen, you 4 really have much of a reason to ask that in a survey 4 know, from a timing perspective, but I can tell you 5 method when we can just actually put a song on the 5 when it comes to the pop-up channels, similar to the 6 radio, expose it to our millions of subscribers, and 6 artist channels that we offer today, they have 7 then go look at the corresponding sales data. 7 become very sought after. 8 Q. Mr. Blatter, I'd like to direct your 8 In certain cases it has sort of become a 9 attention to paragraph 30 of your written direct 9 badge of honor to get your own artist channel on 10 testimony. It should be in the binder there. 10 Sirius XM. 11 Now, in this paragraph, Mr. Blatter, you 11 JUDGE STRICKLER: Thank you. 12 state that artists assist Sirius XM to create pop-up 12 BY MR. TREPP: 13 channels, correct? 13 Q. Mr. Blatter, could you please direct your 14 A. That's correct. 14 attention to page 2 of your written direct 15 Q. You state that pop-up channels are 15 testimony, specifically to the first bullet on that 16 short-term artist-hosted channels that showcase the 16 page. 17 artist's catalogue, correct? 17 A. I'm sorry, what page are you on? 18 A. Correct. 18 Q. Page 2. 19 Q. Sirius XM's short-term pop-up channels 19 A. Okay. 20 usually last anywhere from three to 60 days, right? 20 Q. And the first bullet, and even more 21 A. Typically, yes. 21 specifically the language stating: "Given our 22 Q. And the reason that Sirius XM runs pop-up 22 prominence in the car, we compete most directly with 23 channels is to help create a better listening 23 traditional terrestrial radio." 24 experience for its subscribers, right? 24 Is that an accurate recitation of your 25 A. Yes. 25 testimony?

Page 3396 Page 3398 1 Q. Mr. Blatter, could you please direct your 1 A. Yes, it is. 2 attention to paragraph 32 of your written direct 2 JUDGE STRICKLER: I just want to follow 3 testimony. 3 up. I apologize for breaking into your questioning. 4 In talking about pop-up channels in this 4 MR. TREPP: Of course. 5 paragraph, you state that: Artists' willingness and 5 JUDGE STRICKLER: Your response to me was 6 strong desire to have their music featured around 6 artists have turned you down, you say that might 7 the clock refutes any contention that Sirius XM air 7 happen, but from a timing perspective. 8 play substitutes for sales or other paid forms of 8 In what sense was there a timing 9 consumption. Correct? 9 perspective when that happened? 10 A. Correct. 10 THE WITNESS: So we might have proposed 11 Q. You don't have any quantitative evidence 11 to do a pop-up channel with a particular artist and 12 to support your assertion that Sirius XM does not 12 they said: Hey, we would love to do a pop-up 13 substitute for sales or other forms of paid 13 channel with you but our new album isn't coming out 14 consumption, right? 14 until the fall and why don't we wait until then. 15 A. Can you repeat that, please? 15 JUDGE STRICKLER: Any other reasons or 16 Q. You don't have any quantitative evidence 16 that was the sum and substance of the timing 17 to support your assertion that Sirius XM does not 17 discussions for the turndown of the pop-up channel? 18 substitute for sales or other forms of paid 18 THE WITNESS: No, there is no other 19 consumption, right? 19 reasons I can think of at this point. You know, 20 A. Correct. 20 there might have been earlier on in our sort of 21 JUDGE STRICKLER: Excuse me, counsel. 21 development as a company. 22 Did you ever propose a pop-up channel to 22 Artists maybe weren't as aware of the 23 an artist and were turned down by the artist? 23 kind of things we do or the impact these channels 24 THE WITNESS: Where the artist turned it 24 could have on their careers. 25 down or we turned down the artist? 25 So we don't really need to sell them as

58 (Pages 3395 to 3398) Page 3399 Page 3401 1 much as we used to any more. So now it has really 1 MR. TREPP: Your Honors, I think we're 2 been more about timing than it is anything else. 2 likely to be in and out of restricted questioning 3 JUDGE STRICKLER: Did an artist ever 3 for the remainder of the fairly brief examination to 4 reject the pop-up channel because they indicated 4 come. So I would ask that we close the courtroom. 5 they preferred to window their new music somewhere 5 JUDGE BARNETT: Is there any way you can 6 else or just let their plays run on some other 6 consolidate restricted and unrestricted? 7 service before they went on a pop-up channel on 7 MR. TREPP: I think I can. 8 Sirius XM? 8 JUDGE BARNETT: Let's do it that way 9 THE WITNESS: That has never come up 9 then. 10 before. 10 MR. TREPP: All right. 11 JUDGE STRICKLER: Thank you. 11 JUDGE BARNETT: We will start with the 12 BY MR. TREPP: 12 restricted portion. 13 Q. So, Mr. Blatter, directing your attention 13 MR. TREPP: We can start with the 14 back to the language highlighted on the screen in 14 unrestricted portion if that's easier for everyone 15 front of you: Given our prominence in the car we 15 involved. 16 compete most directly with traditional terrestrial 16 JUDGE BARNETT: Or we can start with the 17 radio. 17 unrestricted. Okay. 18 The "our" in that sentence would be 18 BY MR. TREPP: 19 Sirius XM, correct? 19 Q. Mr. Blatter, I'd like to turn back to 20 A. Correct. 20 your testimony and more specifically your testimony 21 Q. And the basis for your testimony is 21 about the sound recording performance complement. 22 research regarding what Sirius XM subscribers 22 In your written direct testimony you note 23 listened to prior to subscribing to Sirius XM, 23 that record companies waive restrictions on the 24 right? 24 sound recording performance complement, correct? 25 A. Well, I don't think I stated that was the 25 A. Yes.

Page 3400 Page 3402 1 basis for it in my testimony. 1 Q. But as far as you know, record companies 2 Q. But it is the basis for your testimony, 2 that waive restrictions on the sound recording 3 correct? 3 performance complement do not waive royalties, 4 A. I don't know if I would sort of cite 4 correct? 5 specific research about that. I think it is pretty 5 A. That is correct. 6 widely known that most Sirius XM subscribers are 6 Q. Could you please direct your attention to 7 listening to us in their vehicles. 7 paragraph 37 of your written direct testimony. This 8 We do know that once they start listening 8 paragraph concerns an exclusive channel devoted to 9 to us in their vehicles, Sirius XM dominates the 9 Garth Brooks, right? 10 amount of listening on a percentage basis relative 10 A. Yes, it does. 11 to the other options they have in their vehicle. 11 Q. Sirius XM views Garth Brooks as a unique 12 Q. James, could you bring up page 188, line 12 artist, right? 13 9 through 17 of the deposition, please. 13 A. Repeat that? 14 Line 9: 14 Q. Sirius XM views Garth Brooks as a unique 15 "Question: What is your basis for the 15 artist, right? 16 testimony that Sirius XM competes most directly with 16 A. Yes. 17 traditional radio?" 17 Q. And Sirius XM was thrilled to have Garth 18 Line 12: 18 Brooks running his own full-time channel at Sirius 19 "Answer: Based on research that I've 19 XM, right? 20 seen, when Sirius XM subscribers are asked about 20 A. Yes. 21 their habits prior to subscribing to Sirius XM, the 21 Q. In fact, according to Sirius XM, Garth 22 strong majority of our subscribers were previously 22 Brooks' addition to Sirius XM made its country 23 spending the time listening to terrestrial radio." 23 offerings second to none, right? 24 That was your answer, correct? 24 A. Yes. 25 A. Yes. 25 Q. We can agree, can't we, that if the

59 (Pages 3399 to 3402) Page 3403 Page 3405 1 addition of Garth Brooks made Sirius XM's country 1 but the live performances that might air on other 2 offerings second to none, the addition of Garth 2 digital music service are so few and far between I 3 Brooks benefitted Sirius XM, correct? 3 really have no reason to quantify that. 4 A. Yes. 4 Q. So to be clear, the answer is, no, in 5 Q. Sirius XM hasn't made any attempt to 5 your written direct testimony you have not made any 6 quantify that benefit, has it? 6 effort to assess whether or not Sirius XM offers 7 A. We do from time to time conduct research 7 more live events than other digital music services, 8 where we measure the listenership and passion levels 8 correct? 9 of our music channels. 9 A. Well, like I stated, I have no reason to 10 We did recently conduct such a study and 10 quantify it because it is just not something that at 11 the Garth channel that was launched last year, you 11 least to date the digital music services do a lot 12 know, did exceptionally well. 12 of. 13 Q. Can you please turn to paragraph 33 of 13 MR. TREPP: I think, Judge Barnett, we 14 your indirect testimony. And in this paragraph you 14 are now left with only the restricted portion of the 15 assert that the Billy Joel channel was so successful 15 examination. 16 at driving sales for Joel's concerts that Sirius XM 16 JUDGE BARNETT: Thank you. 17 brought the channel back in January 2016 and again 17 MR. RICH: Can we clarify that Sirius XM 18 in 2016, right? 18 representatives can stay in the room? 19 A. Yes. 19 MR. TREPP: Oh, absolutely. 20 Q. Your basis for asserting that the Billy 20 JUDGE BARNETT: Thank you, Mr. Rich. 21 Joel channel was successful at driving sales for 21 (Whereupon, the trial proceeded in 22 Billy Joel's concerts is information provided by 22 confidential session.) 23 Joel's touring agent to a colleague of yours at 23 24 Sirius XM, right? 24 25 A. Yeah, that was definitely one of the 25

Page 3404 Page 3412 1 factors, yes. 1 C O N T E N T S 2 Q. You have never spoken with Billy Joel's 2 WITNESS DIRECT CROSS REDIRECT RECROSS 3 touring agent about this testimony, correct? 3 JONATHAN BENDER 4 A. Not about the testimony -- I mean, about 4 3164 3223 5 my written testimony? 5 3229 3334 3336 6 Q. Correct. 6 STEVEN BLATTER 7 A. No. 7 3340 3373 8 Q. Mr. Blatter, could you please direct your 8 9 attention to paragraph 65 of your written direct 9 AFTERNOON SESSION: 3299 10 testimony. 10 11 In this paragraph you discuss exclusive 11 CONFIDENTIAL SESSIONS: 3376-3382, 3406-3411 12 live performances broadcast by Sirius XM, right? 12 13 A. Yes. 13 E X H I B I T S 14 Q. And you were aware that some other 14 EXHIBIT NO: MARKED/RECEIVED REJECTED 15 digital music services also air exclusive live 15 JOINT 16 performances, correct? 16 5 3344 17 A. They might from time to time but I don't 17 15 3344 18 think any digital music service is broadcasting live 18 18 3344 19 performances at nearly the frequency that Sirius XM 19 29 3156 20 does. 20 48 3167 21 Q. Have you made any attempt to quantify 21 SoundExchange 22 whether or not Sirius XM offers more live 22 101 3177 23 performances than other digital music services in 23 102 3169 24 your written direct testimony? 24 357 3201 25 A. I didn't in my written direct testimony, 25 454 3212

60 (Pages 3403 to 3412) PUBLj:C VKRSI()N

Before the UNITED STATES COPYRIGHT ROYALTY JUDGES THE LIBRARY OF CONGRESS Washington, D.C.

) In re ) ) DETERMINATION OF ROYALTY ) Docket No. 14-CRB-0001-WR (2016-2020) RATES AND TERMS FOR ) EPHEMERAL RECORDING AND ) DIGITAL PERFORMANCE OF ) SOUND RECORDINGS (PXB IV) ) )

WRITTEN DIRECT TESTIMONY OF MICHAEL HERRING (On behalf of Pandora Media, Inc.) Introduction

1. My name is Mike Herring. I am the Chief Financial Officer of Pandora Media,

Inc. ("Pandora" or "the Company"). I report directly to Pandora's CEO Brian McAndrews, and I am a member ofthe Company's executive committee. I have served as CFO of Pandora since I joined the Company in February 2013. I am a Certified Public Accountant, and I graduated with a Bachelor ofArts in Economics and Political Science from UCLA in 1991.

2. Before joining Pandora in 2013, I served as Vice President of Operations at

Adobe Systems Incorporated, the well-known computer software company. I joined Adobe in

2009 from Omnitiue Inc., a company specializing in website analytics and online marketing automation. Beginning in 2004, I was Omniture's Executive Vice President and ChiefFinancial

Officer. Before Omniture, I served as ChiefFinancial Officer of MyFamily.corn (now

Ancestry.corn), having joined that company in October 2000. I came to MyFamily.corn from

ThirdAge Media, an Internet company where I served as Vice President, Finance, and before that

Anergen Inc., a Silicon Valley biotech firm, where I served as Controller. 3. As ChiefFinancial Officer ofPandora, I oversee a staff ofnearly 200 people who are responsible for all aspects ofthe Company's finances and accounting, including reporting to the Securities and Exchange Commission ("SEC") and investor relations. The following departments report to me: Business Solutions, Facilities, Finance, Internal Audit, IT Support,

Legal, Tax k, Treasury, and Technical Operations departments. I also play an active role in planning the general business strategy ofPandora. Most notably here, the Business Affairs function — which executes our content licensing strategies, including our relationships with music publishers, record companies, and recording artists — reports directly to me.

4. My testimony below addresses three main topics:

a) The impact ofsound recording royalties on Pandora'sfinances. Pandora has suffered cumulative losses ofmore than since 2005, and has yet to see its first profitable year. A significant contributing factor to those losses is Pandora's staggering sound recording performance royalties, which have totaled more than $ 1 billion since the company launched in 2005. As I set out in my testimony below, those royalties have been generated, not under the statutory license fees set in the )Feb II and 8'eb III CRB proceedings but, rather, pursuant to significantly lower rates negotiated under the auspices ofthe Webcaster Settlement

Acts of2008 and 2009. The stark reality is that ifPandora had paid at the higher 8'eb II and

Web IIIrates, without anything else changing, we would have sustained an estimated $S00 million in additional losses (in other words, cumulative losses ofmore than $ 1 billion) through the end of 2014 and almost certainly would have been forced out ofbusiness years ago. To have stayed in business at those higher rates, Pandora would have had to take drastic measures that would have deprived it ofthe ability to make the necessary investments that have allowed

Pandora to grow its revenue &orr in 2005 to a projected for 2014 and to build a business that has generated enormous royalties for sound recording copyright owners and recording artists. These measures almost assuredly would have included significantly curtailing listening on the service to manage Pandora's per-play royalty expenses (beyond such steps already undertaken in the past), which would have resulted in dramatically lower royalty payments to the record industry and reduced revenue to invest in operating expenses to grow the business. I estimate that the best Pandora could have grown, if subject to the higher 8"eb II and

8'eb III rates, is halfas large as we are today — a level of activity that would have generated, at best, only 40% as much in sound recording royalties. That is, at these higher rates, record labels and recording artists would have been over a halfbillion dollars worse off in the short-term, and much worse off — hundreds of millions of dollars per year in foregone royalties — in the long- term. Going forward, were the Copyright Royalty Judges to set rates for 2016-2020 anywhere comparable to those set in the 8'eb IIIproceeding (currently 0.23 cents per play), Pandora would face the same types of challenges resulting from its diminished ability to invest in the growth and monetization of its audience. We recognize that it is not the Judges'or the record industry's) responsibility to ensure our survival — and we are not looking for a handout — but it is also clear that Pandora would not be a "willing buyer" at rate levels that would make it impossible to for us

(or any other webcaster) to have the prospect of growing into a profitable business.

b) Pandora's creation ofan Internet radio advertising market. In an effort to earn as much revenue as possible to cover its massive royalty costs — more than a billion dollars in the past seven years — Pandora has worked to create, essentially from scratch, a new market for audio Internet radio ads — a long, difficult, and expensive process that is only now coming to fruition. With an ever growing ad sales force, revamped sales platform and infrastructure, and continued education of media buyers, we are now competing directly for advertising dollars that otherwise would go to terrestrial radio statLons. Pandora has invested tens ofmillions of dollars to begin moving ad dollars from a platform that pays no royalties to record companies or artists to a platform that in 2014 alone expects to pay more tha in such royalties — and will pay even more as we continue to grow and improve our ability to monetize that growth. In order to do so, however, royalties must be set at reasonable levels that will allow us to continue making necessary investments in growing our ad sales team, innovating advertising products, and increasing the efficiencies of our ad-delivery technology to deliver a revenue stream sufficient to support a reasonable royalty to copyright holders.

c) I'undora-Merlin Agreement. After months of intense negotiation, Pandora recently concluded its first direct license agreement with Merlin, a global association of independent record labels. That agreement covers the precise performance rights and ephemeral reproduction rights covered under the Section 114 and 112 statutory licenses. I describe the negotiation and main deal points ofthe Merlin Agreement below.

The Outsized Imnact of Sound Recording Rovalties on Pandora's Finances 5. Since its inception in 2005, Pandora's revenues have grown trom underI per year to a projected for 2014.'pproximately S0% of our revenues come through the sale of advertising, with the remaining 20% generated through the sale of our subscription service, Pandora One, which does not have advertisements. To this day, however, the Company's costs exceed the revenues we are able to generate. As detailed in our most recent form 10-Q filed with the SEC (Pandora Exhibit 8), Pandora had a net operating loss of $11.7 million for the three-month period ending June 30, 2014. In the first quarter of 2014, Pandora

'ince Pandora's Initial Public Offering in June, 2011, Pandora has filed 10-K forms on an annual basis and 10-Q forms on a quarterly basis with the SEC. These forms contain, among other things, Pandora' financials, including its revenues and expenses. The most recent 10-K form filed with the SEC is part of Pandora's 2014 Annual Report, which is attached hereto as Pandora Exhibit 6. The 10-Q forms for the first two quarters of 2014 are attached hereto as Pandora Exhibit 7 and Pandora Exhibit 8. experienced an even greater net operating loss of $28.9 million. Since Pandora was launched in

2005, the Company has amassed a cumulative net loss ofmore than

6. Pandora's single most significant cost is the royalties paid to SoundExchange to publicly perform sound recordings, currently accounting for half its total costs and half its revenues. Figure 1 below, covering the period 2007-2014, reflects this reality:

Fig. 1 RESTRICTED]

By comparison, Pandora currently pays the performing rights organizations (ASCAP, BMI, and SESAC) and music publishers approximately ~ ofrevenue to publicly perform musical works. Pandora's remaining costs can be broadly grouped into the following categories: (i) sales and marketing costs; (ii) product development costs; and (iii) general and administrative costs. These cost categories are not limited to the general categories of costs that any music service must cover, but also the unique investments that Pandora has made that differentiate it from its closest competitors — terrestrial radio, Sirius XM, and other non-interactive webcasters. For example, we have spent more tha to date in developing and refining the Music Genome Project, including, among other things, the work done to develop the MGP and the 30 full-time musicologists who analyze the music that Pandora offers. In addition to developing the MGP, Pandora has spent more tha creating and refining the necessary infrastructure, hardware, and software (including our playlist creating algorithms) to offer a world-class radio product for each of our approximately 77 million active users. 'n this and all other figures, the financial information depicted for 2014 is estimated based on the first eight months ofthe year. 7. Because of its sound recording performance royalty burden, which imposes a payment obligation with respect to every listener play, Pandora has repeatedly been forced to moderate the growth of its listener hours. For example, in July 2009, listeners to our advertising- supported desktop service were limited to 40 hours of listening per month, a cap that was not lifted until September 2011. In March 2013, Pandora temporarily limited listening hours on its mobile platform when mobile usage increased so drastically that we could not sell enough advertising inventory to cover the royalty costs associated with the extra hours. These caps on listening are blunt, but effective, instruments to curtail Pandora's royalty obligations. For example, the 2013 imposition ofthe mobile listening cap had the effect of reducing Pandora' royalty obligations by nearly ~. They also, however, benefit neither Pandora nor the record industry over the longer term.

8. The sound recording payments reflected in Figure 1 above were those made to

SoundExchange pursuant to a 2009 agreement calling for royalties at substantially lower levels than those set by the Judges in the 8'eb II and F'eb IIIproceedings. Had these rates not been available to Pandora, and had we been remitted to paying at the 8'eb II and 8'eb IIIrates, our financial picture would have been substantially worse. Pandora's sound recording royalty expense alone, barring any other changes, would have eaten up more than 80% of its entire revenue base — and an additional $800 million in expenses — over the 2007-2014 period, as represented in Figure 2 below. Indeed, in both 2008 and 2012, the royalty rates set in the Web II and 8'eb III proceedings would have generated for Pandora a sound recording royalty expense accounting alone for more than 100% of our revenues. Fig. 2 [RESTRICTED]

9. 'f Pandora had been able to survive at all under the 8'eb II and 0'eb III rates„ it would have done so at great cost to building a viable long-term business. Among other measures, Pandora would have been forced to more severely limit the amount of listening on the service to reduce its royalty obligation; this, in turn, would have led to fewer total listeners, fewer listening hours against which advertising could have been sold, and much less investment in our advertising infrastructure, compounding the revenue losses from decreased listening. The combination ofthese factors would have created a vicious downward cycle, with the realistic prospect that Pandora would have been unable to acquire a large enough audience to be a meaningful platform on which advertisers would want to spend carefully allocated advertising dollars — not to mention being unable to fully invest in a sales force adequate to bring in those dollars.

10. If the rates set in this proceeding for the 2016-2020 license term were to be set at levels comparable to the Web III rates, it would have a similarly crippling effect on Pandora' finances that would force Pandora to dramatically reduce listening, which would reduce royalties paid, both in the near and long-term. While our advertising rates and revenues are increasing

steadily (as I discuss below), that revenue growth would be far outstripped by the costs we would

incur from royalty rates set at Web III levels.

. It defies logic to suggest that we

would willingly negotiate royalty payments that would cause us to continue to sustain net losses

for many additional years; it is equally implausible that record labels operating in a competitive marketplace would insist on royalties at a level that would jeopardize the continued flow of

enormous royalties from by far the largest webcaster in operation.

Pandora's Contributions to the Internet Radio Audio Advertisin Market

11. Pandora derives more than 80% of its revenue from the sale of advertising.

Pandora has consistently sought to maximize advertising revenue by expanding the types of advertisements that we offer, and expanding our sales and sales marketing staffto compete for advertising spending that would otherwise go elsewhere, including to traditional radio stations.

As I describe in this section, we are seeing impressive gains both in the rates we are able to charge advertisers and in our overall ads sales levels, but it is an ongoing process that requires significant and ongoing investment — in infrastructure, engineering, and above all ad sales staff- to maintain our upward trajectory. To the extent we experience a rapid increase in our royalty payments, it necessarily will impede our ability to make such investments and grow our revenues in a way that ultimately will benefit both Pandora and sound recording owners and artists.

12. At the time of Pandora's launch, most Internet radio services relied primarily on display ads (banner ads and the like) just like any other Internet-based services, and there was no meaningful market for in-stream audio ads on webcasts like one would hear on traditional radio stations. Advertising agencies typically reflected this split, maintaining separate online or

"digital" ad sales apart from the traditional (and much larger) portion oftheir budgets that covered radio, television and print ads. Not only were the digital and radio budgets kept separate, but different personnel at advertisers and agencies tended to handle each segment ofthe ad-buying strategy. Pandora recognized from the beginning that for it to be able to maximize its advertising revenue, Pandora needed to tap into the $ 15 billion radio advertising market, first by selling "national" audio ads (which would run across our entire user base), and eventually, as we gained sufficient scale in individual markets, by competing for local advertising dollars — far away the largest and most lucrative component ofthat $ 15 billion market. Pandora recognized that it would have to endure potentially significant financial losses for a period oftime in order to grow an audience large enough to attract local advertisers and thereby effectively monetize our listener hours and become a profitable business.

13. Advertisers, however, were not familiar with Pandora, where it fit within the radio market, or what role it might play in their radio advertising strategies and budgets. Reflecting the split mentioned above, they tended to lump Pandora together with other digital platforms for visual advertising (such as "banner" or other display ads on Google or Facebook), not audio advertising. Advertisers'esitance to expand their radio budgets to Pandora also was fostered by their inability to place advertisements with Pandora through the automated ad-placement systems to which they had grown accustomed in the terrestrial radio context, as well as a lack of any

'adio advertising is typically divided into three categories. "Network" ads are not geographically targeted, and run on stations nationwide; these ads make up only about $ 1 billion ofthe overall radio advertising market. "Spot" ads are typically purchased by large, national advertisers but are regionally or locally targeted; these ads account for the vast majority of the market, at about $ 10 billion. The remainder of the market is composed of "local" ads, which are typically limited to a single market. To sell spot or local ads, a service must have sufficient audience scale in those given markets to be attractive to advertisers seeking to reach a specific audience. metrics that would allow them to compare Pandora's reach in the same terms traditionally used

to measure the reach ofterrestrial radio stations.

14. The need to develop an audio advertising market for Internet radio intensified

with the introduction ofthe Pandora mobile app in 2008 and the dramatic shift from desktop to

mobile listening that resulted. (Today, more than 80% of our listening is on mobile and other portable devices.) Digital banner ads became less attractive for Pandora's advertisers, since mobile listeners tend not to look at their screens or interact with the service as much as desktop

listeners. Mobile access also led to a massive increase in listening hours. Spending on mobile

advertising, however, lagged behind this shift in audience, making it difficult for Pandora

effectively to monetize this massive change in listener behavior and find buyers for all the advertising space it had to offer.

15. To effectively break into the $ 15 billion radio advertising market (and, more particularly, the local and national "spot" markets comprising $ 14 billion of that total), Pandora needed to build relationships with entirely new advertising buyers and to communicate with them in the language they used when purchasing radio advertising space. One tool that helped us change traditional conceptions was Pandora's development of listener metrics to allow advertisers to compare Pandora's reach to that of traditional radio stations. A consumer research firm called Arbitron (now called Nielsen Audio) tracked and reported listener data on terrestrial radio, essentially showing advertisers how many people (on average) are listening at any given time, and in what local area —. crucial information for an advertiser looking to determine where to place its advertisingdollars. Arbitrondidnottracklistenermetrics forPandora. Itwas effectively impossible for Pandora to break into the radio advertising market without being able to report to advertisers in the language they used to purchase radio spots: "cume" (measuring the

10 total number of unique listeners to a station in a given period of time) and "average quarter hour"

(AQH) (the average number of listeners for a given station in a quarter hour). To that end, in

May 2012, Pandora announced a partnership with Triton Digital, the second largest provider of

radio audience measurement data behind Arbitron. This partnership with Triton took

considerable effort on Pandora's part, but these efforts have borne fruit: it is now possible to

quantify the size and scale of Internet radio at both national and local levels.

16. Another major challenge that impeded Pandora's ability to break into the traditional radio advertising market was the fact that Pandora was not integrated into the ad- buying software platforms used by media buyers and ad agencies, through which approximately

70% of spot purchases are made. We tackled this challenge head-on: in March 2013, Pandora announced that Triton's metrics for Pandora would be integrated into STRATA and Mediaocean, the two most popular audio ad buying platforms, to faciHtate comparisons between different audio advertising options. Earlier this year, Pandora also was integrated into Telmar, the advertising industry's leading strategy and planning platform that is routinely used in connection with the buying platforms. As a result, Pandora is now available on popular strategy and buying platforms alongside its terrestrial radio competitors.

17. A third challenge Pandora faced in tapping into the radio advertising market was its initial inability to get a foothold in the massive market for local advertising (leaving us relegated to the much smaller and less lucrative national network slice ofthe radio advertising pie) because Pandora could not target ads to users in specific locations. We eventually developed the ability to use listener zip codes to track and serve them ads locally. After starting

's Bloomberg.corn recognized when it reported this development in March 2013, the change helped "eliminate[] a hurdle for ad buyers and Pandora." A true and correct copy ofthe article Pandora Gains Access to $14 Billion Radio Ad-Sales Market, published on March 5, 2013 (available at www.bloomberg.corn/news/2013-03-05/pandora-gains-access-to-14-billion-radio-ad-sales-market.html) is attached hereto as Pandora Exhibit 9.

11 with the top ten local radio markets in the first quarter of 2012, Pandora now uses that capability

to sell local ads across virtually all major metro survey areas in the U.S. (We are the top ranked

station in terms of weekly listening in 14 of the top 15 markets.) Pandora's efforts to unlock

local audio advertising are set forth in our most recent investor presentation from the second

quarter of 2014, a true and correct copy of which is attached hereto as Pandora Exhibit 10.

18. In conjunction with the above-described innovations — which came at a cost of approximately ~ since 2011 — Pandora has had to invest heavily in growing its sales organization. Today, just like most of our terrestrial radio competitors, Pandora operates local

sales teams in radio markets, large and small, across the country. As we have penetrated more deeply into local ad sales markets, we have often hired advertising salespeople directly away from local radio stations. Between 2012 and 2013, Pandora grew from about 700 employees to nearly 1,100, and most ofthe new hires were advertising salespeople, including in New York,

Los Angeles, Dallas, San Francisco, Chicago and Seattle; by the end of 2014, we expect to have grown to about ~ employees. We now have local sales forces in 36 of the top radio markets, with plans to continue to invest more deeply in our existing markets and to expand physical coverage to Since Pandora launched, we have spent tens of millions in building this sales force — and a robust sales support organization to assist it. The total budget for our sales organization has increased trom~ ~ in2010t in 2014, and it comprises about 60% of our employees. 19. In addition to investments in our advertising sales force, we have heavily invested in new advertising products. Serving advertisements is not as easy as one might think.

'he growth of this sales organization has been critical to Pandora's ability to gain a foothold in the local advertising market, as was recognized by the New York Times in April 2012. A true and correct copy of the article Pandova Courts Local Advertisers, by Offering Well-Defined Listeners, published in the New York Times on April 15, 2012, is attached hereto as Pandora Exhibit 11.

12 Advertisers want to reach a particular audience at a particular time and in a particular context.

For example, advertisers not only want to reach a particular age and gender demographic, but

also want to know that an advertisement for a competing product was not just served to a

targeted listener. Adding to the complexity of ad delivery, the Pandora service is delivered

through a variety of computing devices, such as PCs, tablets, and smartphones. We have had to

develop technology that allows us to provide the same advertising experience across all ofthese

device types, which have different screen sizes and profiles. In addition, we have developed new

segments of ad products, such as video advertisements and "sponsored" listening, in which a

listener is granted some period of ad-free listening in exchange for engaging with an

advertisement.

20. Our efforts to drive increased monetization of Internet radio through the sale of audio ads are finally beginning to pay off. Our local advertising sales revenue has grown approximately 180% between the first half of 2013 and the same period in 2014, and a substantial proportion of our spot sales are now directly traceable to terrestrial ad budgets.

Pandora has increased not just its overall sales, but the ad revenue it earns for given levels of usage, particularly with respect to mobile listening. Ad RPM (i.e., revenues per thousand listening hours that are attributable specifically to advertising) specific to mobile and other connected devices increased from $22.25 in the second quarter of 2012 to $32.56 in the second quarter of 2013 to $36.00 in the second quarter of 2014 — a nearly 62% increase in just two years. Our total ad RPM (desktop and mobile combined) has grown from $37.89 in the second

'e know this based on the language the buyers use when negotiating price on our advertising space; that is, rather than negotiate based on CPM (cost per thousand impressions) as they would do in the traditional digital space, they are negotiating off cost per rating point, as they would do in the radio space. We can also tell that we are dealing with the same ad buyers as radio based on their titles or roles within their agency or company — for example, our salespeople routinely interact with ad buyers whose job titles are Broadcast Account Manager or similar.

13 quarter of 2013 to $40.11 in the second quarter of 2014. That said, we still have much work to

do. The mobile audio advertising market is still quite nascent, and advertisers'uying on new platforms historically has lagged significantly behind the shift in audience to such platforms.

But we are beginning to see real changes in the understanding of advertisers regarding Pandora' role in the radio market. As these trends continue, we are hopeful that shift will be reflected in improved sell-through and rates closer to those the terrestrial radio industry has been able to achieve.

21. The bottom line is that Pandora has invested tens of millions of dollars to begin moving ad dollars from a platform that pays no royalties to record companies or artists to a platform that, in 2014, will pay more than ~ in such royalties — and will pay even more as we continue to grow and improve our ability to monetize that growth. Pandora's size and popularity is not itself a basis for a major increase in royalty rates. Royalties should instead be set at levels that will enable the unique platform that Pandora has built to be sustainable, that will permit the necessary investments to be made toward that end, and that will thereby best emulate what a willing buyer (Pandora) would be willing to pay willing sellers in a competitive market (record labels and artists in such a market) seeking to continue to benefit from the sizeable and growing royalty streams that only a successful Pandora can afford them.

Pandora's Direct License with Merlin

22. Over the past couple ofyears Pandora has begun to explore negotiating licenses directly with record labels rather than relying solely on the statutory license. As an initial matter, we undertook an investigation into whether it would be possible to perform the sound recordings of certain independent record labels more than Pandora currently did. As a management team,

Ad RPM on desktop, which is a relatively established advertising market, has been over $60 for some time, and was $62.43 in the second quarter of 2014.

14 we wanted to understand whether Pandora could actually realize a financial benefit by securing

performance rights at economically advantageous rates; that is, ifPandora had an economic

incentive to perform certain sound recordings more than others, could Pandora actually perform

such sound recordings more frequently without affecting listenership? We instructed our

Engineering team, specifically our scientists, to design and conduct experiments to answer this

question. These experiments were designed and implemented in the summer and fall of 2013.

Based on the results ofthose experiments, which demonstrated that Pandora could alter the spins of oertain independent labels by ~ or more, we began to formulate a strategy to negotiate direct licenses with record labels.

23. As part ofthis effort I and others at Pandora have had discussions with a number

ofrecord labels regarding the possibility of direct licensing. From our perspective, direct

licensing offers an opportunity that relying solely on statutory licensing does not: letting the marketplace itself determine the reasonable level of fees for sound recording performance rights in an environment in which individual licensors can compete with one another for the opportunity for increased plays on Pandora.

24. On June 16, 2014, Pandora entered into its first direct license, with Music and

Enterta.inment Rights Licensing Independent Network, B.V. ("'Merlin") — a global rights agency that represents thousands of leading independent record labels. The Merlin agreement is attached hereto as Pandora Exhibit 12, and covers the public performance ofrecordings in the repertories ofparticipating Merlin labels, as well as ephemeral copies necessary to make such performances. The Merlin agreement took several months to negotiate, growing from a simple two-page term sheet to the more extensive agreement we eventually executed. Its final form

On July 11, 2014 Merlin and Pandora entered into the First Amendment to the Merlin Agreement. A true and correct copy ofthe First Amendment is attached hereto as Pandora Exhibit 13.

15 reflects substantial give and take from both sides, including a number ofprovisions that Merlin

insisted upon, without which the agreement almost certainly would not have been reached. For

example, our initial proposal, made in November of2013, proposed a

, which Pandora believed would fairly compensate record labels for their sound

recordings while enabling Pandora to grow its business (and, ultimately, pay record labels even

more royalties). We ultimately accepted Merlin's request for

, but succeeded in obtaining various other provisions allowing

Pandora to achieve its targeted

The key terms ofthe Merlin Agreement are set out in the following paragraphs.

25. Term. The Merlin agreement provides for an initial

26. Royalty Payments. The Merlin agreement incorporates a

, in which Merlin labels receive

16 27. Steering.

28. Compensable Performances.

'pecifically, Pandora can increase performances of Merlin-label tracks by as much as without increased payments, which results in a decrease in the effective per-performance payment. Pandora fully intends to increase spins ofMerlin-label tracks by at least the required

17 29.

30. Promotional Benefits. In addition to the price-related terms set forth above, the

Merlin agreement provides covered labels with certain additional promotional benefits. For

example:

Advertising availability:

Fan Outreach:

Bumpers:

18 Metrics:—

Merlin Label Stations:

Pandora Presents/Premieres: 12

Collectively, the cost ofthese promotional benefits is estimated by Pandora to be minimal when compared to the other terms ofthe agreement.

31. The "steering" aspect ofthe Merlin agieement is worthy of special attention.

Unlike with industry-wide agreements at fixed per-performance rates that apply equally to all performances, this agreement provides Pandora with the financial incentive, for the 6rst time, to increase its use ofMerlin-label tracks without an increase in fees — thus lowering its per- performance payment. As noted above, in order to ensure there is no degradation to the current listening experience and prior to entering into direct license negotiations, Pandora has, over the

'hese events are described in more detail in the Written Direct Testimony of Simon Fleming-Wood.

19 last 15 months, spent considerable time testing the impact that increasing or decreasing the spins

of individual labels has on the overall quality ofthe service. As detailed in the accompanying

testimony of Stephan McBride, Pandora has considerable flexibility to substitute tracks of one

label for those of others — both independent and "major" labels' ifprovided with the financial

incentive to do so, with no discernible impact on listener satisfaction metrics.

32. Based on this testing, we anticipate that we will be able to take full advantage of

the lower-priced per-performance rates that result from the increased use of Merlin tracks. We

are currently identifying the particular tracks covered by the Merlin agreement and executing the

changes to our playlist algorithms to put such steering into practice on our service. That

identification process should be complete early in the 4th quarter of 2014, at which point we anticipate being able to increase the spina of Merlin-label tracks by ~ (the increase necessary to cover the above-described guarantee) or more, a figure our testing suggests can

easily be one without any impact on listening.

33.

'or exam le, our ex eriments revealed that we could "steer" towards or awa from each independent labels), as well as by well more than without any material impact on user satisfaction or listening patterns. 14

20 34. As ofthe date of this testimony, approximately~ Merlin member labels have

"opted in" to the Merlin agreement. Thus, as oftoday, the Merlin agreement grants Pandora the necessary sound recording public performance and ephemeral reproduction rights — the same rights as are at issue in this proceeding — to play more the~ tracks recorded b~ artists on our service. Before accounting for any "steering" ofplays, the sound recordings covered by the agreement account for approximately ~ of all spina on Pandora. 35. Numbers aside, the Merlin agreement covers recordings by some ofthe most popular and prominent artists played by Pandora, including winners of Grammys and other major record-industry awards. To provide just a few examples:

Merge: This label's catalog includes the album The Suburbs by the acclaimed band Arcade Fire. The Suburbs won the Grammy for Album ofthe Year at the 53rd Annual Grammy Awards in 2011. Arcade Fire's debut album Funeral was nominated for the Grammy for Best Alternative Music Album in 2005. Both Funeral and The Suburbs are certified platinum by the RJAA.

o Beggars: This record company owns and distributes several other labels, including XL Recordings, Matador, and 4AD. The Beggars Group's catalogs cover the popular bands:

o Vam ire Weekend: This band's third studio album, Modern Vampires of the City, won the Grammy for Best Alternative Music Album in 2014. Vampire Weekend's albums Contra and Vampire Weekend are both certified gold by the RIAA, as is their single, "A-Punk."

e interpol: This band's albums Turn ou the Bright Lights and Antics have both been certified gold by the RIAA. They have released three albums with Matador that each peaked at ¹2 on the Billboard Alternative Albums. and Top Rock Albums charts.

21 o The National: This band released two albums with 4AD — Trouble Will Find Me and High Violet — that each peaked at ¹3 on the Billboard 200 chart. Trouble Will Find Me also was nominated for a Grammy Award in the Best Alternative Music Album category in 2014.

The Beggar Group's catalog also includes Belle 8. Sebastian's album Write About Love, which in 2010 reached ¹15 on the Billboard 200 chart, and peaked at ¹2 on the Top Independent Albums chart, ¹3 on the Alternative Albums chart, and ¹4 on the Top Rock Albums chart. All ofthese artists regularly perform to sold-out audiences and frequently perform at major music festivals such as Governor' Ball, Coachella, and Austin City Limits.

~ Epitaph: This label released The Offspring's hit album Smash in 1994. Fueled by hit singles "Come Out and Play," "SelfEsteem," and "Gotta Get Away," the album set an all-time record for most units sold by an independent label band, at 16 million records. Smash has continued to sell consistently well in the twenty years since its release, and has been certified six-times platinum in the U.S. Epitaph's catalog also includes Bad Religion, whose albums Chrishnas Songs and New Maps ofHell both reached ¹7 on the Hard Rock Albums chart.

ANTI-: This sister label to Epitaph released the acclaimed album The Whole Love by Wilco, which peaked at ¹5 on the Billboard 200 chart in 2012 and was nommated for a Grammy Award in the Best Rock Album category. ANTI- also released Tom Waits's album Mule Variations, which won the Grammy Award for Best Contemporary Folk Album in 2000 and was certified gold by RIAA. Waits's albums Real Gone and Bad as Me each topped Billboard's Top Independent Albums chart. Alternative country artist Neko Case released her fifth studio album Middle Cyclone with ANTI-, which peaked at ¹3 on the Billboard 200 chart and topped the Independent Albums Chart in 2009; Case's 2013 ANTI- album The Worse Things Get, The Harder IFight, The Harder I Fight, The More ILove You also topped the Billboard Independent Albums chart and hit ¹5 on the Top Rock Albums chart.

Domino: This label's roster includes Animal Collective and Arctic Monkeys. Domino celebrated its first UK ¹1 single with Arctic Monkey's "I Bet You Look

22 Good on the Dancefloor" in October 2005. Animal Collective's 2009 album Merriweather Post Pavilion topped the year-end "best of'ists of dozens of critics around the country.

Razor dk Tie: One ofthe largest privately-owned independent labels in North America, Razor A Tie has sold over 40 million units and has won multiple Grammy Awards in its more than 20 years in the recording business. Razor k Tie's diverse catalog represents rock artists such as All That Remains and For

Today, as well as world music such as artist Angelique Kidjo, whose album Dj in Dj in won the 2007 Grammy for Best Contemporary World Music Album, and Ladysmith Black Mambazo, who were nominated for a Grammy Award in the Best World Music Album category in 2012. Razor k Tie is also a leader in children's music: their Kidz Bop series has amassed more than twenty ¹1 albums on the Billboard Kids'lbum chart, fourteen Top 10 hits on the Billboard 200 chart, and nine gold records.

Jagjaguwar: Co-founded by Darius van Arman, a member of SoundExchange's board of directors, this label's catalog covers indie rock band Bon Iver, which won the 2012 Grammy Award for Best New Artist and whose album Bon Iver, Bon Iver won the 2012 Grammy Award for Best Alternative Music Album.

ATO: This label's catalog includes the rock band Alabama Shakes, which was nominated for a Grammy Award in 2012 for Best New Artist and for a BRIT Award for Emerging Artist ofthe Year. Alabama Shakes'ebut albion Boys & Girls peaked at ¹6 on the Billboard 200 chart, topped the Independent Albums chart, and is certified gold by the RIAA. ATO has also four albums by rock band My Morning Jacket, including the 2011 album Circuital and 2009 album Evil Urges, both ofwhich were nominated for Grammy Awards in the Best Alternative Music Album category. Soul singer Allen Stone's self-titled album was re-released by ATO in 2012, and peaked at ¹4 of Billboard's Top Heatseekers chart.

36. We are hopeful that the Merlin agreement will be the first of many direct deals that more fully embrace the workings of a competitive marketplace in which labels compete with

23 each other on price to have their works performed on the Pandora service. In fact, we have already started to discuss such possibilities with a number of labels that are not represented by

Merlin. These negotiations, while ongoing, appear to be promising.

Pandora's Pro osed Rates and Terms

37. The Rates and Terms proposed by Sirius XM are intended to conform with

Professor Shapiro's analysis ofthe Merlin Agreement and to make certain other technical and conforming changes, several ofwhich I highlight here:

Definition of "Performance": Because Pandora and other Internet radio providers may offer transmissions to listeners outside the U.S. (where they will be licensed according to the legal requirements ofthe foreign country), the regulations should make clear that only those transmissions to users in the United States are properly compensable under the Section 112 and 114 licenses.

Definition of "Revenue": Consistent with our Merlin agreement, the applicable revenue base is limited to advertising and subscription revenue on channels making eligible transmissions subject to the Section 112 and 114 statutory licenses. Pandora revenue earned from other activities, or channels making only incidental performances of sound recordings, should be (and is) excluded.

Direct-license credit: The proposed regulations provide a credit for directly licensed performances so that the Licensee is not double paying for the same performances (i.e., one payment to the direct licensor, one payment to SoundExchange). Consistent with the Judges'uling in the Satellite II proceeding, in situations where the Licensee is paying under the percent-ofrevenue prong ofthe rate formula, the regulations provide a discount from fees owed proportional to the percentage of Licensee performances that are directly licensed.

e Ephemeral recording fee: Consistent with past proceedings and the Merlin agreement (which has no separate ephemeral recordings fee), Pandora proposes that the royalty payable for ephemeral recordings be included within the Section

24 114 royalty. We have proposed a technical modification to the prior language, which prescribed a royalty for ephemeral recordings to be used "solely to facilitate transmissions for which [Licensee] pays royalties." Because the definition of "Performance" exempts certain uses (for example, incidental uses) from payment, the allowance for ephemeral recordings should not be limited "solely" to paid uses.

Late payment fee: We propose a revision to make clear that in the event the Licensee's payment and statement of account are late, only a single late fee shall be assessed. A single late fee is more than sufficient to motivate licensees to make timely payments and accounting; duplicative payments (which would add up to 3.0% per month, or 36% per year) are unnecessary, and would be unreasonable and usurious.

Statement of account signatiue: SoundExchange has recently taken the position with Pandora that we cannot revise or adjust a previously submitted statement of account because the first version submitted was certified as accurate under the existing signature requirement in the regulations. Although we make every attempt to ensure that our statements of account are true and accurate, that should not prevent us &om revising and resubmitting those statements if, in good faith, we discover that we have miscalculated our statutory liabilities in some way. The object should be getting it right and making sure SoundExchange's members are paid properly. Accordingly, we propose a revised regulation to address this situation.

25 Before the UNITED STATES COPYRIGHT ROYALTY JUDGES THK LIBRARY OF CONGRESS Washington, D.C.

) In re ) ) DETERMINATION OF ROYALTY ) Docket No. 14-CRB-0001-WR (2016-2020) RATES AND TERMS FOR ) EPHEMERAL RECORDING AND ) DIGITAL PERFORMANCE OF ) SOUND RECORDINGS (SF'V) ) )

DECLARATION OF MICHAEL HERRING

I, Michael Herring, declare under penalty ofperjury that the statements contained in my

Written Direct Testimony in the above-captioned proceeding are true and correct to the best of my knowledge, information, and belief. Executed this 6 day ofOctober 2014 in Oakland,

California.

Michael l. err' PUBLIC VERSION

Before the UNITED STATES COPYRIGHT ROYALTY JUDGES THE LIBRARY OF CONGRESS Washington, D.C.

) In re ) ) DETERMINATION OF ROYALTY ) Docket No. 14-CRB-0001-WR (2016-2020) RATES AND TERMS FOR ) EPHEMERAL RECORDING AND ) DIGITAL PERFORMANCE OF ) SOUND RECORDINGS (WEB IV) ) )

AMENDED WRITTEN REBUTTAL TESTIMONY OF MICHAEL HERRING

(On behalf of Pandora Media, Inc.)

Introduction

1. My name is Mike Herring. I am the Chief Financial Officer of Pandora Media,

Inc. (“Pandora” or “the Company”). I previously provided testimony during the direct phase of

this proceeding.

2. I offer this rebuttal testimony to address several issues raised in SoundExchange’s

written direct statement: (a) SoundExchange’s contention, discussed in the testimony of

SoundExchange’s economist David Blackburn as well as several fact witnesses, that Pandora is

artificially limiting advertising inventory and deferring short-term profitability in favor of user

growth; (b) the premise that the interactive-service benchmarks on which SoundExchange relies

serve as suitable proxies for the statutory license fees to be paid by a non-interactive service like

Pandora, especially when viewed against the economic terms of the direct licenses that Pandora

has been entering into; (c) SoundExchange’s suggestion that recording artists are not being fairly

compensated at the royalty rate Pandora currently pays for sound recording performances; and PUBLIC VERSION

(d) SoundExchange’s proposed revisions to the terms and definitions that govern statutory licensees.

“Sell More Ads” Is a Straw Man

3. In an attempt to divert the Judges’ attention away from SoundExchange’s exorbitant rate proposal, which surely fails the “willing buyer/willing seller” standard,

SoundExchange has offered testimony from several witnesses in an attempt to legitimize that rate proposal on the basis of Pandora’s supposed ability to pay the extraordinary fees sought.

SoundExchange’s witnesses – chief among them Dr. Blackburn – argue that Pandora is purposely not doing enough to monetize its listener base – mainly by failing to increase the number of ads we stream per hour of listening (what we refer to as our “ad load”).

SoundExchange further contends that Pandora is intentionally foregoing short-term revenue and profitability so as not to sacrifice growth of its user base.1 The transparent thrust of this testimony is: “Pandora, just sell more ads and you will be able to afford to pay what we are asking.”

4. As I understand the standard applied in this proceeding, the Judges are to determine the rate to which a willing buyer and willing seller would agree in a workably

1 These themes pervade SoundExchange’s direct case. For example, SoundExchange’s principal economic expert, Daniel Rubinfeld, testifies that “Pandora has recently indicated that it has reached an ‘inflection point’ in its profitability, having made a policy choice in the past to sacrifice profits in return for growth.” Written Direct Testimony of Daniel Rubinfeld (“Rubinfeld WDT”) at ¶ 83 n.65; id. ¶ 117. See also SX Introductory Memorandum at 14 (“webcasters often forego short-run profitability in favor of user and market share growth”); Written Direct Testimony of David Blackburn (“Blackburn WDT”) at ¶ 54 (Pandora has an “expressed strategy of foregoing short-run profits in favor of longer-run growth . . . .allowing Pandora to keep revenues even lower . . . .”); id. ¶¶ 55-58, 63-81 (Discussion of Webcasters’ “Incentive to Increase Marketshare” and Pandora’s strategy to do so); Written Direct Testimony of Aaron Harrison (“Harrison WDT”) ¶ 13 (“webcasting services have been slow to increase advertising inventory because they are currently focusing on growth in number of users and listener hours, rather than monetization and profit”); Written Direct Testimony of Dennis Kooker (“Kooker WDT”) at 14; Written Direct Testimony of Simon Wheeler (“Wheeler WDT”) ¶ 38 (“I remain skeptical of the immediate ability or desire of many online ad-supported webcasters to generate revenue. . . . It seems to me that many are still trying to attract a critical mass of users by providing to them as much music as they can.”).

2 PUBLIC VERSION

competitive market. The Judges are not, as I understand it, tasked with determining the rate any particular party theoretically could pay and remain in business. The rate that Pandora is theoretically capable of paying is simply not informative to the Judges of the rates at which

Pandora would be a “willing buyer” of statutory sound recording performance rights.

5. SoundExchange’s supposed evidence along these lines is flawed for numerous additional reasons. To start, it ignores the fact that the royalty formula under which Pandora pays SoundExchange includes a per-play as well as a percent-of-revenue prong. Accordingly, even if Pandora did not run a single ad and earned nothing, SoundExchange’s members would still be paid for every play; indeed, under such payment metric, Pandora incurred over

in statutory royalties in 2014 alone, making us the single largest statutory licensee and accounting for approximately half of SoundExchange’s statutory royalty distributions. Both

Pandora and SoundExchange have proposed a continuation of per-play rates as an element of their rate proposals going forward.

6. SoundExchange’s singular focus on Pandora’s current ad loads also fails to capture real-world dynamics, including the fact that the more ads Pandora runs, the less music it performs, meaning reduced per-play royalties. Similarly, to the extent Pandora loses listeners because of increased ad loads, it would mean both fewer payable performances and less ad revenue, meaning fewer dollars flowing into SoundExchange’s coffers. As detailed below, the cumulative impact of a seemingly innocuous decline in listening hours can result in very significant declines in revenue over time.

7. Even more fundamentally, SoundExchange’s breezy critique of Pandora’s business strategy misapprehends the realities of business economics, particularly in the marketplace in which we operate: the market for lean-back radio style listening. It fails to grasp

3 PUBLIC VERSION

the challenges advertising-supported non-interactive music services face in achieving long-term

profitability, as well as the higher long-term value such services promise over subscription-based

models. I address these considerations in some detail in the succeeding sections, and summarize

my key observations below.

8. First, SoundExchange’s myopic focus on short-term profitability ignores the

fundamental realities of growing a business like Pandora to scale. Because the business

decisions that Pandora makes today affect its profitability in the future, Pandora is maximizing

profitability, just not necessarily the profitability of the current quarter. Future profits are always

a concern for Pandora—like any business—and lower future profits resulting from doing things

to increase profits today (e.g., by increasing ad loads too quickly) are “costs” that offset the

“value” of today’s higher profits. Pandora is keenly focused on revenue growth, and has

invested all of its revenue and more2 back into its business, investing in engineering, sales, sales

support, and other critical components necessary to grow an ad-supported business at scale. As

shown in Figure 1 below, Pandora has actually increased the monetization of its listeners at the

same time it has grown its total listening hours. It seems elementary that such investment in

long-term profitability stands to serve not only Pandora and its investors, but also the entire

record industry, well.

2 Pandora had a cumulative net loss of approximately as of the end of 2014. Combined with its total revenue of nearly , Pandora’s total investment is almost over the life of the Company.

4 PUBLIC VERSION

Figure 1 [RESTRICTED]

9. Second, by stating or implying that but for services such as Pandora, everyone who listens to music would subscribe to a paid service that generates a higher average revenue per user (“ARPU”) grossly mischaracterizes the market for music consumption.3 There is a clear and meaningful difference between lean-back services such as Pandora and lean-in services such as Spotify. As Figure 2 below shows, the vast majority of music consumers “lean back” when listening to music.4

3 See Harrison WDT ¶ 16 (suggesting “that Pandora is streaming music to users who might otherwise pay for a subscription or use a higher ARPU streaming service”); see also Kooker WDT at 15, 17; Wilcox WDT at 4-5; Wheeler WDT ¶¶ 30, 35, 38. 4 Edison “Share of Ear” study, Fall 2014. (“Lean-In” defined as owned music and Spotify, but does not include audio/video lean-in services such as YouTube and Vevo. “Lean-Back” defined as AM/FM radio, Sirius, TV Music Channels, Pandora, iHeart Radio and iTunes Radio.). See also Pandora Exhibit 14 (Interview of Tom Conrad, Pandora’s former Chief Technology Officer, at GigaOM RoadMap 2011) at 1:34-1:43 (“About 80 percent of music consumption, by hours, happens in . . . radio.”) (available at https://gigaom.com/2011/11/11/pandora-roadmap-2011/).

5 PUBLIC VERSION

Figure 2 [RESTRICTED]

10. As Spotify’s CEO stated, “I don’t really view [Pandora] as a competitor. . . . We don't want to be the radio service. . . We want to be the place where you store and collect, where you build your playlist for your dinner party or your workout. That is very different from

Pandora.”5

11. Consumer research shows that few people who listen to music spend much money on music; the vast majority of the music-consumers view hearing ads as the appropriate “price” to pay to listen to music.6 According to a leading music consumer research firm, 77% of music

5 Georg Szalai, Spotify’s Daniel Ek Zings Dr. Dre's Beats Music, The Hollywodd Reporter (January 22, 2014), available at http://www.hollywoodreporter.com/news/spotifys-daniel-ek-zings-dr-672509; see also Pandora Exhibit 14 at 10:19-13:40. 6 See, e.g., MIDiA Research, U.S. Music Consumer Deep Dive at 7 (January 2015) (“subscriptions remain fundamentally niche in reach and scale with just 7% penetration. With more than a decade of product offerings the US subscription market is hardly virgin territory. Much of the addressable base for $9.99 subscriptions has been tapped. The next wave of subscribers will require cheaper tiers. In the meantime Apple’s re-launch of Beats Music – whatever name it finally hits upon – will undoubtedly have the most opportunity to expand the market further. Much initial growth however will likely come at the direct expense of the subscription incumbents.”); Morgan Stanley, 1st Annual MS Music and Radio Survey at 2 (January 14, 2015) (finding that 70% of survey respondents indicated little or no willingness to pay for a subscription service). See also Pandora Exhibit 14 at 4:46-5:18 (“Fully half of this country doesn’t spend any money on music in a year. Another 40 percent spends about $14 – 15, something like a single CD a year, and then there is just the 10 percent of people – like you and I – that buy a lot of music. If you want

6 PUBLIC VERSION

listeners spend almost nothing on music.7 As part of its rebuttal case, Pandora is submitting a new consumer survey the results of which confirm:

x the very small percentage of listeners to lean-back services like Pandora who would consider paying any significant money to subscribe to a lean-in service;

x that if free online music services were no longer available, most consumers would revert to broadcast radio, watch music videos or listen to music on YouTube or Vevo, or simply listen to less music, as opposed to subscribing to a lean-in service; and

x that time spent listening to online, non-interactive services is mostly replacing time spent listening to broadcast radio or is new time that would not have been spent listening to any music at all—almost none of it is taken from listening to lean-in services like Spotify.

While lean-back services such as Pandora may (today) generate a lower ARPU, the size of the lean-back market is so much larger than the lean-in market that, in my opinion, the long-term revenue opportunity for record labels and recording artists is in the lean-back market.

12. The rates Pandora has proposed in this proceeding represent, in my opinion, the best rates for maximizing the long-term revenue of record labels and artists by enabling services such as Pandora to invest in the scale, systems and staff necessary to monetize the Internet radio listener audience. Conversely, the rates SoundExchange proposes would decrease the royalties paid by Pandora by diverting revenues needed by Pandora to grow our listenership and advertising revenue (and thereby long-term royalty payments to the record industry) to short- term royalty payments. As I discussed in my direct testimony, if SoundExchange’s proposed rates were adopted, Pandora may have to restrict listener hours as we did when we capped mobile listening in 2013, which had the effect of reducing our royalty payments. See Written

Direct Testimony at ¶ 7.

to play in that 90 percent opportunity, which is what gets us excited, you have to devise products that have a compelling ad-supported experience.”). 7 See MIDiA Research, Music Consumer Segmentation (December 2014) at 6-7.

7 PUBLIC VERSION

I. GROWING AN ADVERTISING SUPPORTED BUSINESS

13. The elements required to create and maintain an economically successful advertising supported business are straightforward. At a high level, these elements fall into three categories:

x Scale: having a verifiably measurable audience of sufficient size to be of value to advertisers;

x Systems: providing advertisers with efficient, trustworthy mechanisms to plan and transact advertising campaigns; and

x Staff: having the sales and support resources necessary to convince advertisers to entrust their business with you and to reliability execute campaign after campaign.

Putting all of these elements together, however, is extremely difficult, especially in the context of a new type of advertising (such as internet delivered audio advertising) and a new form of ad delivery (such as mobile phones).

A. Scale: Growing a Measurable Audience of Sufficient Size

14. The first element required for success is attracting an audience of sufficient size.

Advertisers generally want to limit their spending to a relatively small number of media properties, each of which contributes meaningfully to the advertiser’s goals. Subscale media properties simply are not worth the time and effort of advertisers and their agencies.

1. Growing Audience

15. A nationwide media property that seeks ultimately to attract both national and local advertising spending will typically seek to grow audience in the following sequence:

a. National Audience

16. First is the need to grow a national audience of sufficient scale to be of value to national advertisers. This audience is typically referred to as the National Network Audience.

Large, national and international companies such as auto manufacturers (e.g., Ford), big box

8 PUBLIC VERSION

retailers (e.g., Walmart), and consumer packaged goods companies (e.g., Procter and Gamble) that want to reach a wide, national audience with little or no local differentiation are the primary

National Network advertisers. National Network advertising is typically sold at relatively low rates because the advertiser is more interested in reach (i.e., brand awareness) than targeting (i.e., activating consumers).

17. In the circumstance in which the platform distributes content to attract audience and pays for that content on a percentage-of-revenue basis, growing a national audience is relatively easy. That is, the additional incremental audience does not necessarily increase the cost of content distributed. In the circumstance in which distributed content is paid for on a per- use basis, however, growing a national audience can be extremely expensive. For example,

Pandora paid more than in sound recording performance royalties before it grew its audience to the point where it ran its first national audio ad campaign, which was for FOX TV’s

American Idol in January 2009.8

b. Spot Audience

18. Second is the need to grow a local audience in each of the top 10 - 20

metropolitan areas of sufficient scale to be of value to advertisers who focus on these top

markets. In National Spot advertising, large advertisers seek to target specific (usually large)

markets. For example, a home improvement retailer such as the Home Depot might want to

advertise snow blowers in Boston, Chicago, Denver, Detroit, Philadelphia, Pittsburgh, New

York, Minneapolis-St. Paul, and St. Louis. A variant of this type of advertising is commonly

known as Local Spot advertising. Extending the Home Depot example to Local Spot

advertising, Home Depot may be interested in advertising snow blowers in Boston, shovels in

8 Initially, Pandora was limited to selling display ads, which it could not monetize at levels it has reached with audio advertisements.

9 PUBLIC VERSION

New York and deicers in Chicago. In order to attract National and/or Local Spot advertising, a

platform such as Pandora needs an audience equivalent to the number 4 or number 5 terrestrial

radio station in that market (approximately 6-8% of the relevant demographic). National and

Local Spot advertising is sold at a premium to National Network ads because of the increased

audience targeting.

c. Local Audience

19. Third is the need to grow an audience of sufficient scale in more and more local

markets to be of value to advertisers who limit their focus to just one of these individual local

markets. Here, again, a platform needs at least 6-8% of the targeted demographic in a local

market to attract Local ad spend. Local ads are sold at the highest rate because they are the most

targeted. National, regional and local advertisers are all competing to reach a targeted

demographic in a targeted local market.

20. While Local advertising commands the highest rate, building a local audience

large enough to attract those Local ad dollars is the most challenging audience to grow. Other

participants in this proceeding have expressed similar challenges in monetizing a local audience.9

21. In addition to being the most difficult audience to grow, Local ad sales are also

the most “expensive” to grow. National Network and National and Local Spot advertising is

often bought directly by the advertiser or by an advertising agency. Advertising agencies may

control the advertising budgets of multiple large advertisers. Therefore, a relatively small sales

force located in a few markets can be built that targets these large national advertisers and their

agencies.

9 See Written Direct Testimony of Ben Downs, Vice President and General Manager of Bryan Broadcasting, Inc., at 10 (describing how difficult it is to sell ads in local markets when audience is too small; “In order to even begin to interest advertisers in our streaming audience, we need to increase our listener base significantly.”); id. at 11 (explaining that a small local online audience leads to a lack of demand [that] limits the number of ads our stream provider can sell on our streams”).

10 PUBLIC VERSION

22. In contrast, the only way to sell Local ads is by having “feet on the street.” A local sales force is very expensive to build; each local market requires dedicated sales, sales management and sales operations resources. Growing a local sales force means constantly reinvesting the profits of the business into future revenue growth, which is exactly what Pandora has done over the last three years, going from a few local sellers in a handful of markets to local sellers in markets. The financial resources Pandora has historically had available to reinvest in growing its local salesforce are severely constrained by the very high royalty obligations it pays to SoundExchange. But for these royalty payments, Pandora could have invested even more in developing its local salesforce and could be generating even more revenue than it currently is. By our estimates, if Pandora’s royalty obligations had been only 10% lower, thus allowing us to increase sale-force investment and avoid caps on listening hours, Pandora could have grown total revenue by an additional and also have generated an additional

in royalties to the record industry.

2. Measuring the Audience

23. Advertisers require that the audience they are “buying” is verifiably measurable.

For example, when making advertising buying decisions on terrestrial radio platforms, advertisers usually compare stations’ Nielsen Audio (formerly Arbitron) ratings. Nielsen Audio radio ratings measures the size of a station’s audience by collecting data from a random sample of a population throughout the United States, primarily in 294 metropolitan areas, using either its proprietary Portable People Meters (“PPMs”) or a written diary.

24. When Pandora launched in late 2004, there was no equivalent rating service to measure Internet radio audience. This lack of verifiable audience measurement created problems

11 PUBLIC VERSION

not only for Pandora, but as well for terrestrial broadcasters that also simulcast their signal

online.10

25. The lack of verifiable audience measurement was not a critical issue when

Pandora launched because it did not have an audience of sufficient scale to attract even National

Network advertising. As Pandora grew, it began measuring its audience using internal server log

files. This internal measurement was not sufficiently verifiable to satisfy most advertisers, so in

2010 Pandora engaged Edison Research, a market research firm with a particular emphasis on

media research. Edison took Pandora’s log files and compiled audience calculations for our top

10 markets. Because Edison was using Pandora’s internal log files, this measurement was also

not sufficiently verifiable to satisfy most advertisers. In 2011 Pandora approached both Nielsen

(then Arbitron) and Triton Digital regarding measuring Pandora’s audience in a way that

satisfied the standards of the Media Rating Council (“MRC”).11 Triton uses a beacon-based

technology to measure Pandora’s audience. This technology is similar to Nielsen Audio’s PPMs,

but Triton’s measurement is based on census calculations, while Nielsen Audio’s measurement is

based on survey panels. In May, 2012, Triton announced an audience rating system for Pandora

(and other Internet radio platforms) that extended into local markets. This provided, for the first

time, verifiable audience measurement that satisfied advertisers.12

10 See, e.g., Written Direct Testimony of John Dimick, Senior Vice President of Programming & Operations at Lincoln Financial Media Company, at 7 (“[A] major problem with converting that understanding into advertising dollars has been the lack of a demonstrated audience or a consistent ratings boost based on the streaming listenership. While streaming audience measurement remains in its infancy, advertisers have a high comfort level with over-the-air ratings.”) (emphasis added); id. at 9 (“Advertisers base their buys and the rates they are willing to pay on consistent, demonstrated ratings.”) 11 The Media Ratings Council is an industry-funded organization created at the behest of Congress to review and accredit audience rating services. In addition to setting minimum standards for media rating research, the MRC monitors the activity of rating services through annual external audits of rating service operations performed by a specialized team of independent CPA auditors. 12 Triton obtained MRC accreditation in March 2014.

12 PUBLIC VERSION

26. Obtaining Triton’s MRC-accredited ratings was the critical gating item to

Pandora’s ad inventory being integrated into media-buying platforms of Mediaocean,13 Strata,14 and Telmar,15 the largest media buying and planning platforms for National Spot and Local Spot radio advertising. As discussed below, integration into these types of media buying systems is part of the infrastructure investment that an ad-supported business must build to be successful at scale.

B. Systems: Providing Advertisers with Efficient Mechanisms to Plan and Transact

27. All major ad spending, both national and local, hinges on a set of systems that (1) enable advertisers to efficiently and reliably plan, execute and measure advertising campaigns and (2) allow platforms to deliver the right audience across all campaigns, especially at scale.

Certain systems, such as the independent verified measurement of audience and media buying, require working with third-parties, such as Pandora’s integrations with Triton and Mediaocean,

Strata, and Telmar.

28. Beyond audience measurement and media buying, other systems must enable the platform to efficiently transact and execute a high volume of advertising campaigns. At the scale of a platform such as Pandora, such transactional activity must be managed automatically (i.e., via computers) lest the volume of transactions consume countless hours of manual administration.

13 Mediaocean is an advertising services company that provides advertising purchasing and other services to clients, handling over $130 billion of advertising spending a year. 14 Strata Marketing Inc. is a software company that provides a suite of tools for planning buying various forms of media, including TV, cable, radio, print, digital and outdoor, and handles more than $50 billion worth of media buys annually. 15 Telmar is a leading supplier of advertising and media information software that allows advertisers and ad agencies to plan and optimize media purchases across digital, print, broadcast and outdoor platforms.

13 PUBLIC VERSION

29. For Pandora today, these systems encompass a wide variety of business functions necessary to produce, schedule and deliver advertising on behalf of national and local advertisers. Getting to this point, however, involved a considerable amount of investment in time and engineering resources. For example, when Pandora began selling ads, it had to invest in ad operations, specifically, ad media servers that could deliver audio and display ads to listeners. As its ad volume increased, Pandora had to invest in order management, specifically the order management software that manages the workflow from the execution of the sale

(commonly referred to as an insertion order) to the trafficking of the ad to listeners. Initially,

Pandora relied on third-party software. Due to our unique business (i.e., the sale of both display digital and audio advertisements), these third-party systems often required more than ten people across various departments to process a single order. Pandora invested in the development of its own proprietary order management software known as , which has reduced the average order processing time significantly.

This greatly increased our order management efficiency, enabling our sales efforts and reducing our operational costs. Pandora continues to invest significant time and resources in adding functionality to and other advertising technology solutions to drive towards improving the advertising buying experience while also optimizing pricing yield and increasing operational efficiency. For example,

14 PUBLIC VERSION

30. As its ad volume increased further, Pandora invested in growing its ad trafficking organization. Pandora now employs people who work on building advertising “creatives”; e.g., the display, audio, and custom advertiser video ads that must be capable of being rendered on a wide variety of desktops, tablets and mobile phones. Pandora’s ad volume, and the creative requirements that flow therefrom, is many magnitudes higher than terrestrial radio. In the first instance, Pandora serves a larger volume of a wider variety of ads (display, audio and video) than any terrestrial radio broadcaster also engaged in webcasting. Because of the wide variety of devices and the multitude of screen sizes commonly used by consumers, Pandora’s creatives must be compatible with a very wide range of formats. In the second instance, Pandora delivers individual ads to individual listeners, in comparison to a terrestrial radio broadcaster, which can deliver a single audio ad to all of its listeners at the same time (that is the definition of

“broadcast”). Said differently, to deliver 1,000,000 “impressions,” Pandora has to deliver

1,000,000 ads, while a large terrestrial broadcaster such as Z-100 in New York might only have to deliver 15 or 16 ads and Pandora has to be able to track each “impression” so that it can report back to the advertiser on the success of the campaign.

31. Because personalized Internet radio services such as Pandora deliver individual streams to individual listeners, in addition to higher volume of ads served, Pandora has had to invest in far more complex systems of ad management. That is, just as Pandora delivers personalized music, it also delivers “personalized” ads. Therefore, Pandora has had to build an entire infrastructure that could identify the “right” ad to play to the “right” listener at the “right” time. Because we collect a listener’s age, gender, and zip code at registration, we know these basic demographic facts about our audience. If an advertiser requested, we can, for example, target women aged 25-35 in New York with a particular advertisement. This ability to target

15 PUBLIC VERSION

specific audiences enables us to attract advertising, as well as obtain higher rates for those ads,

because advertisers know that their ads are being delivered to those listeners that are most likely

to be consumers of their products.

32. To obtain even higher ad rates, Pandora engages in far more sophisticated

audience targeting.

Pandora employs engineers to support this kind of advertising technology.

33.

C. Staff: Selling and Support Resources to Attract and Retain Advertisers

34. Scale and systems are necessary but not sufficient; ultimately, advertisers and

their agencies need convincing that their advertising spend on a service like Pandora will be

successful. Making this “sell” – accessing the right people and convincing them to try something

“new” – requires experienced, talented and trained sales representatives and support staff

16 PUBLIC VERSION

physically present in every place where major decision makers do business. Building a team capable of selling hundreds of millions of dollars of advertising all across the U.S. is a major challenge, requiring years of effort and investment.

35. Moreover, this staffing effort must be carefully coordinated lest it fall out of sync with the scale of the media property’s audience or the sophistication of its systems. That is, hiring must keep pace with growing inventory to ensure it is sold, but our sales force cannot grow so fast such that it out-paces inventory and our sellers sit idle. To accomplish all of these objectives, Pandora has hired nearly people in sales and sales management and another in sales support roles. Of those new hires, have been specifically added to our Local sales, sales management and sales support teams.

II. PANDORA’S EXPERIENCE

A. Pandora’s Audience Growth

36. In December 2005, after its first full year of operation, registered accounts logged onto and listened to music on Pandora. In December 2014, registered accounts logged into and listened to music on Pandora.16 As Figure 3 demonstrates, based on size of audience, Pandora was the largest radio “station” in all of the top 25 DMAs.

16 The total number of listeners is likely higher, since evidence shows that of our listeners share accounts.

17 PUBLIC VERSION

Figure 3 [RESTRICTED]

B. Pandora’s Ad-Load Growth

37. Pandora is very serious about revenue growth. In fact, one of our “Principles” is

“Revenue is our oxygen.” Pandora recognizes that generating revenue is the way we will grow our company, by reinvesting that revenue into software engineers to build a better service, sales reps to sell our ad inventory, and marketing efforts to increase our listenership. Pandora’s entire revenue organization, including the Chief Revenue Officer, has, since 2006, been heavily incentivized to generate as much revenue as possible.

38. The key to generating revenue growth is growth in the number of our listening hours. The greater the number of listening hours, the larger the amount of ad inventory we have

18 PUBLIC VERSION

to sell. However, because Pandora pays a per-track, per-listener royalty, growing listener hours

means increased content costs. As described in my original written direct statement, on at least

two occasions Pandora has capped listening—thereby driving down listener hours—when our

growth in listener hours was outstripping our ability to monetize those additional hours.

1. Supply Does Not Drive Demand

39. For Pandora, like most ad-supported services, there is a complex interrelationship

of factors that must be weighed, balanced and optimized to maximize revenue. Pandora employs

people in revenue operations (e.g., pricing and yield management) and sales operations (e.g.,

sales analytics) to undertake this task. In ascertaining how Pandora can sell the highest number

of advertisements for the highest price, these employees estimate such effects as (i) the impact

increased ad inventory would have on the pricing Pandora could realize for its ads, and (ii) the

impact increased ad load would have on overall listener hours. The notion, pressed by

SoundExchange, that simply increasing ad inventory would generate increased ad revenue,

without taking account of the need carefully to manage both the prices Pandora can obtain for

selling advertising and levels of listenership, fails to account for these marketplace realities.

a. Pandora, like all advertising-based services, seeks to optimize its “sell thru” rate

40. Because supply does not drive demand, an ad-supported service such as Pandora

seeks to optimize its sell-thru rate. Where demand closely matches supply, Pandora is able to

charge its maximum rate. The flip side, of course, is that demand fluctuates and keeping supply

too close to demand risks having unmet demand. Pandora, therefore, tries to keep its sell thru

rate for audio ads in the 25 – 54 demographic around of inventory.17 As Pandora achieves

17 Pandora is able to “sell” this demographic to advertisers at higher RPMs than it can “sell” the 13 – 24 or 55+ demographics.

19 PUBLIC VERSION

this level, it increases its ad load. In just the last two years, Pandora has increased its ad inventory by more than , while increasing ads sold by . Because Pandora only increases ad load when it receives the strong market signal that demand does or is about to outstrip supply,

Pandora has been able to increase ad prices and revenue per listener hour, particularly for audio ads, against our primary demographic.

b. Pandora Optimizes for the “Lifetime Value” of Each Listener

41. Another fallacy of the “Simply Sell More Ads” critique is its failure to take account of the need to optimize ad load in a manner that maximizes the revenue generated over the “life” of a listener.18 We know that serving no ads generates no revenue, and that serving nothing but ads generates no listening (and, therefore, no revenue). The key is to find the right balance point between those extremes that will maximize long-term revenue.

42. The triple goal of ad-supported services is to (1) increase usage, (2) increase ad load, and (3) increase revenue per user. Pandora has been able to achieve all three. To do so,

Pandora rigorously studies the impact of additional advertising on listener engagement, including studying listener behavior to determine the impact of various ad loads on listener engagement.

In the most recent such study, we were able to determine the impact on listener hours of various levels of ad load. As Figure 4 shows, higher ad load results in reduced listener hours.

18 See PAN_CRB_00065198 (LTV Model 08.26.2014).

20 PUBLIC VERSION

Figure 4 [RESTRICTED]

43. A reduction in listener hours reduces our ad inventory, which reduces our ability

to generate ad revenue.

By empirically studying the effects of ad load, Pandora can maximize long-

term revenue generation by optimizing ad load to yield maximum listener hours and utilize

targeting and segmentation to yield maximum revenue per advertisement.

2. Pandora Invests in Demand Creation

44. Even though supply does not drive demand, Pandora is not without means to stimulate demand. In fact, everything described above are investments Pandora has made in creating demand for advertisements. The sales people, especially the Local sellers, are creating demand by knocking on doors and explaining the value to advertisers of advertising on

our platform. The people in Pandora’s Sales Marketing group research the market and build

collateral to help our sellers most effectively tell that story. The audience measurement by

Triton Digital and subsequent integrations with Mediaocean, Strata and Telmar create demand

by making advertising on our platform more analogous to purchasing terrestrial radio ads. By

21 PUBLIC VERSION

enabling Pandora to more efficiently deliver and report on the effectiveness of ads run on our platform, Pandora’s investment in automation such as Slingshot drive demand.

3. More Ads vs. “Better” Ads?

45. A variant of the “simply sell more ads” gloss over reality is, “Pandora, FM stations broadcast more ads per hour than you; why don’t you just be more like them?”

Pandora’s answer is that the approach taken by terrestrial radio stations would not maximize revenue in the context of a service like Pandora. First, and perhaps most important, there is substantial data indicating that, while FM stations broadcast a higher number of ads, many of those ads are not listened to by consumers. It is critical to understand that in the FM world, consumers can and frequently do change stations when a pod of ads begins on a given station.

According to research by Added Value, a global marketing consultancy with over 30 years of experience, 79% of terrestrial radio listeners usually change the station when an advertisement comes on. Thus, there is a very significant difference between the number of ads an individual radio station broadcasts in an hour and the number of ads consumers actually listen to in an hour.

46. By contrast, on Pandora consumers cannot skip an ad. Even if they change from one Pandora station to another, the second station will not start playing until the ad is done playing. Thus, talking about the number of ads an individual FM radio station broadcasts in an hour and the number of ads played on Pandora in an hour is an apples-to-oranges comparison. In one case, the consumer actual listens to the ads (Pandora); in the other case they quickly switch to another station (FM radio).

47. The relevant comparison is not how many ads Pandora runs each hour versus FM radio, but how much revenue is generated by each hour of listening. By this metric, Pandora actually monetizes better than FM radio does in three of the largest radio markets, where Pandora has for the longest time had a dedicated local sales force. For example, in San Francisco,

22 PUBLIC VERSION

Pandora’s revenue per thousand hours of listening on desktop is , which is higher than average terrestrial radio revenue per thousand hours of listening. Pandora achieves this higher

RPM with a lower ad-load for the listener, which drives longer listening, which creates additional ad inventory; i.e., a true virtuous cycle. Over time, we expect that favorable metric to extend to additional markets.

Pandora’s Direct Licenses Reflect a Superior Benchmark to SoundExchange’s On-Demand Benchmark

48. SoundExchange has built its case around record-company agreements with on- demand services such as Spotify and Rhapsody. Pandora, however, has executed direct license agreements that we believe provide a much better benchmark. First, these agreements cover non-interactive webcasting of the kind covered by the statutory license, meaning, they reflect what Pandora as a willing buyer has demonstrated it is willing to pay to willing sellers for the very rights at issue in this proceeding. Second, the Pandora agreements reflect the workings of competition. Our agreement with Merlin, for example, covering the repertories of numerous independent record labels, provides for a lower effective per-play rate as we “steer” our plays toward directly-licensed repertory, and thus provides us with a competitive incentive to play directly-licensed tracks more heavily than we would otherwise.

49. As I noted in my direct testimony, the Merlin agreement covers significant and critically-acclaimed sound recordings, including winners of Grammys and other major record- industry awards. At this year’s 57th Annual Grammy Awards, artists associated with Merlin labels won the awards in categories for Best Jazz Vocal Album (Concord); Best Jazz

Instrumental Album (Concord); Best Bluegrass Album (Concord); Best Folk Album (ATO);

Best Gospel Album (eOne); and Dance/Electronic Album of the Year (Warp).

23 PUBLIC VERSION

50. Since signing the Merlin agreement, we have succeeded in increasing directly-

licensed spins by approximately . This means that the effective rate we are paying Merlin on behalf of participating labels is per ad-supported play. Based on those Merlin labels that have opted in to the agreement, the license covers about of our total performances.

51. In January, we signed a direct license agreement with Naxos, a leading classical

music label. See Pandora Exhibit 15.

Pandora’s Payments to Artists

52. In October 2012, Pandora posted on its company blog the payments that Pandora

would make to certain artists in the coming twelve months (i.e., October 2012 through

September 2013).19 In that post, Pandora stated:

Have you heard of Donnie McClurkin, French Montana or Grupo Bryndis? If you haven’t you’re not alone. They are artists whose sales ranks on Amazon are 4,752, 17,000 and 183,187, respectively. These are all working artists who live well outside the mainstream – no steady rotation on broadcast radio, no high profile opening slots on major tours, no front page placement in online retail. What they also have in common is a steady income from Pandora. In the next twelve months Pandora is on track to pay performance fees of $100,228, $138,567 and $114,192, respectively, for the music we play to their large and fast-growing audiences on Pandora.

And that’s just the tip of the iceberg. For over two thousand artists Pandora will pay over $10,000 dollars each over the next 12 months (including one of my favorites, the late jazz pianist Oscar Peterson), and for more than 800 we’ll pay over $50,000, more than the income of the average American household. For top

19 Pandora and Artist Payments, The Pandora Blog (October 9, 2012), http://blog.pandora.com/2012/10/09/pandora-and-artist-payments/.

24 PUBLIC VERSION

earners like Coldplay, Adele, Wiz Khalifa, Jason Aldean and others Pandora is already paying over $1 million each. Drake and Lil Wayne are fast approaching a $3 million annual rate each.

53. In response to testimony from SoundExchange witnesses Fletcher Foster (Iconic),

Raymond Hair (AFM), and Michael Huppe (SoundExchange), individually and collectively seeking to convey the impression that without “strong” (i.e., significantly higher) statutory rates than the rates under which Pandora is now operating, recording artists will be deprived of “fair compensation,” I want to update the above referenced numbers. In 2014, Pandora paid or more to more than recording artists and their respective labels. In fact, Pandora paid the equivalent of the median U.S. household income of $52,000 to more than artists and their labels. Pandora paid or more to nearly artists and their labels. And what about

Donnie McClurkin, French Montana, and Grupo Bryndis? Pandora paid them (and their labels) about , and , respectively, increases of , and since the October 2012 blog post. In short, sweeping and unsupported generalizations as to allegedly systematic under-compensation of artists under Pandora’s prevailing rate structure are wholly unfounded. Pandora’s more than in annual royalties overall and its indisputably generous payments to the most productive artists and labels utterly belies such a notion.

SoundExchange’s Proposed Amendments To The Terms Should Be Rejected

54. In this portion of my rebuttal testimony, I respond to SoundExchange’s proposed

Rates and Terms for the 2016-2020 rate period, including, but not limited to, proposed modifications to Sections 380.2, 380.3, 380.4, and 380.6 of the Copyright Royalty Board’s regulations, 37 C.F.R. §§ 380.2, 380.3, 380.4 & 380.6.

A. SoundExchange’s Proposed Definition of Revenue Should Be Rejected

55. As part of SoundExchange’s rate proposal, which includes a percent-of-revenue prong, SoundExchange has proposed a new definition of “Gross Revenue.” The governing

25 PUBLIC VERSION

webcasting regulations, which to date have been based on a per-play royalty rate, have not

previously contained such a definition of revenue. The new Gross Revenue definition that

SoundExchange proposes here, however, is very different from the revenue definitions adopted

by the Judges in the past for other types of services that do pay based on a percentage of their

revenue, including satellite radio (37 C.F.R. § 382.11), preexisting subscription services (Music

Choice) (37 C.F.R. § 382.2) and cable/satellite music services (37 C.F.R. § 383.2).

56. Notably, SoundExchange’s proposed revenue definition here includes certain

unrelated sources of revenue that the Judges have typically and rightly excluded – often over the

objection of SoundExchange – in the regulations applicable to the other licensee categories:

a) Product and Sales: SoundExchange’s revenue definition would require

Pandora and other licensees to include revenue from “sales of products and services offered as

part of or through the Service, including revenue from products and services that are Bundled

with the Service.”20 SoundExchange’s revenue definition also includes “Revenue from any

software or other product associated with the Service,” including “placement fees for such

software or other product.” Assuming Pandora is the licensed “Service” under this definition

(we “make eligible transmissions”), this broad definition could be read to sweep in all manner of

equipment sales, media player software and other “app” sales, and, most notably, advertising or

subscription revenue paid for other non-statutory services offered by Pandora. For that matter, it

could include Pandora t-shirts and coffee mugs, if those items were sold “through” or “associated

with” the Service.21

20 “Service” is not limited to transmissions covered by the statutory license, but rather is more broadly defined as a service that “makes eligible transmissions,” even if such transmissions comprise only part of the service offering. “Bundled” is defined as a situation where the user must take/use/receive both the Service and some other product/service. 21 The only exception allowed by SoundExchange is for products where the Service is Bundled with other products or services “that do not involve the Service.” This exception will have limited effect in shielding

26 PUBLIC VERSION

By contrast, such unrelated revenues are explicitly excluded from the definition of “Gross Revenues” for satellite radio. See, for example, 37 C.F.R. § 382.11 at (3)(i)

(excluding revenue from sale and/or license of equipment and/or other technology, including bandwidth and receiving devices); (3)(ii) (excluding revenue from intellectual property licenses);

(3)(vi)(B) (excluding revenue for “products and/or other services offered for a separate charge where such channels use only incidental performances of sound recordings”); and 3(vi)(D)

(excluding revenue from “products and/or other services for which the performance of sound recordings and/or the making of Ephemeral Recordings is exempt from any license requirement or is separately licensed”).

b) Advertising Agency Commissions: SoundExchange’s proposed revenue definition includes advertising revenue, as is common, but does not allow for the deductions of ad agency commissions paid by the licensee (which deduction is also common). That not only is unfair, it is at odds with the revenue definitions governing other statutory licensees,22 not to mention in Pandora’s agreements with ASCAP and BMI. SoundExchange’s witnesses have provided no explanation or justification for parting with standard industry practice or the prior regulations on this point.

c) Data: SoundExchange’s proposed revenue definition includes “Revenue generated by the use of exploitation of data gathered or generated from the Service.” The regulations governing satellite radio, by comparison, include only advertising and subscription unrelated revenue. First and foremost, a Service might offer other optional products or services – ticket sales, for example – in some arrangement less formal than a “Bundle” (which, under SoundExchange’s definition, requires a situation where the user must take/use/receive both the Service and other product/service). Second, SoundExchange provides no guidance to determine whether a Bundled service is or is not “involved” with the licensed Service. 22 See, e.g., 37 C.F.R. § 382.11 (“gross revenues” definition) at (1)(ii) (excluding “advertising agency and sales commissions”); § 382.2(e)(2) (excluding “advertising agency commissions not to exceed 15% of those fees incurred to a recognized advertising agency not owned or controlled by Licensee”); § 383.2(g)(1)(ii) (same).

27 PUBLIC VERSION

revenue. See 37 C.F.R. § 382.11. Those regulations also exclude revenue for “current and future data services offered for a separate charge.” 37 C.F.R. § 382.11 at (3)(vi)(A).

d) Bad Debt: SoundExchange’s proposed revenue definition does not allow licensees to reduce the revenue base on account of bad debt. Bad debt expense reflects revenue that was initially booked as earned but that was not ultimately collected from the customer.

SoundExchange is therefore proposing, again, that it and its members should get a cut of revenue that is never actually collected. The revenue definitions governing satellite radio and preexisting services exclude “bad debt expense” (§ 382.11 at (3(v)) and “bad debts actually written off during the reporting period” (§ 382.2 at (2)), as does the definition that applies to New

Subscription Services (37 C.F.R. § 383.2 at (g)(viii) (requiring inclusion only of bad debts

“recovered”)).

e) Other Fees: SoundExchange’s proposed revenue definition leaves out another exclusion found in the satellite radio revenue definition: the exclusion for “Sales and use taxes, shipping and handling, credit card, invoice, and fulfillment service fees.” Instead, it includes a deduction solely for sales, excise, and use taxes. The other items, however, no different than taxes, reflect fees that licensed services collect, not for the music service itself, but to cover costs charge by third parties for delivering the service – most of which are simply passed through to such third-parties (such as the local taxing authorities). In the case of credit card fees, while Pandora recognizes revenue for subscriptions paid by credit card, the credit card companies deduct their fees off the top prior to passing the revenue to Pandora. As a result, the revenue actually collected by Pandora is less (by the amount of the fees) than what is initially recognized. Absent such a deduction, Pandora would be paying royalties on revenues it does not actually collect.

28 PUBLIC VERSION

57. SoundExchange and its members have no entitlement to streams of revenue that

are paid for items other than statutory streaming or (as to bad debt and credit card fees) are never

collected in the first place. There is no reason SoundExchange should receive a share of the

Company’s revenues earned for lines of business, service offerings, data sales, or software and

equipment sales that are not covered by the statutory license merely because Pandora is required

to pay a royalty for performances that are covered by its statutory license. This could include the

launch of an interactive streaming service (which would be separately licensed from sound

recording copyright owners) or a business establishment service (where there is an exemption for

the public performance of sound recordings). And there is even less reason for Pandora to be

taxed on revenue that is never collected.

58. For just these reasons, such unrelated revenues are, as noted above, excluded from

the definition of “Gross Revenues” for satellite radio and other categories of licensees. These

exclusions grew from the Judges’ recognition in the Satellite I proceeding that “[i]n order to

properly implement a revenue-based metric, a definition of revenue that properly relates the fee

to the value of the rights being provided is required.”23 The Judges accordingly defined “Gross

Revenue” through a variety of exclusions in order to “more clearly delineate the revenues related

to the value of the sound recording performance rights at issue.”24 Notably, when

SoundExchange attempted to eliminate those exclusions in the Satellite II proceeding, the Judges

rejected that attempt, explaining that they were “driven by the admonition in SDARS–I to

include only those revenues related to the value of the sound recording performance rights at

issue in this proceeding. The Judges are satisfied that the exclusions permitted in the current

23 See Determination of Rates and Terms for Preexisting Subscription Services and Satellite Digital Audio Radio Services, Docket No. 2006-1 CRB DSTRA (“Satellite I”), Fed. Reg. Vol. 73, No. 16 p. 4087 (Jan. 24, 2008). 24 Id.

29 PUBLIC VERSION

Gross Revenues definition remain proper.” 78 Fed. Reg. 23054, at 23072 (citing SDARS–I, 73

FR at 4087) (internal citations omitted).25 SoundExchange’s witnesses do not explain why the

situation should be different here.

59. In sum, in order to ensure that SoundExchange is only paid on the portion of

revenues derived by a licensee from operations under the statutory license, the regulations must

account for the possibility that a service such as Pandora may be involved in multiple lines of

business that fall outside of the scope of this proceeding, and may earn revenues from the

operation of such services that are entirely distinct from the operations for which the Judges are

currently establishing a royalty rate. The regulations should make clear that revenues of the

licensee that are derived from any other service operated by the licensee are not to be included in

the revenue base on which royalties are payable to SoundExchange.

60. Pandora’s definition of revenue does just that: it includes all money earned

according to GAAP26 derived from making eligible transmissions in the United States, and

excludes revenue from activities other than making eligible transmissions or from eligible

transmissions outside the United States.

25 The Judges continued: “In defining Gross Revenues, the Judges plainly stated that it was their intention to unambiguously relate the fee charged for a service provided by an SDARS to the value of the sound recording performance rights covered by the statutory licenses.” Satellite II, 78 Fed. Reg. at 23072 (citing SDARS–I, 73 FR at 4087). 26 SoundExchange’s proposal, in addition to the problems noted above, defines revenue to include “all amounts paid, payable, credited, or creditable to Licensee, received or receivable by or on behalf of Licensee, or recognized by Licensee as revenue under United States Generally Accepted Accounting Principles (U.S. GAAP).” That proposal is redundant at best and confusing at worst: it is not at all clear, for example, how revenue that is “credited” or “creditable” is different than revenue that is “payable.” It also creates the possibility for audit disputes, i.e., that SoundExchange or its auditors might identify revenues never recognized or received by Pandora as nonetheless being somehow “creditable” or “payable.”

30 PUBLIC VERSION

B. Ephemeral Recordings Definition

61. SoundExchange has proposed to retain the language in Section 380.3 of the current regulations, which states:

Ephemeral recordings. The royalty payable under 17 U.S.C. 112(e) for the making of all Ephemeral Recordings used by the Licensee solely to facilitate transmissions for which it pays royalties shall be included within, and constitute 5% of, the total royalties payable under 17 U.S.C. 112(e) and 114.

62. Pandora has two concerns with SoundExchange’s proposal. First, as indicated by the bolded language above, it appears to permit the making of ephemeral recordings only for sound recordings for which a performance royalty is paid. However, the definition of

“Performance” proposed by SoundExchange (which mirrors the current definition and Pandora’s proposed definition) exempts certain performances from payment, for example, “incidental” performances including “transitions in and out of commercials.” See 37 C.F.R. § 380.2. This creates the possibility (likely unintended) that ephemeral copies of sound recordings that are used by a service for non-compensable performances under Section 114 might not be authorized under the regulations. To remedy this issue, Pandora has proposed the following simple change to Section 380.3:

Ephemeral recordings. The royalty payable under 17 U.S.C. 112(e) for the making of all Ephemeral Recordings used by the Licensee solely to facilitate transmissions for which it pays royalties made pursuant to 17 U.S.C. 114 shall be included within, and constitute 5% of, the total royalties payable under 17 U.S.C. 112(e) and 114.

This edit clarifies that a webcaster is licensed to make ephemeral copies of sound recordings the performances of which are either payable or exempted from payment under Section 114.

63. In addition, the regulations adopted by the Judges should affirmatively state that statutory licensees are permitted to make as many ephemeral phonorecords of sound recordings as the licensee needs to optimize their statutory service. Clarification of this point – which I

31 PUBLIC VERSION

don’t understand SoundExchange to be contesting – is important because the statute provides that a statutory licensee is entitled “to make no more than 1 phonorecord of the sound recording

(unless the terms and conditions of the statutory license allow for more).” 17 U.S.C. §

112(e)(1) (emphasis added). The terms of the statutory license should call for “more,” for the obvious reason that statutory internet radio streaming necessarily involves making multiple copies to facilitate transmissions in different streaming formats and at different bit rates, to have backup copies available for disaster recovery purposes, and to handle the volume of a popular national streaming service, among other reasons. To eliminate any doubt, we recommend adding the following sentence to Section 380.3: “A Licensee is authorized to make more than one

Ephemeral Recording of a sound recording as it deems necessary to make noninteractive digital audio transmissions pursuant to 17 U.S.C. 114.”

C. Accelerated Timing of Royalty Payments

1. 30-Day Payment Window

64. SoundExchange has proposed amending Section 380.4(c) of the regulations so that the timing of royalty payments be shortened from 45 to 30 days in order to expedite the royalty distribution process for artists and copyright owners. SoundExchange also calls for corresponding changes in the delivery deadlines for statements of account and reports of use.

These changes should be rejected for several reasons, both procedural and substantive.

65. First, as the Judges know, there is a separate rulemaking proceeding addressing notice and recordkeeping regulations. In that proceeding, SoundExchange has already sought to shorten the period for delivery of reports of use from 45 to 30 days. Pandora and a number of other statutory licensees have opposed that recommendation and filed detailed comments explaining their opposition. There is no reason to rule on that topic here as well when it has been fully litigated elsewhere – or to give SoundExchange a second avenue for pursuing its desired

32 PUBLIC VERSION

result. Moreover, if the Judges reject SoundExchange’s proposal in that separate proceeding to shrink the timing for submitting reports of use (thus maintaining the current 45-day window), it would make little sense to rule here that statements of account and payments should be submitted in a shorter 30-day window. As should be obvious, the statements of account and payment calculations are premised on the performances contained in the reports of use; if the reports are still being completed and vetted when the payment comes due (under the shorter 30-day window), it raises the possibility that the statements of account and payments could suffer from inaccuracies or – more likely – licensees would essentially be forced to complete their reports of use in 30 days to avoid such inaccuracies even though the deadline would be 45 days.

66. As Pandora explained in the separate recordkeeping rulemaking, there are also solid substantive reasons not to reduce the time period for delivering reports of use from 45 to 30 days – reasons that apply equally to monthly payments and statements of account (as

SoundExchange proposes here):

SoundExchange proposes that licensees be required to deliver ROUs in 30 rather than 45 days. . . . The alleged reason for accelerating payment is to enable expedited payments to copyright owners and artists. . . .

There is simply no basis for accelerating the delivery of ROUs under the existing regulations. As SoundExchange appears to acknowledge, it is already able to distribute the overwhelming majority of royalties paid to it, and there appears to be little demand for accelerating payments by 15 days. Moreover, it strains credulity that reporting to SoundExchange could be improved by giving statutory licensees less time to complete ROUs. If SoundExchange is truly receiving ROUs that are incomplete or untimely, then perhaps the extant regulations are already too burdensome or the time period for reporting is too short. There is no evidence that accelerating reporting obligations will improve the quality of reporting, which is SoundExchange’s stated goal.

. . .

The preparation of ROUs can be challenging. As described above, Pandora currently plays over 1.5 million unique sound recordings. This means that for any given month, Pandora generates an ROU with no fewer than 1.5 million rows of data. To compile those over 1.5 million rows of data, Pandora must determine the

33 PUBLIC VERSION

play counts for each one of those 1.5 million songs from billions of server instances (i.e., each time a user received a transmission of a sound recording). It takes significant resources to generate those reports and track performances at the volume at which Pandora operates.

. . .

Although Pandora can and does comply with today’s reporting requirements within 45 days, accelerating ROU delivery while at the same time increasing the information to be included within ROUs while being subject to financial penalties for failing to do so accurately, will impose a significant and unreasonable burden on Pandora (and likely many other statutory licensees).

See Reply Comments of Pandora Media, Inc. in Recordkeeping Rulemaking at 15-17 (Sept. 5,

2014) (attached here as Pandora Exhibit 16).

67. The significant burden imposed upon licensees by shortening the payment and

reporting deadlines from 45 to 30 days will not improve the quality of royalty calculations or

reporting. Rather, accelerated payment and reporting will likely only lead to more hardships on

licensees and greater likelihood of errors in payment calculation and reporting. That is simply

not justified. The regulations should not impose significant additional burdens upon licensees

for a marginal (if any) speedup in SoundExchange royalty allocation and distribution –

especially if it introduces the possibility of errors in such distributions. SoundExchange recently

touted that it distributed $773 million in royalties in 2014, an increase of 31% over 2013

distributions. This was accomplished with payment and reporting deadlines of 45 days. The

system is not broken and no fix is required.27

27 Marketplace agreements for the same rights at issue in this proceeding support retention of a 45-day payment period. See Agreement between Pandora Media Inc. (“Pandora”) and Music and Entertainment Rights Licensing Independent Network B.V. (“Merlin”), § 12.a.

34 PUBLIC VERSION

D. Verification of Royalty Payments.

1. Definition of “Qualified Auditor”

68. SoundExchange attempts to expand the universe of entities that can conduct a

verification of a statutory webcaster by proposing elimination of the current requirement that a

“Qualified Auditor” be a certified public accountant. Instead, SoundExchange seeks to allow

any “person, who by virtue of education or experience, is appropriately qualified to perform an

audit to verify royalty payments related to performances of sound recordings.” The Judges

should reject this proposal for the same reasons they rejected a similar effort in Webcaster II.

69. Importantly, it is unclear what “by virtue of education or experience” means.

What “education” or “experience” – apart from the rigorous accreditation and standards applied to CPAs – is appropriate for conducting an audit of a statutory licensee? Must the education or experience be in copyright law, network design or engineering, finance, accounting, or something else? Is one survey course in accounting sufficient? SoundExchange does not answer these questions.

70. But the Judges in Webcaster II did. There they noted that the purpose of verifications is not to provide them at the lowest price but to establish a high degree of credibility in the results of the verification. That credibility is achieved by requiring verifications solely by those who are governed by accepted standards and practices of auditing and standards of conduct:

By eliminating the requirements that an auditor be a CPA and independent from SoundExchange, SoundExchange is seeking to transform the prior verification process into what it calls ‘‘technical audits.’’ Technical audits would, in SoundExchange’s view, reduce its costs by allowing in-house technical experts to conduct the audits rather than outside CPAs, who might lack the technical capability for the data processing and analysis and may be more expensive than in-house personnel. The Copyright Royalty Judges have reviewed the record company/ music service agreements submitted by the parties and note that some agreements permit technical audits. Others, however, require the auditors to be

35 PUBLIC VERSION

CPAs, and that the auditor be independent of both the licensor and licensee. While technical audits by in-house personnel might be cheaper for the Collective, we conclude that it is more important, in the interest of establishing a high level of credibility in the results of the audit, that the auditor be independent of both parties. Likewise, we find that requiring the auditor to be certified further raises confidence levels in the audit. CPAs have experience in the field of accounting, are familiar with the accepted standards and practices for auditing, and are governed by standards of conduct. If technical skills are required to process the data of a Service, the auditor can request assistance. In sum, the Copyright Royalty Judges are requiring that the auditor be certified and independent of both SoundExchange and the Service being audited.

Final Rule and Order in Digital Performance Right in Sound Recordings and Ephemeral

Recordings, Docket No. 2005-1 CRB DTRA, 72 Fed. Reg. 24084, 24109 (May 1, 2007) (internal citations omitted).

71. This statement remains true today. I note that American Institute of Certified

Public Accounts, Inc. (“AICPA”) recently revised its Code of Professional Conduct (effective

December 15, 2014) (the “Code”), to clarify the rules that apply to all “members,” which includes a “member” in public practice, in business, and even to those who are retired or unemployed. See Code Section 0.100.02. These rules are designed to ensure the objectivity and integrity of members of the accounting profession. See Code Sections 0.300.030.02 and .04.

The Code states that “[i]ntegrity is measured in terms of what is right and just” (Section

0.300.040.04) and prohibits the “subordination of the [auditor’s] judgment” to the interests of its client – here, SoundExchange. Code Section 0.300.050.03.

72. I have significant concerns about permitting an audit by a non-Qualified Auditor, as that term is currently defined in 37 C.F.R. § 380.2, that is not bound by these obligations.

Specifically, I am troubled by the prospect of an auditor not governed by the Code conducting an audit solely to determine whether a licensee has underpaid royalties to SoundExchange rather than determining the accuracy of payments – whether underpayments or overpayments – under the statutory license. Following discussions with a representative of a music service that has

36 PUBLIC VERSION

been audited by one of SoundExchange’s preferred auditors, I understand that the auditor (not a

CPA, and therefore not a Qualified Auditor) has conducted regular audits solely to determine

underpayments to a licensor. This approach could be particularly problematic where licensees

undertake direct licensing efforts alongside reliance upon the statutory license to secure rights to

sound recordings, as Pandora has done. If SoundExchange’s auditor were not bound by

standards of conduct, the auditor could search solely for evidence that a licensee had mistakenly

claimed direct-license credit for a track not actually covered by the direct license (and thus

underpaid SoundExchange), but not attempt to determine whether the licensee had overlooked

performances where it could have claimed a direct-license credit (and thus overpaid

SoundExchange).

73. Restricting audits to Certified Public Accountants will help avoid this result and

ensure that the integrity of the audits is not brought into question by the auditor only looking for

underpayments. To further ensure the independence of an auditor, the auditor should be

prohibited from conducting audits on a contingency fee basis. Qualified Auditors should either

be compensated on a flat fee or time and materials basis, but excluding any time spent consulting

with SoundExchange, which should not be at the expense of the licensee. Finally, the

regulations should be amended so that an auditor must determine whether an underpayment or

overpayment has been made to the collective. Statutory licensees should not be penalized if they

inadvertently overpay royalties to SoundExchange.28 In the event of any such overpayment, the

regulations should provide for a credit or refund to the licensee, with interest. As interest is

28 Pandora’s proposed rates and terms amend § 380.6(c) to add the following: “Subject to Section 380.6(e), a Qualified Auditor must determine the accuracy of royalty payments made to the collective, including whether an underpayment or overpayment of royalties has been made, and the Qualified Auditor may not be compensated on a contingency fee basis.”

37 PUBLIC VERSION

currently chargeable for any underpayment of royalties, interest should also be credited for any

overpayment of royalties.29

2. Acceptable Verification Procedure

74. SoundExchange has also proposed the deletion of Section 380.6(e), which

provides that an audit of a licensee already performed in the ordinary course of business shall

“serve as an acceptable verification procedure.” This provision has been included in all of the

regulations adopted for Webcasting since Webcasting I. It ensures that a licensee that is already

subject to an audit by a Qualified Auditor in the normal course of business will not have to be

burdened with a second audit by SoundExchange. I understand that it was negotiated by

representatives of the recording industry and licensees in 2002, and there is no reason for that

provision to now be excised from the regulations.

75. Where a statutory licensee is audited by a Qualified Auditor and the Qualified

Auditor has concluded that the licensee’s financial statements present fairly, in all material

respects, the consolidated results of the licensee’s operations and cash flows in conformity with

U.S. generally accepted accounting principles (GAAP), that audit should suffice and preclude

any further audits under CRB regulations. Pandora’s auditors would not be able to reach that

conclusion and certify Pandora’s financial statements if Pandora was not accurately calculating

its liabilities to SoundExchange, particularly where Pandora’s payments to SoundExchange are

Pandora’s single largest expense.

29 Pandora’s First Amended Proposed Rates and Terms adds the following Section 380.6(h): “Make-up payments or credits. Upon the conclusion of the verification and the resolution of all claims between the Collective and the Licensee, (i) the Licensee shall, in the case of any underpayment, remit the amount of any agreed-upon underpayment to the Collective, as mutually agreed by the Collective and the Licensee, which agreement may, but need not, include installment payments, with interest, at the rate specified in Section 380.4(e) and (ii) the Collective shall, in the case of any overpayment, credit the account of the Licensee in the amount of any agreed upon overpayment to the Collective, with interest, at the rate specified in Section 380.4(e).”

38 PUBLIC VERSION

76. If the Judges accept SoundExchange’s proposal on the theory that a routine financial audit may not devote sufficient attention to server log records and other more technical aspects of a licensee’s royalty reporting mechanisms, then the Judges should at least retain the right of a licensee to have a Qualified Auditor conduct a verification of server logs in the ordinary course of business to serve as an acceptable form of verification. SoundExchange is currently in the position to wield an audit as a weapon, and licensees should not be subject to such harassment by either SoundExchange or the record companies that control it.

E. Unclaimed Funds

77. SoundExchange has proposed in its Proposed Rates and Terms to push forward the current regulation for the treatment of so-called “unclaimed funds.” Current regulations provide that in the event a sound recording copyright owner or featured artist entitled to a portion of a pool of royalties (presumably the royalties for a calendar month) fails to register with the collective within three years of the date of first distribution of all or any portion of that pool of royalties by SoundExchange, SoundExchange may retain those unclaimed funds

78.

Pandora therefore proposes the following amendment to

380.8:

If the Collective is unable to identify or locate a Copyright Owner or Performer who is entitled to receive a royalty distribution under this subpart, the Collective shall retain the required payment in a segregated trust account for a period of 3 years from the date of distribution. No claim to such distribution shall be valid after the expiration of the 3-year period. After expiration of this period, the Collective shall handle such funds in accordance with may apply the unclaimed funds to offset any costs deductible under 17 U.S.C. 114(g)(3). The foregoing shall apply notwithstanding the common or State law or statutes of any State.

39 PUBLIC VERSION Capital Reporting Company Day 12 In Re: Determination of Royalty Rates (Public) 05-12-2015

3087 Before the

UNITED STATES COPYRIGHT ROYALTY BOARD

Library of Congress

Washington, D.C.

------: In Re: : Docket No. : 15-CRB-0001-WR Determination of Royalty : (2016-2020) Rates and Terms for : Volume 12-PUBLIC Ephemeral Recording and : Pages 3087-3092 Digital Performance of : Pages 3205-3357 Sound Recordings (Web IV) : ------:

PUBLIC SESSION

Washington, D.C.

Tuesday, May 12, 2015

The hearing in the above-entitled matter was convened at 9:12 a.m.

BEFORE COPYRIGHT ROYALTY JUDGES:

SUZANNE M. BARNETT, CHIEF JUDGE

DAVID R. STRICKLER, JUDGE

JESSE FEDER, JUDGE

3331 3333

1 A. You're the liberal arts major. You 1 Q. And why don't you go back one more 2 tell me. 2 position backwards. 3 Q. You'll take my word at 396 hours? 3 A. And then I joined Adobe through the 4 A. Okay. 4 acquisition of Omniture, which was a software 5 Q. Okay. Is there anything that you're 5 service business that did Web analytics and 6 aware of, from your two years of work on this case, 6 marketing optimization for companies. 7 that the average listener to interactive services 7 Q. And are your job -- or your employment 8 listens to 396 hours a month of interactive 8 history spelled out in your written direct 9 streaming radio, on average? 9 testimony at the beginning, is that right? 10 A. I'm not aware of any evidence. I'm not 10 A. It is. 11 sure it's a question I've investigated. 11 Q. What are your job responsibilities at 12 Q. Okay. 12 Pandora? 13 A. But I think the answer is no. 13 A. Well, as the chief financial officer, I 14 MR. KLAUS: No further questions, Your 14 have pretty broad responsibilities. But a lot of 15 Honor. 15 the standard finance and accounting, compliance, 16 CHIEF JUDGE BARNETT: Thank you, 16 SEC reporting responsibilities. 17 Dr. Kendall. 17 I also have investor relations. As a 18 MR. LARSON: May I approach, Your 18 public company, a relatively large part of my job. 19 Honor? 19 The chief information officer reports 20 CHIEF JUDGE BARNETT: You may. 20 to me. So all of our tech operations, IT systems, 21 MR. LARSON: Your Honor, Pandora calls 21 enterprise systems, all of that also rolls up 22 as its next witness Michael Herring. 22 through my office. 23 CHIEF JUDGE BARNETT: Please raise your 23 I also have our facilities and security 24 right hand. 24 operations. 25 MICHAEL HERRING, 25 And then last but not least, I also

3332 3334

1 called for examination by counsel, and after having 1 supervise our business affairs function, which is 2 been first duly sworn to tell the truth, the whole 2 responsible for our content licensing part of our 3 truth, and nothing but the truth, testified as 3 business. 4 follows: 4 Q. And so would that include music 5 DIRECT EXAMINATION BY COUNSEL FOR 5 licensing? PANDORA 6 A. That would include music licensing. 6 BY MR. LARSON: 7 Q. Are you a CPA? 7 Q. Good afternoon, Mr. Herring. 8 A. I am. 8 A. Hello. 9 Q. Did you file written direct testimony 9 Q. Thank you for your patience in waiting 10 in this matter? 10 to come on the stand today. 11 A. I did. 11 A. Thank you for having me. 12 Q. Can you just take a look at that 12 Q. Can you state your name for the record, 13 demonstrative that I've put in front of you. 13 please. 14 That's the little sheets. 14 A. Michael Herring. 15 A. Yes. 15 Q. And what is your job title? 16 Q. Does that accurately reflect the basic 16 A. I'm the chief financial officer of 17 areas of testimony in your written direct 17 Pandora. 18 testimony? 18 Q. How long have you been at Pandora? 19 A. It does, actually. Yes. 19 A. Since February 2013. So a little over 20 Q. And did you file written rebuttal 20 two years. 21 testimony in this matter? 21 Q. Okay. And where were you before 22 A. I did as well. 22 Pandora? 23 Q. Okay. Can you flip over to the second 23 A. Immediately before, I was working for 24 page of the demonstrative. 24 Adobe Systems as the vice president of global 25 A. Okay. 25 operations.

3335 3337

1 Q. And does that slide reflect the topics 1 MR. LARSON: Should we address these 2 that are generally covered in your written rebuttal 2 one at a time, Your Honor, or -- 3 testimony? 3 MR. CHOUDHURY: Yeah. I can do it 4 A. It does. 4 whichever way is most convenient. 5 Q. And do you understand that you are here 5 CHIEF JUDGE BARNETT: All right. Yes. 6 today to discuss your rebuttal testimony as 6 So, Mr. Larson? 7 compared to your direct testimony? 7 MR. LARSON: Yes. So -- well, this is 8 A. Yes. I understand that. 8 a statement by Spotify's CEO, Mr. Ek, regarding his 9 Q. Okay. And turning to the topics in 9 view of whether or not Pandora is a competitor. 10 your rebuttal testimony, will you be discussing 10 Frankly, we felt it would be useful, 11 today the third topic on the list, Pandora's direct 11 since there's a lot of testimony on that subject or 12 license agreements? 12 discussion of that topic, to hear from Spotify 13 A. Other than -- or no. I'm not actually 13 itself what its views are on the subject. 14 going to talk about that here. We'll talk about 14 CHIEF JUDGE BARNETT: Overruled. 15 that in relation to my direct testimony. 15 Go ahead. Next one. 16 Q. Okay. And will you be discussing the 16 MR. CHOUDHURY: If we go to Paragraph 17 fourth and fifth topics that are identified here on 17 51, we believe this is improper rebuttal. 18 the slide? 18 MR. LARSON: Shall I proceed, Your 19 A. We won't be talking to them in depth. 19 Honor? 20 We will rely on my written testimony other than 20 CHIEF JUDGE BARNETT: Yes, please. 21 some discussion of revenue definitions. 21 MR. LARSON: So this is an agreement -- 22 Q. So I'll direct you to the binder in 22 or a discussion of an agreement which is appended 23 front of you, which has been -- take a look at the 23 between Pandora and Naxos. I guess a couple 24 first tab, if you would. 24 responses to this. 25 A. Okay. 25 JUDGE FEDER: Before you get to that,

3336 3338

1 Q. It's Pandora Exhibit 5016. 1 are we getting into restricted information? 2 Do you see that? 2 MR. LARSON: No. I don't think, for 3 A. Yes. 3 purposes of this discussion, I'll need to address 4 Q. Do you recognize that as your written 4 anything that's restricted. Thank you. 5 rebuttal testimony in this case? 5 First of all, the parties actually had 6 A. I do. 6 an agreement that rebuttal witnesses could discuss 7 MR. LARSON: Your Honor, I would offer 7 new developments that occurred after the direct 8 Pandora Exhibit 5016 into evidence. 8 testimony if they were related to topics in the 9 MR. CHOUDHURY: Your Honor, we have a 9 direct testimony of the witness. 10 number of objections that we had submitted before 10 And I think this pretty plainly fits 11 to this written direct testimony. I'm happy to go 11 under that. This is a new direct license that 12 through them now. 12 Pandora signed after the direct testimony was 13 CHIEF JUDGE BARNETT: Okay. 13 submitted, not unlike the Merlin agreement. 14 MR. CHOUDHURY: So first objection 14 JUDGE STRICKLER: The January -- new 15 would be in Paragraph 10, the amended rebuttal 15 articles of January of 2015? 16 testimony. 16 MR. LARSON: That's correct, Your 17 JUDGE STRICKLER: This 5016, is this 17 Honor. 18 the amended written rebuttal testimony? 18 I would say, even absent that 19 MR. CHOUDHURY: Yes. The amended 19 agreement, this pretty squarely fits with what 20 written rebuttal testimony. 20 happened in Satellite II where Professor Knoll, you 21 And I believe counsel said that the 21 may recall, Judge Barnett, on rebuttal offered a 22 amendment is related to Your Honor's order about 22 number of new direct licenses that SiriusXM had 23 this testimony. That's the amendment. 23 entered since his direct testimony. Those were 24 Paragraph 10, we think that's plainly 24 admitted. 25 hearsay. 25 I think this also fits under the

3339 3341

1 judge's decision related to the 3E agreements, 1 CHIEF JUDGE BARNETT: Okay. Exhibit 2 which were held to be allowable testimony by 2 5016 is admitted. 3 Professor Rubinfeld, I think on the theory that 3 (Pandora Exhibit No. 5016 was admitted 4 they represented agreements that the other side 4 into evidence.) 5 hadn't considered. 5 BY MR. LARSON: 6 So I think, for all those reasons, not 6 Q. Mr. Herring, could you look at Pandora 7 to mention just the judge's general discretion, 7 Exhibit 5018 in your binder. 8 that this agreement be informative and should come 8 A. Yes. 9 in. 9 Q. And do you recognize that as the 10 MR. CHOUDHURY: And, Your Honor, on 10 Pandora and Naxos agreement that we were just 11 that point, as you'll recall, Section 3E, the 11 discussing? 12 Court's order was that it was improper rebuttal. 12 A. I do. 13 And the Court ordered discovery and allowed 13 MR. LARSON: Your Honor, I would offer 14 actually for surrebuttal by the services on that 14 Pandora Exhibit 5018 into evidence. 15 point. 15 CHIEF JUDGE BARNETT: 5018 is admitted. 16 To the extent counsel is representing 16 (Pandora Exhibit No. 5018 was admitted 17 that this is equated to the Section 3E, we 17 into evidence.) 18 certainly weren't allowed surrebuttal on the Naxos 18 MR. LARSON: And since we're dealing 19 agreement. 19 with some procedural matters, we believe it would 20 JUDGE STRICKLER: How do you respond to 20 be, sort of for the sake of efficiency, preferable 21 counsel's point that you had an agreement that 21 to have Mr. Herring, when he returns for his direct 22 information about new developments could be 22 testimony where he'll be doing a detailed 23 introduced? 23 discussion of the Merlin-Pandora agreement, to also 24 MR. CHOUDHURY: Well, you know, I don't 24 discuss the Merlin-Naxos agreement at that time 25 actually -- I mean -- again, I may be 25 rather than today, if that's okay with Your Honors.

3340 3342

1 misunderstanding -- I don't actually understand 1 MR. CHOUDHURY: Your Honor, this is 2 that that was the agreement. 2 testimony -- as his own demonstratives testify, 3 I thought the agreement was that direct 3 these are in his rebuttal report. And we should be 4 witnesses could respond to rebuttal witnesses, but 4 able to examine him -- because he's not just 5 they could not rely on new data or studies. 5 describing his testimony. 6 Now, it may be that counsel is 6 If you look at that paragraph's 7 representing that this is not a new data or study; 7 testimony, he's comparing it to the interactive 8 so, therefore, it's allowable. That was just my 8 services. And he is making a comparison in his 9 understanding. And again, you know, I -- honestly, 9 testimony. 10 I can't take a stand here. I don't actually think 10 I'm not saying that Mr. Herring should 11 that was the agreement, but I don't want to 11 be precluded from discussing in the direct phase 12 misrepresent because -- 12 the Merlin report. But I am saying we should be 13 JUDGE STRICKLER: Maybe you should see 13 allowed to examine him about the contents of his 14 the agreement. 14 testimony that is in his rebuttal submission. 15 CHIEF JUDGE BARNETT: I think this is a 15 MR. LARSON: If I may, just so there's 16 factual development. 16 no misunderstanding, my request was around Naxos. 17 Your objection is overruled. A 17 And even though it appears in the rebuttal 18 relevant factual development. 18 testimony, it might more appropriately, just for 19 Mr. Choudhury? 19 efficiency reasons, be discussed when Mr. Herring 20 MR. CHOUDHURY: Yes. Otherwise, you 20 is here on direct. 21 know, subject to our continuing objection about the 21 So it's not a matter of precluding 22 Pureplay, that, you know, we said it should be -- 22 cross on that topic, just what -- when it makes the 23 the testimony should be originally admitted but we 23 most sense to do it. 24 can discuss it later, we have no further 24 MR. CHOUDHURY: And, Your Honor -- 25 objections. 25 well, it would preclude cross in this phase of the

3343 3345

1 proceeding. And that's our point. 1 business and -- in terms of not driving revenue but 2 CHIEF JUDGE BARNETT: It's in his 2 really driving revenue efficiency and how we grow 3 rebuttal testimony. 3 revenue in the context of growing our service. 4 MR. LARSON: Fair enough. 4 So the first one, ad load, is -- you 5 CHIEF JUDGE BARNETT: Is that all, 5 know, importantly, it represents the maximum number 6 Mr. Choudhury? 6 of ads per hour that we might serve in a given 7 MR. CHOUDHURY: On that, yes. 7 time. This is a threshold that's set within the 8 CHIEF JUDGE BARNETT: Okay. Thank you. 8 system. And that's a threshold that changes over 9 And so 5018 is admitted. 9 time. 10 BY MR. LARSON: 10 JUDGE STRICKLER: I was going to ask 11 Q. Mr. Herring, let's start with the first 11 you, maximum subject to what constraint? 12 topic that's on the list of topics related to your 12 THE WITNESS: It's a constraint set by 13 rebuttal testimony. 13 Pandora. 14 A significant section of your rebuttal 14 JUDGE STRICKLER: And what are the 15 testimony describes Pandora's development of the 15 constraining -- without getting into the details of 16 Internet radio advertising market; is that correct? 16 it, what are the constraining factors? 17 A. That is correct. 17 THE WITNESS: The constraining factors 18 Q. And just take a look at Paragraph 3 of 18 for maximum ad load are what we think the right 19 your rebuttal testimony. 19 level to optimize for revenue, both looking 20 Does that spell out the reasons that 20 short-term and long-term. 21 you have included this discussion in your 21 So it takes into account our -- how 22 testimony? 22 much ads we think we can sell and at what price 23 A. It does. 23 we're selling those ads. So what the optimal 24 Q. And just briefly, could you relate what 24 num -- you know, revenue generating ability is on 25 parts of SoundExchange's case you are responding to 25 that hour.

3344 3346

1 in this section of your testimony? 1 Second thing it looks at is the impact 2 A. Well, specifically the three main 2 of serving those ads on the loyalty or the 3 assertions, the first of which is that we are 3 retention of that listener over time. So not just 4 purposefully not monetizing our listener base; 4 in that existing session but, you know, how 5 second, that, if we wanted to monetize further, we 5 frequently that listener will come back and listen 6 could just add more ads or increase the number of 6 to Pandora and how long they listen in the future. 7 ads we sell, and that we're currently not selling 7 And one of the benefits of having a 8 as many ads as we could; and lastly, you know, if 8 connected service is we can measure those things 9 we just sold more ads, then we could afford to pay 9 very granularly and adjust our ad load to optimize 10 the rates that SoundExchange is asking. 10 either the number of ads and the number of times 11 Q. Okay. So before we unpack that, let's 11 people come back as well as the revenue that's 12 do one more thing, which is to take a look at some 12 generated by those ads during those hours. 13 of the terminology that you use in this section. 13 JUDGE STRICKLER: People who come back 14 Could you turn to the next 14 are -- that's a component of market share, correct? 15 demonstrative, please. 15 THE WITNESS: It is. When we talk 16 And do you recognize this demonstrative 16 about market share, we talk about 10 percent of the 17 as identifying some of the terms that are used in 17 market. It's the hours listened to radio. 18 your rebuttal testimony? 18 So it's important that we deal -- when 19 A. I do, yes. 19 we talk about growing market share, we talk about 20 Q. And could you just walk the Court 20 getting more and more people's radio listening. 21 through these terms and explain what they mean and 21 JUDGE STRICKLER: Thank you. 22 the relation between them? 22 THE WITNESS: Sure. 23 A. Sure. I'd be happy to. 23 BY MR. LARSON: 24 So these are important terms in 24 Q. Why don't you continue through the 25 understanding how Pandora -- how we talk about our 25 slide, if you would, with moving on to the second

3347 3349

1 line, ad inventory. 1 the optimal price to build the right kind of 2 A. So when we talk about inventory to sell 2 business. 3 or ad inventory, it's a simple calculation of that 3 We have invested aggressively in 4 ad load or the number of ads per hour times the 4 building out our sales team. There are over 430 5 number of hours that we will stream in a given 5 bag carrying salespeople, quota bearing 6 month or quarter. 6 salespeople, in, you know, national markets as well 7 Then we apply to that a sell-through 7 as the, you know, 37 local markets around the 8 rate. And this is one of the big efforts of our 8 country with 138 people in those local markets. 9 sales team. But essentially what percentage of 9 We sell to digital buyers. We sell to 10 that ad load, the number of ads -- or what 10 radio buyers. So we built a massive infrastructure 11 percentage of that ad inventory, the number of ads 11 to address the opportunity. And we think we're 12 we'll have in a month we'll be able to sell based 12 industry leading in that capacity when it comes to 13 upon our sales capacity. The number of salespeople 13 Internet radio. 14 we have, the number of clients, what our pipeline 14 We, you know, have tried to push that 15 looks like, et cetera. 15 sell-through rate up aggressively and catch up with 16 You take that -- those in -- 16 the number of hours we're driving. 17 sell-through rate times the ad inventory. And 17 And our hours have continued to grow 18 that's the quantity portion of our revenue. If you 18 aggressively. You know, hitting 10 percent market 19 think of revenue as very simply just the price we 19 share was an all-time high last quarter. But we've 20 sell something at times the number we sell or the 20 been growing those hours at the same time we've 21 quantity, the quantity comes from the inventory 21 been growing sell-through rate faster. And so that 22 times the sell-through rate. And the price is the 22 RPM continues to grow. 23 average price we're selling those ads for. That 23 And I think we've done that as well as 24 gets our revenue number. 24 we possibly can. It's been a top priority of the 25 And a very important metric that we 25 company. You think half our company is essentially

3348 3350

1 measure internally -- and it's going to be 1 made up of sales or sales-support employees. So 2 discussed a lot, I believe, during these 2 these are people who are a hundred percent 3 proceeding -- is RPM or revenue per thousand hours. 3 compensated on hitting, you know, the most 4 It's important to remember -- so that's 4 aggressive revenue number that we can hold them to 5 the dollar revenue we're able to generate per 5 within constraints. 6 thousand hours of playing music for our listeners. 6 Those constraints are around yield and 7 That's a thousand hours to individual listeners. 7 maximization, driving the right kind of advertising 8 It's important to remember we stream a 8 balanced with what our long-term opportunity is. 9 million hours of music to our listeners. That's 9 Q. So has Pandora actually increased its 10 the equivalent, you know, of -- or a million 10 ad load in recent times? 11 streams to our listeners is equivalent to 16 or 17 11 A. So again, ad load being the maximum 12 plays on, you know, Z100, a large New York radio 12 number of ads, since I joined the company, which 13 station. So it's -- just keeping that in mind as 13 hasn't been that long ago, we have increased our ad 14 we talk about the number of streams, number of 14 load more than 50 percent, closer to 70, 80 15 hours. 15 percent. 16 Q. Thank you. 16 When I joined, it was about three ads 17 So with that description in mind that 17 per hour. We increased it to four ads per hour and 18 you just gave, what is your reaction to the 18 then six ads per hour. And frankly, just recently, 19 suggestion by various SoundExchange witnesses that 19 in the last few months, we've increased in some 20 Pandora should simply increase its ad load and sell 20 markets to seven ads per hour. 21 more ads? 21 So we're just talking about audio ads 22 A. Well, you know, I think we have been 22 here, but it's the one that's probably most 23 working very hard to do that, frankly. That's been 23 applicable to this conversation. 24 a major focus of our business for a long time, is 24 Q. And how do you decide when it's 25 to sell as many ads as we can and to sell them at 25 appropriate to increase the ad load?

3351 3353

1 A. So that is both a science and an art. 1 A. I do. 2 So when we talk about the science of it, it is -- 2 Q. Is Pandora attempting to reach the 3 it involves how aggressively we can move that ad 3 point where it can run as many ads as terrestrial 4 load up based upon how it affects listening. 4 radio? 5 So if we add to many ads in an hour, it 5 A. Well, the short answer is no. And 6 restricts -- it reduces the amount of listening 6 there's good reason for that. That criticism is 7 that occurs. And we can measure that directly. 7 frankly somewhat naive. It doesn't understand how 8 JUDGE STRICKLER: You said that you 8 our business actually works. 9 increased the ad load from three to four, up to six 9 In terrestrial radio, they have one 10 ads per hour since you've been at the company. And 10 source of revenue, essentially, which is add more 11 I know you mentioned it, but it's not in the 11 audio ads. So getting to 12 or 15 minutes or -- 12 rebuttal testimony. 12 you know, certain sports -- AM stations run as many 13 When did you start again? 13 as 22, 23 minutes of ads an hour. 14 THE WITNESS: Oh, sorry. February 14 That is -- Pandora would be ignoring 15 2013. So about 26 months ago. 15 maybe the largest part of the opportunity. Keep in 16 JUDGE STRICKLER: Thank you. 16 mind our audio advertising is sold aggressively 17 THE WITNESS: Sure. 17 into radio budgets, which is about a $17 billion 18 BY MR. LARSON: 18 opportunity. 19 Q. And you mentioned that you have, I 19 But we also sell digital display 20 think you said, people looking at this. 20 advertising. On both Web and mobile we sell video 21 Just give us a sense of how many people 21 advertising. Together those are 40-, $50 billion 22 and what type of people are looking at this 22 marketing opportunities. 23 question of ad load. 23 So the addressable market for Pandora 24 A. Right. So a lot of people in the 24 expands by a multiple just by selling different 25 company. We have 18 people who just think about 25 kinds of advertising. So we optimize across

3352 3354

1 yield optimization at the company -- so these are 1 multiple forms of advertising and opportunity and 2 in our sales organization -- that are constantly 2 as well as thinking about the long game, which is, 3 looking for -- looking at the inventory we have 3 you know, how do we drive the optimal revenue and 4 available and what yield we're getting out of that 4 profitability as a business not in Q2 2015 but in 5 availability and what our pipeline looks like in 5 2017, 2020, 2025, 2035. 6 terms of approaching a hundred percent sell-through 6 You know, we talk about being a hundred 7 rate. 7 year company. It seems maybe a little precocious 8 Now, we don't target a hundred percent 8 as a nine-year-in sort of business model. But 9 sell-through rate. We target about an 85 percent 9 everyone started at zero at some point. And we 10 sell-through rate. 10 believe that there's a lot of potential to grow 11 That has a lot to do with the dynamics 11 that the right way. 12 of the ad market and the need to be able to 12 Now, let's put that in context. If you 13 guarantee delivery. Since, even though we have a 13 believe that terrestrial radio -- and we do in our 14 lot of -- we can predict hours -- in a -- nothing 14 own analysis -- has their RPM or revenue per 15 is for sure. Lots of things can cause hours to 15 thousand hours of listening at about $73. You 16 fluctuate. So about 85 percent is considered sold 16 know, in Q4, which is our seasonally best 17 out. 17 advertising period, you know, we topped $50 RPMs 18 When we start hitting that point, 18 overall for the first time. 19 that's when we look to try and raise ad load, which 19 So we were -- you know, a substantial 20 increases capacity. And then we work to drive that 20 percentage of terrestrial radio's RPM while 21 again. 21 delivering a fraction of the audio ad load. 22 Q. You mentioned in your testimony that 22 What that does is it expands our 23 some SoundExchange witnesses have faulted Pandora 23 ability over time, we believe, to drive top line 24 for not running as many ads as terrestrial radio. 24 revenue. And that's what powers the rest of our 25 Do you remember that? 25 business.

3355 3357

1 JUDGE STRICKLER: When you say you 1 CERTIFICATE OF COURT REPORTER 2 were a substantial percentage of terrestrial radio 2 3 when you hit 50, what percentage were you? 3 I, Bonnie L. Russo, do hereby certify that the 4 THE WITNESS: Well, if terrestrial 4 foregoing transcript is a true record of the 5 radio's at $73, so it's approximately two-thirds, 5 proceedings to the best of my ability, that I am 6 right? 60, 65 percent. 6 not related to or employed by any of the parties 7 BY MR. LARSON: 7 involved in these proceedings, and, further, that I 8 Q. And you believe it's possible for 8 am not a relative or employee of any attorney or 9 Pandora to get up to the RPMs you see on 9 counsel employed by the parties hereto, or 10 terrestrial radio? 10 financially interested in the proceedings. 11 11 A. Well, I know it's possible. In certain 12 12 markets where we've been in -- had local sellers, ______13 which is our highest price offering because it's 13 Notary Public 14 the most targeted -- you know, that's the other 14 15 thing that really differentiates us from 15 My Commission Expires: 16 terrestrial radio is we can target down to the user 16 May 16, 2016 17 based on lots of demographic data, geographic data, 17 18 psychographic data that's gleaned from their 18 19 activities. 19 20 You know, we're able to drive RPM in 20 21 those markets, you know, in excess of the 73, $75 21 22 that we think terrestrial radio gets. Our top 22 23 three markets today operate well above that number. 23 24 CHIEF JUDGE BARNETT: Mr. Larson, we 24 25 have to stop. 25

3356

1 MR. LARSON: Okay. 2 CHIEF JUDGE BARNETT: We will be at 3 recess until 9:00 in the morning. 4 (Whereupon, the proceeding was 5 concluded at 4:23 p.m.) 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

3358 Before the

UNITED STATES COPYRIGHT ROYALTY BOARD

Library of Congress

Washington, D.C.

------: In Re: : Docket No. : 15-CRB-0001-WR Determination of Royalty : (2016-2020) Rates and Terms for : Ephemeral Recording and : Volume 13 Digital Performance of : Pages 3358-3667 Sound Recordings (Web IV) : ------:

RESTRICTED MATERIAL REDACTED PURSUANT TO PROTECTIVE ORDER

DAY 13

Washington, D.C.

Wednesday, May 13, 2015

The hearing in the above-entitled matter was convened at 9:08 a.m.

BEFORE COPYRIGHT ROYALTY JUDGES:

SUZANNE M. BARNETT, CHIEF JUDGE

DAVID R. STRICKLER, JUDGE

JESSE FEDER, JUDGE

3367 3369

1 - - - 1 "radio market." What did you mean by that? 2 MICHAEL HERRING, 2 A Well, that is a general term we refer to 3 having been previously duly sworn, testified as 3 all radio listening. So that includes terrestrial 4 follows: 4 radio listening. It includes satellite radio 5 CONTINUED DIRECT EXAMINATION BY 5 listening and includes all Internet radio COUNSEL 6 listening, including Pandora. So when we say 6 FOR PANDORA MEDIA 7 we're 9, 10 percent of radio listening in the 7 BY MR. LARSON: 8 United States, that's Pandora's share of the total 8 Q Good morning, Mr. Herring. 9 number of hours that's being consumed of the radio 9 A Good morning. 10 format. 10 Q I asked you yesterday about the argument 11 JUDGE FEDER: Do you include Spotify or 11 that Pandora should just increase its ad load and 12 other on-demand services in radio? 12 sell more ads, and you indicated, I believe, that 13 THE WITNESS: We do not. 13 that would affect listening. Do you recall that? 14 BY MR. LARSON: 14 A I do. 15 Q Why is the statistic about your share of 15 Q So with respect to advertising sales, 16 the radio market notable in terms of advertising 16 why is the level of listening a concern? 17 opportunity? 17 A Well, the level of listening is an 18 A That statistic was actually developed to 18 important factor in the number of ads we have to 19 support our sales marketing efforts. So radio 19 sell. As we talked about yesterday, ad inventory 20 advertising in particular, I'm looking for a 20 is directly related to both ad load and the number 21 certain reach within their target audience before 21 of hours. So if we adversely affect the listening 22 they'll even consider buying into that audience, 22 experience too greatly, it can limit the number of 23 buying advertising to address that audience. And 23 hours they actually stream and therefore limit the 24 that's true for national, it's true for national 24 advertisements we have available to sell. 25 spot radio, so buying specific markets, whether it 25 Q Let me ask a related question. If you

3368 3370

1 increase the supply of ads, would there 1 be the southeast or the southwestern United 2 necessarily be demand for that additional supply? 2 States. It's true with local markets. When you 3 A No. In my experience, supply does not 3 get to local markets, you need to have 6 to 4 drive demand. In fact, we've spent millions of 4 8 percent of the target market or the equivalent 5 dollars building demand, developing demand at 5 of the top five terrestrial radio markets within 6 Pandora. Our efforts to have our product -- our 6 any given local market before advertisers will 7 service measured by Triton and certified by the 7 consider buying on your platform. It's purely an 8 MRC, which is a governing body that establishes 8 opportunity cost for them to even go through the 9 the accuracy of that measurement, and then have 9 machinations of evaluating your service and 10 that integrated into the planning platforms and 10 creating the creative, and placing their ad, et 11 the buying platforms that advertisers use, all of 11 cetera. They have to know they're getting the 12 that is about getting demand, the 17 billion 12 right reach for their advertising, they'll get the 13 dollars of radio demand, to consider Pandora as 13 right ROI. 14 part of their advertising campaigns. 14 Q What's the size, roughly, of that local 15 So we're still in the early days of 15 advertising sales market? 16 that. If you think about terrestrial radio has 16 A Well, it's the majority. So if you 17 been selling to that audience now for decades, 17 think about the 17 billion dollars in total radio 18 we've been selling to that audience now for about 18 advertising in the United States, it's about two 19 three years. So we're still early in the 19 thirds of that number. So that is why we have 138 20 development phase of getting that demand out. 20 of our 430 sales professionals targeted 21 Just adding more ad capacity doesn't necessarily 21 specifically in market to address local 22 mean it gets sold. It's actually a process that 22 advertising. We also have additional inside sales 23 makes that happen. 23 that address the markets, the local markets -- 24 Q You mentioned yesterday that Pandora is 24 beyond the 37, we have local people in. We 25 10 percent of, what I believe you called, the 25 address that with inside salespeople because

3371 3373

1 that -- that is incredibly valuable advertising, 1 be spent every quarter just to drive that revenue. 2 not just because it's large, but it's also at a 2 Last quarter it was 35 percent. As we move 3 higher price because it's more targeted. 3 through the year, we get leverage on that percent 4 Q Following on what you just said, your 4 of revenue, and it gets into the mid -- low to mid 5 written rebuttal testimony describes at some 5 20 percentage points of our revenue just to drive 6 length the investments made by Pandora in sales 6 that revenue on an ongoing basis. 7 force and systems and the like. How has Pandora's 7 Q Let's shift gears a bit, if we could. 8 profitability been impacted by those investments? 8 A Okay. 9 A Well, so if you think about 9 Q If the judges were to adopt rates in the 10 profitability in two dimensions, either it's gross 10 range suggested by SoundExchange here, what do you 11 profit and then there's operating profit or net 11 project to be the likely impact on Pandora's use 12 profit. Gross profit is purely just the revenue 12 of music? 13 less the cost of driving that revenue. And those 13 A Well, you know, the rates proposed by 14 investments have actually increased that gross 14 SoundExchange are significantly higher, let's just 15 profit pretty significantly. But in order to 15 say for argument, 80 percent higher than the rates 16 continue to drive our own market share of that 16 we pay today. So if our current content costs are 17 advertising over the last few years, we've 17 at 50 percent of revenue and our operating 18 invested aggressively in sales and marketing. In 18 expenses are at 44 percent of revenue, so we have 19 fact, our two points of investment have been in 19 a 6 percent EBITDA margin, increasing that 20 product development sales and marketing, but the 20 50 percent 80 percent would essentially take our 21 vast majority to sales teams, to our sales 21 content cost, under today's monetization 22 execution. 22 thresholds, to 90 percent of our current revenue 23 That investment is what has, you know, 23 base. 24 eaten up all of our gross profit, essentially. 24 Obviously, that 4 to 6 percent would 25 You know, last year we had about 6 and a half 25 make us, you know, immediately very unprofitable.

3372 3374

1 percent of an EBITDA margin, or earnings before 1 Now we wouldn't operate at a minus 40 percent, 2 interest, taxes, depreciation, and amortization. 2 minus 35, 40 percent negative operating margin, so 3 So we follow that metric because it is -- it's a 3 we would have to take dramatic steps to reduce 4 cash flow approximation or free cash flow 4 costs to stay in business. You can only operate 5 approximation. 5 at a large loss for a short period of time. 6 So investors look at EBITDA as a way of 6 Because content costs are our largest 7 understanding whether you're profitable or not, 7 expense line, the first thing we would look at is 8 you know, taking out some of the noise of stock 8 how do we reduce those costs. Obviously, price is 9 comp, et cetera. If our gross profit has been 9 set in this room. So quantity becomes the other 10 improving, that EBITDA margin has been improving, 10 side of that thing. So we would immediately start 11 not as quickly because we've been reinvesting so 11 reducing costs by reducing quantity. We're 12 aggressively into the sales opportunity. 12 reducing the hours of listening. There's lots of 13 Q Are those investments in the sales 13 ways we can do that. We have done that twice in 14 opportunity, as you call it, up-front investments 14 our past, as is in my testimony, where we've 15 or are they ongoing in nature? 15 limited listening hours. 16 A So this isn't like where you buy an 16 Q Are those the caps you spoke of? 17 apartment building and you put a big outlay up 17 A Those are the caps. We've done it most 18 front and then you charge rent and you pay a low 18 recently in 2013. I was not at Pandora the first 19 maintenance over time and you earn your return. 19 time we did it in 2009, but I was on the team that 20 In the advertising business, it's a constant 20 executed that cap in 2013. 21 investment. You know, I like to say that every 21 It would have to be much more draconian. 22 quarter we start at zero in terms of driving 22 That reduced our costs about 20 percent during the 23 revenue for that quarter. So that investment is 23 six months that we had that -- our content costs, 24 an ongoing amount, ongoing amount of investment. 24 about 20 percent during the six months that cap 25 It's -- 20 to 30 percent of our revenue needs to 25 was in place. But, obviously, if that structure

3375 3377

1 was 80 percent higher, we would have to look for a 1 THE WITNESS: I do, yes. 2 much, much larger impact. 2 JUDGE STRICKLER: Okay. With that 3 Q When you say -- 3 backup, how did they affect Pandora's financial 4 CHIEF JUDGE BARNETT: Excuse me. We had 4 and economic position, those listening caps, if 5 at least one witness who suggested that perhaps 5 you can answer both with regard to 2009 and 2013 6 executive salaries would be a place where cutting 6 separately if there's a distinction. 7 could happen. Is that something Pandora would do? 7 THE WITNESS: That's a great question. 8 THE WITNESS: So executive salaries 8 So the short-term effects were actually relatively 9 today probably represent, in 2015 -- I just want 9 positive in the context of our financial 10 to get this right, so pardon me if I pause. We 10 statements in both cases. I'll speak directly to 11 have eight people on our management team. It 11 2013 just because I was one of the architects of 12 would be about maybe right around 3 percent of 12 that program. 13 total expenses. So you could cut executive 13 It reduced the hours significantly, 14 salaries in half. The first thing that would 14 which brought our content costs down 15 happen is you would probably lose those 15 significantly, which allowed us to take those 16 executives. Keep in mind, Silicon Valley is a 16 dollars that we saved from not paying royalties 17 incredibly competitive place, not just for 17 and invest them aggressively in sales and drive 18 engineers, which you read about quite often, but 18 monetization improvements that much faster than we 19 also for executives that drive these technology 19 otherwise would have been able to do. That 20 companies. 20 allowed us to create gross margin pretty 21 Now, that may be not something people 21 dramatically. 22 like to hear, but that is -- the reality is that 22 Let me give you an example, in Q1, 23 innovative, aggressively growing, and 23 2013 -- remember, I started in February 2013. In 24 market-defining businesses like Pandora need the 24 that quarter our mobile RPM was around $24. So 25 people who are most skilled and most experienced 25 $24 of revenue per thousand hours. Our mobile

3376 3378

1 in driving those businesses to lead them. So 1 LPM, or licensing costs per thousand hours, in 2 compensation packages often are directly 2 that quarter was around $21. So the gross 3 associated with that. 3 margin -- 4 So with attracting the best and the 4 JUDGE STRICKLER: That's pre or post 5 brightest, hopefully -- hopefully we're hiring the 5 cap? 6 best people we possibly can. But even in that 6 THE WITNESS: That's pre cap. So Q1, 7 context, relative to royalties, it's a much 7 2013. The cap went in place March -- 8 smaller percentage that would -- that we could 8 approximately March, April of 2013. 9 impact through limiting executive salaries. 9 JUDGE STRICKLER: So in Q2. 10 JUDGE STRICKLER: Another question. You 10 THE WITNESS: Yes, so really, 11 talked about the caps on listening. 11 effectively, from a cost perspective, in Q2. 12 THE WITNESS: Yes. 12 That, obviously, is not much margin, 13 JUDGE STRICKLER: You weren't there for 13 right? It's 2 -- $2 or $3 of per thousand hours 14 the first round in 2009 but you were there for the 14 streamed. And that's just gross margin, before we 15 second round in 2013. 15 pay for anything -- any servers to stream the 16 THE WITNESS: Yes. 16 services, salespeople to go sell the revenue, 17 JUDGE STRICKLER: Let me just ask you a 17 accountants to count, you know, to collect the 18 background question. Did you get any 18 revenue and then pay the royalties. 19 institutional knowledge by looking into the books 19 Reducing the cost and investing in sales 20 as to how the listening caps affected Pandora in 20 allowed us to grow monetization significantly in 21 2009? 21 that time frame. Keep in mind, here we are two 22 THE WITNESS: I did, yes. 22 years later, in the same quarter, that $24 is now 23 JUDGE STRICKLER: So you have an 23 $38. It depends on how you measure it, but it's 24 understanding of how the listening caps affected 24 multiples higher. 25 Pandora both in 2009 and 2013? 25 JUDGE STRICKLER: 38, you say?

3379 3381

1 THE WITNESS: Yes. 1 THE WITNESS: The licensing under the -- 2 JUDGE STRICKLER: That's mobile RPM in Q 2 JUDGE STRICKLER: Oh, just fewer hours. 3 2013? 3 THE WITNESS: Just fewer hours, yes. 4 THE WITNESS: No, that's Q1. Sorry, Q1, 4 The rate per hour doesn't change with the caps. 5 2015. So I was saying two years later, that's 5 JUDGE STRICKLER: So your net for cost 6 much higher, but back then it was much lower. 6 of goods was 7, went up to 7. 7 JUDGE STRICKLER: But the cap -- 7 THE WITNESS: Yes, net gross profit 8 THE WITNESS: Yes. 8 approximately, yes. 9 JUDGE STRICKLER: My question was before 9 JUDGE STRICKLER: But you didn't keep 10 and after the caps. There's the caps in Q 2015. 10 that up. You eventually eliminated the cap. 11 So let's take 2013. 11 THE WITNESS: We did. 12 THE WITNESS: So as we move through 12 JUDGE STRICKLER: If it was more 13 2013 -- sorry. Let me slow down. We're able to 13 profitable from quarter to quarter, from 3 to 7 on 14 take -- to improve our RPM significantly in Q2, 14 the RPM basis, why did you remove the caps at all? 15 2013. 15 THE WITNESS: Because in that six-month 16 JUDGE STRICKLER: So we talk apples to 16 period -- 17 apples, this will be mobile RPM. 17 JUDGE STRICKLER: Which six months now 18 THE WITNESS: Yes, mobile RPM. 18 are we talking about? 19 JUDGE STRICKLER: What was mobile RPM in 19 THE WITNESS: Six months is the cap 20 Q2, 2013 after the caps? 20 time, so from March -- 21 THE WITNESS: I don't remember exactly. 21 JUDGE STRICKLER: The two Qs, Q1 and Q2. 22 I think it was around $28 or so. 22 THE WITNESS: Yes, Q1 and Q -- Q2 and 23 JUDGE STRICKLER: So approximately 28, 23 Q3. We were able to dramatically increase our 24 which is an improvement over the 24 from the 24 ability to monetize in that two months. So that 25 quarter before, when there were no caps. 25 when we got to September, essentially, we were

3380 3382

1 THE WITNESS: Yes, and that's a benefit 1 comfortable releasing the cap in the context of 2 of two things, growing revenue efficiency and 2 being able to monetize each hour better than we 3 lower hours. Remember, RPM is the -- the 3 were before. 4 numerator is our revenue. The denominator is the 4 JUDGE STRICKLER: When you decided to 5 number of hours. So we've restricted the 5 remove the cap and you removed it in Q3, 2013; is 6 denominator, made it lower, and increased the 6 that right? 7 numerator, grew the revenue. 7 THE WITNESS: Yes, the end of Q3. 8 JUDGE STRICKLER: Did the increase in 8 JUDGE STRICKLER: Did you have a concern 9 the numerator have anything to do with the caps? 9 that, well, maybe we should -- when you were 10 THE WITNESS: Only in the context is it 10 wrestling with the decision, that maybe we should 11 freed up dollars to invest in salespeople. Does 11 keep the cap in place since we're continuing to 12 that make sense? 12 monetize -- I mean, we can see the RPM improves, 13 JUDGE STRICKLER: Yes, it does. So that 13 maybe we should just keep doing that for a while 14 was 28. 14 but you're going to have to figure out if and when 15 How about the mobile licensing costs. 15 you're going to make the switch -- 16 That's what, the 38 -- what's the cost? 16 THE WITNESS: Yes. 17 THE WITNESS: Yes, they would have 17 JUDGE STRICKLER: -- and it made more 18 stayed the same. On a per-hour basis, the 18 sense to grow the revenues, the net revenues? 19 licensing costs stay at $21 approximately. Just 19 THE WITNESS: Yes. So that brings us to 20 we had fewer total hours. 20 the second part of my answer to the question. So 21 JUDGE STRICKLER: But I thought with the 21 why did we take it off so soon if it had such a 22 caps, you have fewer hours. 22 positive impact on our financials. It's because 23 THE WITNESS: Fewer hours but the cost 23 any time you're restricting listening, you're 24 per hour stays the same. 24 kicking your most loyal and ardent fans off your 25 JUDGE STRICKLER: Why is that? 25 service. You're telling them to stop listening.

3383 3385

1 And loyal art and music listeners eventually will 1 BY MR. LARSON: 2 go somewhere else, whether they go and listen to 2 Q Just a couple follow-up questions on 3 terrestrial radio again or they go back to piracy. 3 that conversation. You described that in the cap 4 A lot of our listeners, you know, were streaming 4 situation you were able to, I think you said, 5 off pirate sites. Or they go use some other 5 invest aggressively in sales and marketing; is 6 service. 6 that right? 7 We don't have any research as to where 7 A That's correct. 8 they go, but we know millions of people hit that 8 Q Now, if the statutory royalties here 9 cap every month. And they either stop listening 9 were increased to levels sought by SoundExchange, 10 to music, which was unlikely, because they've 10 how would that affect your ability to similarly 11 already hit 40 hours in a month, or they went 11 invest aggressively, as you called it? 12 somewhere else. 12 A Well, we started talking about this a 13 JUDGE STRICKLER: Did you, in fact, see 13 little bit. We would start by reducing hours 14 any of that migration in Q2 of 2013? 14 significantly. That's problematic for long-term 15 THE WITNESS: We did. It's hard to 15 viability of the business because you start 16 measure exactly. We saw a lot of people hit the 16 driving away your most loyal listeners. You limit 17 cap. We saw some of those people -- 17 your ability to grow. Our ability to grow is 18 JUDGE STRICKLER: When you say "a lot," 18 predicated on our ability to grow market share, 19 how many, percentage-wise, hit the cap? 19 both from a listening perspective as well as a 20 THE WITNESS: I would say about 4 to 20 revenue perspective. 21 5 percent of the 70 million or so. 21 But probably most critical in terms of 22 JUDGE STRICKLER: I'm sorry to 22 the long-term prospects of Pandora is we would 23 interrupt. 23 dramatically reduce our ability to invest. You 24 THE WITNESS: It sounds like a small 24 know, we're talking about driving those gross 25 percentage. You know, it's your typical -- you've 25 profits up and reinvesting that back in the

3384 3386

1 got very loyal users who consume a lot down to the 1 business. When you collapse my gross profit 2 long tail who listen a few hours. Very, very 2 margin, you eliminate my ability to add additional 3 normal in a media business like this. It is very 3 dollars, whether it's additional engineers to 4 uncomfortable to make your most loyal user 4 develop product or, as I've already said, 5 unhappy, but you have to do that sometimes to do 5 additional salespeople and marketing resources to 6 what's right for the business. In this case, 6 drive revenue. 7 that's what the cap was about. We wanted to do 7 My expectation is that -- and we may 8 that only for as short a time as possible because 8 even have to reduce the resources we already have. 9 eventually we would lose those people forever. 9 That's a pretty big dramatic change, an 80 percent 10 JUDGE STRICKLER: How did the imposition 10 increase. So, you know, the result would likely 11 of the cap in Q2 impact the total royalties paid 11 be certainly less revenue, you know, decline in 12 for sound recordings to SoundExchange? 12 usage, I would guess. Maybe -- we would obviously 13 THE WITNESS: That's the -- we believe 13 try and retain the usage that was most profitable, 14 it lowered our royalty obligation about 14 but we would be losing our ability to develop 15 20 percent. We paid 20 percent less in those two 15 unprofitable segments, which is one thing we do 16 quarters than we would have otherwise. 16 now. By the way, we developed those unprofitable 17 JUDGE STRICKLER: That was the gain you 17 payments and still pay full freight from a royalty 18 had in that cost reduction, and, correspondingly, 18 perspective because we're paying on a per-song 19 there was a reduction in revenue to SoundExchange 19 basis. That would, you know, almost certainly 20 and hence to the record companies and the artists? 20 lead to a pretty dramatic reduction in royalties 21 THE WITNESS: Yes, it was probably, my 21 paid. 22 guess over that two-quarter period, it was about 22 JUDGE STRICKLER: And if the revenue 23 20 million dollars. 23 caps, excuse me, the listening caps cause a loss 24 JUDGE STRICKLER: Thank you. 24 of revenue to SoundExchange, that would seem to 25 THE WITNESS: Sure. 25 be, at least that analysis alone, a lose-lose for

3387 3389

1 both Pandora and for the record companies. But 1 listening cap, where are these people going? 2 then the question would become the one about 2 THE WITNESS: So, the research I've seen 3 migration. When you have the listener caps, where 3 says that a small, single-digit percentage of 4 do these individual listeners migrate -- to where 4 those people would consider paying for some 5 do they migrate? If they migrate to terrestrial 5 alternate service. The rest would go listen to 6 radio, it remains a lose-lose. If they migrate to 6 terrestrial radio, revert to piracy, or even if 7 another noninteractive service, where they're not 7 they went to other free services that are ad 8 burdened by the cap, it's a loss to Pandora. 8 supported like YouTube or something -- whether 9 THE WITNESS: Yes. 9 those survive in the current environment, I don't 10 JUDGE STRICKLER: It's no loss to the 10 know -- on an ARPU basis, they will get a much, we 11 radio companies and SoundExchange. If they 11 believe, much less dollars on a per-user basis. 12 migrate to interactive, likewise, it's a loss to 12 JUDGE STRICKLER: "They" being 13 Pandora, but no loss or a gain to the radio 13 SoundExchange and the record labels? 14 companies. 14 THE WITNESS: SoundExchange and the 15 So if we were looking at both sides of 15 record labels. 16 the ledger, that is to say, buyers and sellers, 16 We're monetizing on a per-hour basis 17 for us to complete the picture, we would have to 17 and, therefore, on a per-user basis, much higher 18 know where the -- to be able to understand where 18 than any other free service out there because of 19 these listeners go if you impose the price cap -- 19 all the investments we've talked about here, 20 the listener cap. Excuse me. 20 because we've been able to make those investments 21 THE WITNESS: Yes. So there are plenty 21 and because we're good at it. We only think about 22 of economists and people who do research and such 22 the music. We only think about monetizing free 23 who are much better at looking at the future, I'm 23 music. Other companies that have ulterior things 24 sure, than me testifying in front of you -- 24 going with it are selling operating systems or 25 JUDGE STRICKLER: Just -- 25 using data to influence these other parts of their

3388 3390

1 THE WITNESS: -- but I'll give you my 1 business or they're selling handsets. Their 2 opinion. 2 priorities are not always optimizing the 3 JUDGE STRICKLER: Okay. Despite your 3 monetization of that free user. 4 humble background. 4 JUDGE STRICKLER: Question for you. 5 THE WITNESS: I'm just an operator. I 5 Maybe this is beyond your area, but if you have a 6 just run a business. And that has its place, I 6 listener cap, you're now stopping listeners from, 7 suppose, in these conversations. 7 obviously, greater listening, listeners who have 8 So, first of all, it would absolutely be 8 already decided that they like a noninteractive 9 a lose-lose. We pay half of SoundExchange's -- 9 service. That's why they went -- now, maybe they 10 half of SoundExchange's revenue and a bigger 10 like only the Pandora noninteractive service. But 11 percentage of that, of their growth, comes from 11 do you have any knowledge as to whether or not 12 Pandora alone. Forget about Internet radio. So 12 those individuals simply migrated to another 13 you would think that that would be an important 13 noninteractive service since they've already 14 revenue stream for SoundExchange and then 14 revealed some sort of a preference for 15 subsequently the record labels to collect that. 15 noninteractive as opposed to going back to 16 JUDGE STRICKLER: Isn't it an argument 16 terrestrial radio? In other words, how can you 17 by Pandora in this proceeding that its value -- 17 keep them down on the farm once they use Pandora? 18 one of its great values to the radio companies is 18 THE WITNESS: Well, first of all, I 19 that it gets individuals who migrate from 19 believe that if rates are at this point where 20 terrestrial radio, where no royalty is paid, to 20 Pandora is unable to operate, there isn't anybody 21 Pandora, where it pays up to this huge amount 21 else, independent, at least, who can operate, 22 of -- huge percentage of SoundExchange's revenue? 22 right? I mean, we're -- we have the most 23 THE WITNESS: Yes. 23 effective sales and monetization machine in the 24 JUDGE STRICKLER: If suddenly we 24 industry by a long shot. So if we're unable to -- 25 switched -- throw that car into reverse with a 25 if we have to restrict, anybody else out there is

3391 3393

1 just going to get crushed. 1 problem. 2 JUDGE STRICKLER: Maybe there's some 2 JUDGE STRICKLER: And the smaller 3 proof, though, of that in the marketplace, whether 3 competing services didn't have that problem. That 4 you impose listener caps, whether there are 4 problem rears its head the larger you get. 5 listener caps imposed by other interactive 5 THE WITNESS: That's correct. 6 services. 6 JUDGE STRICKLER: And now let's bring in 7 THE WITNESS: No, because there really 7 the thing you talked about before, which was 8 aren't any other large ones. 8 having a -- my phrase, not yours, a deeper pocket, 9 JUDGE STRICKLER: Who were your major 9 you are a bigger organization. IHeart is Clear 10 competitors -- I understand that you welcome -- 10 Channel. Now we have Amazon -- excuse me, Apple. 11 THE WITNESS: Yes. 11 THE WITNESS: Right. 12 JUDGE STRICKLER: -- competition, but in 12 JUDGE STRICKLER: Another deeper pocket. 13 terms of market share, particularly the time, 13 THE WITNESS: You know, I mean, I don't 14 let's stick with Q1 and Q2 of 2013, who were your 14 know what Apple -- I've been an Apple shareholder 15 competitors in the noninteractive space? 15 for 20 years, but I don't -- and being the best 16 THE WITNESS: Yes. Our listener cap was 16 performing stock I've ever owned in my life. 17 itself -- you know, it's because of our success we 17 Probably if you owned it, I would say it's 18 had to do that. Because we're so much bigger, 18 probably also true. 19 we're paying these costs that we need to be able 19 But they don't give us a lot of detail 20 to afford. The other competitors in the space 20 as how iTunes Radio, for example, is doing. In 21 either have direct deals or are much smaller, you 21 the time frame we're talking about, iTunes Radio 22 know, in the noninteractive space, if you're 22 didn't exist yet so it hadn't launched. 23 talking about iHeartRadio, for example, where 23 One of the reasons why removing the cap 24 they're much smaller in terms of number of users 24 and kicking our most loyal people off, not kicking 25 and have -- 25 them off any longer was important because we --

3392 3394

1 JUDGE STRICKLER: You're talking about 1 you know, it's always increasing competition. We 2 the iHeart customized radio, right? 2 can get them back for only so long, but if we kept 3 THE WITNESS: Yes, exactly, the actual 3 that on for years, we would, of course, lose them 4 iHeartRadio property itself, which is about a 4 to other options. 5 fifth today as large as Pandora. 5 But Apple, of course, could pay, and 6 JUDGE STRICKLER: In Q1 and Q2 of 2013, 6 from my understanding from -- you know, is that 7 iHeartRadio had customized radio and did not 7 they have paid a lot more than Pandora has or a 8 impose listener caps contemporaneous with you? 8 significantly higher percentage. I have not seen 9 THE WITNESS: They didn't impose 9 any of their deals other than one was leaked to 10 listener caps. They were even much smaller, even, 10 the press from an indie a while ago. And even 11 then than they are today. 11 that one showed higher rates for iTunes Radio, not 12 JUDGE STRICKLER: So if you're saying 12 anywhere near the SoundExchange proposed rates but 13 because you are larger you had to impose the 13 still a higher rate. 14 listener caps because of the cost, it makes -- 14 And, you know, I'm sure that is, on its 15 THE WITNESS: And because we're 15 own, very unprofitable for Apple, but that's 16 independent. So -- that's another important 16 because they have a handset business that does 17 thing, is that iHeartRadio also has a parent that 17 pretty well. 18 has other revenue streams. 18 JUDGE STRICKLER: Thank you. 19 JUDGE STRICKLER: Let's move that off to 19 MR. LARSON: Just one. 20 the side for the moment. It sounds like what 20 MR. RICH: If I may, for the judge's 21 you're saying is, given the rate structure and the 21 benefit, a witness you'll be hearing from as early 22 cost structure, there are diseconomies of scale, 22 as tomorrow, Larry Rosin, will be addressing, 23 the bigger you get, the more you lose, depending 23 through his survey, some of the questions, Judge 24 on the rate structure. 24 Strickler, you posed about where would people go. 25 THE WITNESS: That's the first stream 25 I just wanted to indicate to you that his

3395 3397

1 testimony will address that. 1 Q And just -- can you tell us what you 2 BY MR. LARSON: 2 mean when you used the terms "lean-back" and 3 Q Just one follow-up question, if I could. 3 "lean-forward" in these paragraphs? 4 Users who hit the cap when you put them in place 4 A Yes, so, again, lean-back services are 5 could have chosen to subscribe to Pandora One, the 5 radio-like services, one where you hit play and 6 subscription product, then continue listening 6 the service kind of chooses for you. 7 beyond the cap; is that right? 7 Q And lean-forward? 8 A That's correct. 8 A Whereas in lean-forward we consider 9 Q Did you notice any significant uptick in 9 on-demand services. So you go into the service 10 Pandora One subscriptions after the cap was put in 10 and you choose exactly what you want to listen to. 11 place? 11 Q What claims by SoundExchange were you 12 A Listeners, kind of, had two options. 12 responding to in this section? 13 They could pay 99 cents and continue to listen 13 A Well, the claim that those are 14 with ads, or they could pay $3.99 at the time and 14 converging in some sense or there's less and less 15 subscribe, for an ad-free product, to continue 15 of a difference or consumer preference between 16 listening. 16 those two. Our research and our experience has 17 We had several hundred thousand 17 been that listeners have a very clear distinction 18 listeners a month opt for the 99 cents and then 18 in use case of music between those two. The 19 another -- well, the total net increase in 19 80 percent of the -- our survey showed that 20 subscribers in this time period was about 20 80 percent of Americans listen to music -- or 21 1.8 million subscribers. I think a majority of 21 80 percent of listening is done through lean-back 22 those subscribers are those incremental 22 sources. About 20 percent is lean-in. Now, a 23 subscribers who actually subscribed in order to 23 same user might do both. There's like, for 24 continue listening. 24 example, significant overlap between Spotify 25 JUDGE STRICKLER: How many subscribers 25 users, which is a lean-in or on-demand service,

3396 3398

1 did you have prior to putting on the cap? You've 1 and Pandora users, which is a lean-back radio 2 seen growth of 1.8 million, and you were making 2 service. 3 the assumption that most of them did it in order 3 There's significant overlap between 4 to avoid the cap. So you would -- 4 those two user bases because they're such 5 THE WITNESS: I believe it went from 5 different actual use cases. 6 about 1.4 million to about 3 million, you know, or 6 Q Doesn't Pandora state publicly that 7 3.2 million. 7 on-demand services are among its competitors? 8 JUDGE STRICKLER: In one quarter? 8 A We do. 9 THE WITNESS: In two quarters. 9 Q So does that mean that the suggestion 10 JUDGE STRICKLER: In two quarters. So 10 from SoundExchange that you discuss in paragraph 9 11 over 100 percent increase in two quarters. 11 is correct, in your view? 12 THE WITNESS: That's my recollection. 12 A That it's -- I'm sorry? 13 BY MR. LARSON: 13 Q That the services compete with one 14 Q Let's shift gears to another topic 14 another. 15 unless there are any more questions on this one. 15 A Well, they do compete with one another 16 If you could look at paragraphs 9 to 11 16 for music listening generally, and probably at 17 of your written testimony. It's in front of you 17 that margin, that line between 80 percent and 18 in your binder. 18 20 percent, there's some overlap there, but, you 19 A My written? 19 know, in terms of the majority of listening for 20 Q Yes, your written rebuttal testimony. 20 either service, it's not competing directly. 21 It should be the first tab. 21 They're not substitutes for each other. 22 Now, those paragraphs discuss something 22 We talk a lot strategically about should 23 you identify as lean-back versus lean-forward 23 we develop and launch an on-demand product to 24 listening; is that right? 24 complement our radio product. And Spotify has 25 A That's correct. 25 taken similar actions to try and figure out how to

3399 3401

1 tap the 80 percent of listening through their sort 1 why should we then double-dip and pay another 2 of playlist listening and -- so there's 2 royalty on top of that revenue that Pandora 3 increasingly marginal, you know, ways we're trying 3 collects? It's 1 percent of the 99 cents. That's 4 to compete in each other's adjacent markets. I 4 the affiliate fee that comes back to us. But 5 think that's a natural evolution maybe over time. 5 should that revenue also be sort of double counted 6 But our core businesses are very different. 6 in the definition? We don't think that's fair. 7 Q Just a couple more questions, and we'll 7 MR. LARSON: I have no more questions. 8 be finished. 8 Thank you, Mr. Herring. 9 Are you familiar with SoundExchange's 9 JUDGE STRICKLER: I have a question 10 proposed revenue definition in this proceeding? 10 about another subject in your written rebuttal 11 A Yes. 11 testimony -- 12 Q At a general level, what is your 12 THE WITNESS: Okay. 13 criticism of that definition as you spell it out 13 JUDGE STRICKLER: -- that was just 14 in your written rebuttal testimony? 14 uncovered just now about the definition of a 15 A Well, in general, it's just broad and 15 "qualified auditor." 16 incorporates all revenue we could possibly be 16 THE WITNESS: Yes. 17 generating. It's way -- and, essentially, revenue 17 JUDGE STRICKLER: Your testimony is that 18 outside revenue generated by serving the 18 the regulation should continue to require a 19 compensable plays or the music associated with the 19 certified public accountant to be the individual 20 license. That just doesn't, you know, inherently 20 who conducts the audit, correct? 21 make sense to me. If I have -- any time I use 21 THE WITNESS: Correct. 22 that music and stream that music and, sort of, 22 JUDGE STRICKLER: As opposed to 23 advertising or charge a subscription for it, that 23 SoundExchange's position that somebody who is 24 makes sense. I should be paying a license back to 24 qualified by experience and education, even if 25 that music. If I sell a Pandora mug or a T-shirt 25 they are not a CPA, should be permitted to conduct

3400 3402

1 on our site, I don't understand why they should be 1 the audits of the licensing services, correct? 2 paid for those ancillary revenue streams or 2 THE WITNESS: Yes. 3 develop other business lines that are not related 3 JUDGE STRICKLER: Are you -- you're a 4 with actually playing the music. 4 certified public accountant; is that correct? 5 CHIEF JUDGE BARNETT: What percentage of 5 THE WITNESS: I am. 6 your revenue is mugs and T-shirts? 6 JUDGE STRICKLER: Are you aware of 7 THE WITNESS: Today, a very small 7 whether there exists in the market of certified 8 percentage of revenue is anything but music 8 public accountants individuals who have both -- 9 streaming. But we are looking into the future. 9 who are both CPAs and have education and/or 10 And it goes way beyond mugs and T-shirts. 10 experience in this particular marketplace? 11 So, theoretically -- and I'm not a 11 THE WITNESS: Yes. We use auditors out 12 lawyer, so, once again, I'll start out by saying 12 of KPMG, for example, to augment our ability to -- 13 I'm not an expert in the testimony I'm about to 13 our own accounting firm, Ernst & Young -- our own 14 give, but we have a business -- a B-to-B service 14 accounting firm is Ernst & Young. We use auditors 15 where we pay licensing under a different statutory 15 out of KPMG who are experts in this to help us 16 rate related to that business. Pandora gets 16 ensure that we're reporting correctly under 17 played in 14,000 establishments in the United 17 SoundExchange. 18 States. 18 JUDGE STRICKLER: That's by way of 19 Theoretically, under this definition, 19 example, and they were already doing work for 20 you know, that revenue could get pulled under -- 20 you -- 21 under this definition. We get affiliate fees for 21 THE WITNESS: Yes. 22 selling tens and millions of downloads through 22 JUDGE STRICKLER: -- so, theoretically, 23 Pandora. Now, we literally do it more as a 23 they might have a conflict of interest, and that 24 feature for our customers, but significant 24 wouldn't be an example of one that could be used. 25 royalties are paid through the download sales. So 25 THE WITNESS: Right.

3403 3405

1 JUDGE STRICKLER: My question was more 1 CHIEF JUDGE BARNETT: Good morning, 2 general, but thank you for the example. 2 Mr. Choudhury. 3 THE WITNESS: Yes. 3 CROSS-EXAMINATION BY COUNSEL FOR 4 JUDGE STRICKLER: Is there a universe, SOUNDEXCHANGE 5 however small it may be, as far as you know, of 4 BY MR. CHOUDHURY: 6 accountants or auditors who are both certified 5 Q Good morning, Mr. Herring. 7 public accountants and have the education or 6 A Good morning. 8 experience in this industry? 7 Q We met at your deposition. You know I 9 THE WITNESS: Yes, I don't have personal 8 represent SoundExchange. It's good to see you 10 knowledge of those people outside of the ones I 9 again. 11 work with. And that would be a little bit of a 10 I want to begin with a couple of the 12 unicorn, I would believe. What -- the reason that 11 questions that the judges have asked this morning. 13 I'm focused on in my testimony is that certified 12 Judge Barnett, of course, asked you about 14 public accountants can bring in -- are allowed 13 executive compensation. I just want to make sure 15 through their practice to bring in experts to 14 the record is clear. When you were discussing, I 16 whatever level they need in order to get to the 15 think it was the 3 percent, were you including 17 right answer. But they bring to it a level of 16 stock-based compensation? 18 professionalism, a code of conduct, a fairness, a 17 A No. 19 duty to get the right answer rather than the 18 Q How much did Pandora pay in stock-based 20 answer my client wants, all things that a 19 compensation to its executives last year? 21 qualified auditor that is a nebulous definition 20 A I don't know that number off the top of 22 makes me concerned. 21 my head. I'm sure you can remind me what it is. 23 JUDGE STRICKLER: So if I understand 22 Q So let's look at volume 1 of your 24 your testimony correctly, then, it's your 23 binder. If you could take a look at tab 7 in your 25 understanding that certified public accountants, 24 binder. This is Pandora's 10-K, right? 25 A It is.

3404 3406

1 when they're auditing a business in an industry 1 Q You're familiar with this? 2 that they might not have particular familiarity 2 A I am. 3 with will call in someone else, also an 3 MR. CHOUDHURY: Your Honor, at this time 4 accountant, to advise them with regard to the 4 we would move into evidence SoundExchange 158. 5 particulars relating to the industry? 5 MR. LARSON: No objection, Your Honor. 6 THE WITNESS: Yes, that's common across 6 CHIEF JUDGE BARNETT: Exhibit 158 is 7 many industries. They'll bring in technical 7 admitted. 8 experts to help them make sure that the audit is 8 (SoundExchange Exhibit 158 was admitted 9 correct, but the overall audit is still governed 9 into evidence.) 10 by their expertise as auditors and through their 10 JUDGE STRICKLER: Which 10-K is this? 11 code of conduct and their training as CPAs. 11 MR. CHOUDHURY: This is the most recent 12 JUDGE STRICKLER: Do the generally 12 10-K. 13 accepted auditing procedures that govern auditors 13 BY MR. CHOUDHURY: 14 specifically address the ability or the 14 Q Do you report the stock compensation 15 appropriateness of a CPA to bring in such an 15 that Pandora pays out in your 10-Ks? 16 individual when conducting an audit? 16 A Yes. 17 THE WITNESS: I believe they do, yes. 17 Q Do you know where you do? Can you point 18 JUDGE STRICKLER: Thank you. 18 us to it? 19 MR. LARSON: No more questions for me, 19 A It's included in the financial 20 Your Honor. Thanks. 20 statements because the financial statements are 21 MR. CHOUDHURY: I have some binders. 21 under GAAP. So, as I mentioned when I was 22 Volume 1. A lot of it is depositions. Here are 22 discussing EBITDA earlier, that excluded -- our 23 the demonstratives. 23 earnings of 6 percent actually excludes 24 MR. POMERANTZ: Judge Barnett, let me 24 stock-based compensation. We're at a significant 25 hand this over to you. 25 loss percentage if you included that stock comp.

3407 3409

1 Q So if I could ask you to look at 1 listener caps, right? 2 SoundExchange 158 -- page 158094. 2 A Right. 3 A Okay. 3 Q And you recall those questions. 4 Q Does that refresh your recollection 4 You reference some research that you had 5 about how much stock-based compensation Pandora 5 seen. I think you said, I want to make sure we 6 paid in the 12 months ending December 31st, 2014? 6 have this right, that you had seen some research 7 A Sure. 7 that would be single digits of users who might 8 Q How much was it? 8 migrate. Do you recall what we're talking about? 9 A 87 million. 9 A Yes. 10 Q And does Pandora have an estimate of how 10 Q Was that research on migration of -- 11 much stock-based compensation it anticipates 11 during a listener cap? 12 paying out this year? 12 A No, that research -- I think -- I 13 A It will probably be about 120 million. 13 believe that research asked the question if 14 Q What was SoundExchange -- what was 14 Pandora was not available, you know, what 15 Pandora's total royalty payments to SoundExchange 15 alternate service would you use. 16 in 2014? 16 Q And I want to make sure we understand. 17 A About 400 million. 17 So a listener cap, you said about 4 to 5 percent 18 JUDGE STRICKLER: These stock options, 18 of Pandora users were affected by this listener 19 these are payouts or the value of the stock? 19 cap? 20 THE WITNESS: This is how much the 20 A That's my recollection, yes. 21 expense was incurred in the period, which is 21 Q These were your most -- I think you said 22 neither of those things. 22 most loyal customers? 23 JUDGE STRICKLER: So it wasn't an 23 A Correct. 24 exercise of the option. 24 Q That means these are the heaviest users 25 THE WITNESS: No. 25 of music?

3408 3410

1 JUDGE STRICKLER: It was the evaluation 1 A Correct. 2 of the option? 2 Q Now, the research you were citing, did 3 THE WITNESS: It's a Black-Scholes 3 that only look at the heaviest users of music? 4 valuation that's been amortized over the value. 4 A No, we looked at, I believe, all users 5 So if it's an option, for example, it has a huge 5 of Pandora. 6 value, even though the strike price might be at or 6 Q And -- you said that the subscription 7 above the current price so it has zero value to 7 price, I believe you said in response to Judge 8 the actual employee. 8 Strickler's question, it almost doubled over two 9 And then it's amortized over the period 9 quarters; is that right? 10 of the -- of the period -- usually, of four years, 10 A Not the price. 11 an initial period that's set when the option was 11 Q I mean the subscription -- the number of 12 granted. 12 subscribers you had. 13 JUDGE STRICKLER: So is there a positive 13 A Correct. 14 value here if the strike price in a given year is 14 Q And Pandora at that time paid a 15 above the option price? 15 different rate for subscriptions than -- for 16 THE WITNESS: It's still a positive 16 royalty costs for subscriptions than 17 value, yes. 17 non-subscriptions, correct? 18 JUDGE STRICKLER: Does it affect -- does 18 A Correct. 19 it affect cash flows for the year? 19 Q That subscription royalty rate that 20 THE WITNESS: It does not affect cash 20 Pandora paid at the time, that -- how does that 21 flows, no. 21 compare to SoundExchange's proposed rates now? 22 JUDGE STRICKLER: Thank you. 22 A It's about the same. It's important to 23 BY MR. CHOUDHURY: 23 say that, like, that isn't a good thing. Turning 24 Q Now, Judge Strickler also asked you a 24 our business -- 25 number of questions about what happened during the 25 Q Mr. Herring, that's not my question.

3411 3413

1 A Okay. 1 A Correct. 2 Q Your counsel will have time to redirect. 2 Q Now, with respect to Pandora's current 3 A All right. 3 rates and terms, Pandora is currently subject to a 4 Q So let me ask you, you don't actually 4 percentage of revenue as part of a "greater of" 5 know how many users went back to broadcast radio 5 structure, right? 6 during the listener caps? 6 A That's correct. 7 A No. 7 Q That includes a revenue-sharing prong, 8 Q And how many of those users went to 8 right? That includes a percentage of revenue? 9 interactive services? 9 A It does. 10 A No. 10 Q And that includes -- that percentage of 11 Q Or noninteractive services? 11 revenue is done on an entity-wide basis, right? 12 A No. 12 A Yes. 13 Q Or pirate services, right? 13 Q So it includes all of Pandora's revenues 14 A (Shaking head.) 14 operating in the U.S. 15 Q All you know is that your subscriptions 15 A Correct. 16 doubled over those two quarters; is that right? 16 Q Mugs, T-shirts, affiliate fees, 17 A Yes. 17 everything, right? 18 Q Let's talk about your testimony about 18 A Yes. 19 terms, if we could. You recall that we discussed 19 Q I just want to clarify, you know, when 20 some of this at your deposition. 20 Judge Barnett asked you about how large these mugs 21 A Yes. 21 and T-shirts are and you said very small, does 22 Q Now, Mr. Herring, you don't know how 22 very small mean below 10 percent? 23 long it takes for Pandora to generate its 23 A Yes. 24 month-end report of play counts for SoundExchange, 24 Q Below 5 percent? 25 right? 25 A Yes.

3412 3414

1 A I know approximately how long it takes. 1 Q Below 1 percent? 2 Q Well, but you, at your deposition, you 2 A Yes. 3 recall me asking you, right? 3 Q And Pandora has a target model, right, 4 A Yes. 4 that it hopes to reach in terms of margins and 5 Q Do you recall your answer was "I don't 5 costs, right? 6 know how long"? 6 A Yes. 7 A I don't know exactly how long it takes, 7 Q In that target model you say you could 8 that's correct. 8 pay 40 percent of your revenue to content costs on 9 Q And you don't know whether the query is 9 an entity-wide basis, right? 10 a matter of days or weeks, right? 10 A Yes. 11 A No, that's correct. 11 Q And that would include the sound 12 Q And you have an estimate of 12 recording royalties, right? 13 SoundExchange royalties within a couple of weeks 13 A Yes. 14 of the end of each month, right? 14 Q And it would also include the publishing 15 A Yes. 15 royalties, right? 16 Q And at the end of each quarter -- of 16 A Yes. 17 each quarter, you need to, in fact, calculate your 17 Q And so if publishing -- let me ask you, 18 royalty payments to have your earnings calls 18 and I don't know if this is restricted or not, so 19 within three and a half weeks of the end of that 19 I'll hold off until we get to restricted session. 20 quarter, right? 20 In 2013, you paid 53 percent of your 21 A That's correct. 21 entity-wide revenue to content costs, right? 22 Q And you don't know how your current 22 A Yes. 23 process would change if you had to calculate your 23 Q In 2012, you paid 60 percent of your 24 SoundExchange royalties on a 30-day window instead 24 entity-wide revenue to content costs? 25 of a 45-day window; is that correct? 25 A Yes.

3415 3417

1 Q Now, if we look at your written direct 1 requirements on people who are purely CPAs 2 testimony, if you can look -- this is -- we can 2 versus -- who are operating as a CPA and signing 3 use -- it's in tab 3 of the binder you have in 3 opinions. 4 front of you, which is the same document as has 4 Q With respect to your testimony about 5 been admitted into evidence. 5 unclaimed funds, what experience do you have, what 6 A Yes. 6 personal knowledge do you have about the unclaimed 7 Q Let's just keep one big binder in front 7 funds in pooled royalties? 8 of you. 8 A I don't have personal experience related 9 If you could look at paragraph 55, 9 to that. 10 page 25. 10 Q With respect to your license with 11 Now, you cite a number of regulations, 11 Merlin, that is administered in part by 12 correct, in that paragraph? 12 SoundExchange, correct? 13 A Yes. 13 A Correct. 14 It's actually 54. 14 Q That was a decision made by Pandora and 15 Q 54, and then going on to paragraph 55. 15 Merlin, correct? 16 A You're talking about into 55, okay. 16 A Yes. 17 Q Those aren't regulations that apply to 17 Q You also believe that other commercial 18 Pandora, right? 18 webcasters should pay a royalty rate at least as 19 A That's correct. 19 high as Pandora, correct? 20 Q And we've established you're not a 20 A I think that commercial webcasters 21 lawyer and you're not a regulator, right? 21 should pay an equal -- we should all pay the same 22 A That's correct. 22 rate. 23 Q Pandora is a pureplay webcaster, right? 23 Q And that's because you think that it's 24 A It is. 24 better for the webcaster marketplace if costs are 25 Q Which is to say that your main product 25 equal across the landscape, correct?

3416 3418

1 you offer is music, right? 1 A I believe that if costs are fair and 2 A That's correct. 2 equal across the landscape, it's better for 3 Q When you refer to the Satellite I and 3 competition. 4 Satellite II proceedings in your testimony, did 4 Q Just turning back to that listener cap 5 you study the factual record in those proceedings? 5 situation we were talking about -- 6 A No. 6 A Yes. 7 Q Now, with respect to auditors, Judge 7 Q -- were costs equal at that time across 8 Strickler asked you some questions this morning. 8 the webcaster landscape? 9 Let me ask you, if the entity that was conducting 9 A No. 10 the audit had both a CPA and someone with 10 Q Did Pandora pay a higher or lower 11 technical experience, would that satisfy your 11 per-play rate than the commercial webcasters you 12 concerns about making sure that there was the CPA 12 were discussing with Judge Strickler? 13 obligations that you discussed? 13 A We paid the pureplay rates, which are 14 A Well, I would like it to be a CPA firm 14 lower than the Web III rates. 15 that was involved, someone who would sign an 15 Q Now, if we can turn to paragraph 4 of 16 auditor report under the -- under the requirements 16 your written testimony. 17 of a certified public accountant. 17 Now, you agree that the rate that 18 Q And are you drawing a distinction 18 Pandora is theoretically capable of paying is 19 between a firm that has CPAs, in fact, let's say a 19 simply not informative to the judges, right, 20 firm that has CPAs who are working on the audit 20 that's what you testified to here? 21 versus a CPA firm? 21 A Yes. 22 I just want to make sure the record is 22 Q That would hold true of other webcasters 23 clear here. 23 as well, right? 24 A Yes, I am making somewhat of a 24 A I would assume so. 25 distinction there. There are different 25 Q You agree that Pandora is not

3419 3421

1 necessarily attempting to maximize the 1 outside of the rate-setting period, these judges 2 profitability in its current quarter, right? 2 would not know by the end of the rate period 3 A That's correct. 3 whether you attained that profitability, right? 4 Q And, in fact, you're focused on what you 4 A They would know whether we attained it 5 call future profits, right? 5 in three to five years, what level of 6 A Yes. 6 profitability was attained then. What we're going 7 Q In fact, yesterday I think you said you 7 to attain farther out is, frankly, unknowable. 8 might be focusing on profitability in 2017, 2020, 8 The point is that we're investing to build a 9 2025, 2030, right? 9 business over time. 10 A Sure. 10 Q Let me ask you, you agree that it's not 11 Q And this, of course, is to set a rate 11 the judges' responsibility to ensure Pandora's 12 proceeding for the next five years, right? So if 12 survival, right? 13 your focus is beyond those five years, these 13 A Yes. 14 judges can't know if you're going to actually 14 Q And you agree it's not also the record 15 achieve the profitability that you're discussing 15 companies' responsibility, right? 16 within that period of time, right? 16 A Yes. 17 A Well -- 17 Q And you agree that it's not the judges' 18 Q Let me ask it -- 18 responsibility to ensure Pandora's profitability, 19 A I don't understand that question. 19 right? 20 JUDGE STRICKLER: Are you withdrawing 20 A Yes. 21 that question? 21 Q It's not the record companies' 22 MR. CHOUDHURY: Yes, I'm withdrawing it. 22 responsibility either? 23 JUDGE STRICKLER: You asked him a 23 A Correct. 24 question, and he hadn't answered it, and then you 24 Q You agree that the way that Pandora 25 went on to another question. 25 chooses to monetize its business, that's not what

3420 3422

1 MR. CHOUDHURY: Let me withdraw the 1 the judges are here to determine, right? 2 question. 2 A Correct. 3 BY MR. CHOUDHURY: 3 Q And that's also not the record 4 Q The future profits you're discussing, 4 companies' responsibility for Pandora's business 5 are those future profits necessarily within the 5 choices, right? 6 five-year rate period? 6 A Correct. 7 A Well, only the rate -- only the profits 7 Q You agree that Pandora's goal in this 8 that we would operate in the next five years would 8 business is to maximize its entity value or this 9 be within the five-year rate period. 9 long-term profitability we've been discussing, 10 Q Right. So these judges can't know if 10 right? 11 you're going to attain future profits beyond five 11 A Yes. 12 years at the time that -- during the rate period 12 Q And that's not -- and that's the 13 rate set here, right? 13 business strategy you've chosen; is that right? 14 MR. LARSON: Objection to the form. The 14 A That is part of our overall business 15 question is terribly confusing. 15 strategy. It's broader than that, but that's part 16 CHIEF JUDGE BARNETT: Sustained. Can 16 of it. 17 you rephrase, Mr. Choudhury? 17 Q And that's the strategy that informs 18 MR. CHOUDHURY: Sure. 18 your business decisions, right? 19 BY MR. CHOUDHURY: 19 A You know, our purpose doesn't say to 20 Q When you refer to future profits, you're 20 build a profitable business over the long term. 21 referring to profits that may occur beyond the 21 That's one aspect of our goal as a company. It's 22 rate-setting period of this proceeding, right? 22 important that we build a -- revenue is our 23 A Yes. Well, and also during the 23 oxygen, is one of our principles in a business, 24 rate-setting period. 24 but it's one of 13 principles. 25 Q And to the extent those profits occur 25 So I think, yes, it's a big part of what

3423 3425

1 we are operating as, as a company, is to drive 1 paid out, was it, in fact, paid out? 2 revenue and be profitable, but it's not the 2 THE WITNESS: It's expensed on a GAAP 3 core -- why we go to work every day. 3 basis. There's no -- no dollars are actually paid 4 JUDGE STRICKLER: Did you say it's only 4 out associated with that expense. 5 one of your 13 principles? 5 JUDGE STRICKLER: Just so it's clear. 6 THE WITNESS: Yes. 6 BY MR. CHOUDHURY: 7 JUDGE STRICKLER: Sort of a Woodrow 7 Q Pandora's decision to not play more ads 8 Wilson type of comment. 8 today, that reduced profits today, right, like 9 Do we have in the record the 13 9 Pandora's decision -- 10 principles? 10 A I don't believe so. 11 THE WITNESS: No, I don't believe we do. 11 Q -- not to increase its ad load further 12 MR. LARSON: I don't believe we do, Your 12 than it currently exists, right, that affects the 13 Honor. 13 amount of profits it has in the current year? 14 JUDGE STRICKLER: Okay, thank you. Your 14 A I disagree with that. No, that's not 15 ball, counsel. 15 correct. We believe we are serving every ad as 16 BY MR. CHOUDHURY: 16 efficiently as we can today to maximize revenue 17 Q You are making decisions that 17 and, therefore, maximize profits. 18 wouldn't -- you're not -- you are making decisions 18 MR. CHOUDHURY: I think at this point, 19 that do not increase Pandora's profits today, 19 Your Honor, I think we need to go into a 20 right? 20 restricted session. 21 A Well, when we make investment decisions 21 CHIEF JUDGE BARNETT: If there's anyone 22 for the future, we are, by definition, spending 22 in the courtroom who has not signed the 23 dollars that we otherwise could drive to our 23 nondisclosure certificate, please wait outside the 24 profit margins today. 24 hearing room. 25 Q So, for example, one of those -- so last 25 MR. RICH: Your Honor, just to be clear,

3424 3426

1 year Pandora posted, I think, a loss of 1 we have certain internal personnel from Pandora 2 approximately 30 million dollars; is that correct? 2 who I assume can stay for this restricted session? 3 A Yes. 3 CHIEF JUDGE BARNETT: Since this is an 4 Q And last year Pandora spent -- 4 examination of Pandora, yes. 5 A On a GAAP basis. 5 MR. RICH: Thank you. 6 Q That's right. 6 (THIS ENDS PUBLIC SESSION) 7 And Pandora spends about 50 million 7 (RESTRICTED SESSION BOUND SEPARATELY) 8 dollars in product development, right? 8 9 A Yes. 9 10 Q That's a decision you made that affected 10 11 profits today, right? 11 12 A Yes. 12 13 Q It reduced them. 13 14 A Correct. 14 15 Q And Pandora spent 277 million dollars 15 16 last year on sales and marketing, right? 16 17 A Yes. 17 18 Q That's a decision that decreased profits 18 19 today, right? 19 20 A Correct. 20 21 Q And Pandora's decision to pay out 21 22 87 million dollars in stock-based compensation, 22 23 that affected profits today, right? 23 24 A On a GAAP basis, yes. 24 25 JUDGE STRICKLER: Again, when you say 25

3427 3429

1 (THIS BEGINS RESTRICTED SESSION) 1 (THIS PAGE CONTAINS RESTRICTED 2 MATERIAL) 3 2 4 3 5 4 6 5 7 6 8 7 9 8 10 9 11 10 12 11 13 12 14 13 15 14 16 15 17 16 18 17 19 18 20 19 21 20 22 21 23 22 24 23 25 24 25

3428 3430

1 (THIS PAGE CONTAINS RESTRICTED 1 (THIS BEGINS RESTRICTED SESSION) MATERIAL) 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 10 10 11 11 12 12 13 13 14 14 15 15 16 16 17 17 18 18 19 19 20 20 21 21 22 22 23 23 24 24 25 25

3431 3433

1 (THIS PAGE CONTAINS RESTRICTED 1 (THIS PAGE CONTAINS RESTRICTED MATERIAL) MATERIAL) 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 10 10 11 11 12 12 13 13 14 14 15 15 16 16 17 17 18 18 19 19 20 20 21 21 22 22 23 23 24 24 25 25

3432 3434

3435 3437

3436 3438

3439 3441

3440 3442

3443 3445

3444 3446

3447 3449

3448 3450

3451 3453

3452 3454

3455 3457

3456 3458

3459 3461

3460 3462

3463 3465

3464 3466

3467 3469

3468 3470

3471 3473

3472 3474

3475 3477

3476 3478

3479 3481

3480 3482

3483 3485

3484 3486

3487 3489

3488 3490

3491 3493

3492 3494

3495 3497

3496 3498

3499 3501

3500 3502

3503 3505

3504 3506

3507 3509

3508 3510

3511 3513

3512 3514

3515 3517

3516 3518

3519 3521

3520 3522

3523 3525

3524 3526

3527 3529

3528 3530

3531 3533

3532 3534

3535 3537

1 (THIS PAGE CONTAINS RESTRICTED 1 (THIS PAGE CONTAINS RESTRICTED MATERIAL) MATERIAL) 2 2 3 3 4 4 5 5 6 6 7 7 (THIS ENDS RESTRICTED SESSION) 8 8 9 9 10 10 11 11 12 12 13 13 14 14 15 15 16 16 17 17 18 18 19 19 20 20 21 21 22 22 23 23 24 24 25 25

3536 3538

3539 3541

3540 3542

3543 3545

3544 3546

3547 3549

3548 3550

3551 3553

3552 3554

3555 3557

3556 3558

3559 3561

1 (THIS PAGE CONTAINS RESTRICTED 1 (THIS BEGINS PUBLIC SESSION) MATERIAL) 2 MR. HANSEN: Your Honors, if I could 2 3 introduce Tres Williams, in-house counsel for 3 4 iHeartMedia. I believe he can stay for this 4 5 proceeding because any confidential information 5 6 would be iHeartMedia information. 6 7 Ms. Pope will be conducting examination 7 8 of our next witness. 8 9 CHIEF JUDGE BARNETT: Thank you. 9 10 - - - 10 11 MARISSA MORRIS, 11 12 having been duly sworn, testified as follows: 12 13 DIRECT EXAMINATION BY COUNSEL FOR 13 IHEARTMEDIA 14 14 BY MS. POPE: 15 15 Q Good afternoon, Ms. Morris. 16 16 A Hi, good afternoon. 17 17 Q How long have you been working in radio? 18 18 A I have been working in radio for about 19 19 11 years now. I had started out as an intern for 20 20 iHeart -- Clear Channel, which is now iHeartMedia, 21 21 when I was in high school at 16, and then I 22 22 continued throughout college interning for the New 23 23 York stations. 24 24 I worked briefly in promotions in D.C., 25 25 at some of the D.C. stations as well. Then I

3560 3562

1 (THIS PAGE CONTAINS RESTRICTED 1 graduated and took a position with the New York MATERIAL) 2 programming team at Clear Channel and have been 2 3 gradually growing there and been with iHeartMedia 3 4 ever since graduation. 4 5 CHIEF JUDGE BARNETT: Could you -- can 5 6 we back up and would you state your name and spell 6 7 your -- first and last names, please. 7 8 THE WITNESS: Sure. My name is Marissa 8 9 Morris, M-A-R-I-S-S-A, Morris, M-O-R-R-I-S. 9 10 CHIEF JUDGE BARNETT: Thank you. 10 11 BY MS. POPE: 11 12 Q Ms. Morris, what's your current position 12 13 at iHeartMedia? 13 14 A I am the vice president of artist 14 15 relations. 15 16 Q What are your current job 16 17 responsibilities? 17 (THIS ENDS RESTRICTED SESSION) 18 A Day-to-day I work closely with labels 18 19 and managers on building national promotions and 19 20 marketing plans around releases and I book talent 20 21 for iHeartRadio's national events. 21 22 Q Do you have any responsibilities related 22 23 to iHeartMedia's music summits? 23 24 A I do. I also organize our semiannual 24 25 music summits. 25

4184 Before the

UNITED STATES COPYRIGHT ROYALTY BOARD

Library of Congress

Washington, D.C.

------: In Re: : Docket No. : 15-CRB-0001-WR Determination of Royalty : (2016-2020) Rates and Terms for : Volume 16-PUBLIC Ephemeral Recording and : Pages 4184-4226 Digital Performance of : Pages 4324-4410 Sound Recordings (Web IV) : Pages 4468-4509 ------:

PUBLIC SESSION

Washington, D.C.

Monday, May 18, 2015

The hearing in the above-entitled matter was convened at 9:12 a.m.

BEFORE COPYRIGHT ROYALTY JUDGES:

SUZANNE M. BARNETT, CHIEF JUDGE

DAVID R. STRICKLER, JUDGE

JESSE FEDER, JUDGE

4193 4195

1 there would be to any witness appearing here today. 1 last page that that's your signature on the 2 I can't speculate to what I think. It's highly 2 exhibit? 3 unlikely and tangential. 3 A. It is. 4 Needless to say, if upon the public 4 MR. LARSON: Your Honor, we'd offer 5 release of that document Mr. Pomerantz feels there 5 Pandora Exhibit 5007 into evidence. 6 is some basis for having some further colloquy 6 MR. KLAUS: No objection, Your Honor. 7 about it as opposed to, perhaps, briefing its 7 CHIEF JUDGE BARNETT: 5007 is admitted. 8 consequences in the post trial briefing, we would 8 (Pandora Exhibit No. 5007 was admitted 9 be happy to discuss it at that time. 9 into evidence.) 10 MR. POMERANTZ: That's fine, Your 10 MR. LARSON: And we would also offer 11 Honor. 11 Pandora Exhibits 5008 through 5015, which are also 12 CHIEF JUDGE BARNETT: Thank you. 12 in the binder, in evidence. 13 I do think, Mr. Pomerantz, it may be a 13 CHIEF JUDGE BARNETT: 5008 through 14 bit premature. I mean, even though we are not all 14 what? 15 powerful, please be reminded that we do have the 15 MR. LARSON: 5015. 16 ability to call witnesses on our own if we feel 16 Yeah, 08 through 15. 17 it's necessary or -- and appropriate. It's not 17 CHIEF JUDGE BARNETT: 15, okay. Thank 18 something we're planning to do right now, either, 18 you. 19 because we're all operating in a vacuum. 19 MR. KLAUS: And, I believe, Your Honor, 20 So, if necessary, you can raise the 20 these were the exhibits to Mr. Herring's direct 21 issue again later and we can deal with it in 21 testimony. We do object on hearsay grounds to the 22 that -- at that time in some appropriate way. 22 newspaper article that's attached at 5011 and the 23 Good morning, Mr. Larson. 23 newspaper article that's attached at 5013. No 24 MR. LARSON: Good morning. 24 objection to the documents coming in as evidence of 25 CHIEF JUDGE BARNETT: Mr. Herring, you 25 what was stated in a newspaper article, but we do

4194 4196

1 remain under oath. 1 object to consideration of any of the statements 2 MR. LARSON: Yes. Thank you. 2 therein for their truth. 3 May I hand out some binders here? 3 MR. LARSON: Your Honor, we disclosed 4 CHIEF JUDGE BARNETT: Oh, good. 4 these exhibits to SoundExchange on their objections 5 MR. LARSON: Wouldn't be a day at CRB 5 to the exhibit list that they sent to us. They did 6 without binders. 6 not include a hearsay objection. So we believe 7 There are demonstratives inside. 7 that objection is moot. 8 Tucked inside of the binders are a small set of 8 CHIEF JUDGE BARNETT: The objection is 9 demonstratives, as well. 9 overruled. We will give whatever weight they are 10 MICHAEL HERRING, 10 due. 11 having been previously duly sworn, to tell the 11 MR. LARSON: Thank you, Your Honor. 12 truth, the whole truth, and nothing but the truth, 12 BY MR. LARSON: 13 testified as follows: 13 Q. Mr. Herring, could you take a look at 14 DIRECT EXAMINATION BY COUNSEL FOR 14 Slide 1 on the demonstratives in front of you? PANDORA 15 A. Yes. 15 BY MR. LARSON: 16 CHIEF JUDGE BARNETT: I'm sorry. 16 Q. Welcome back, Mr. Herring. 17 Before you do that -- thank you, Judge Feder. 5008 17 A. Thank you. 18 through 5015 inclusive are admitted. 18 Q. We're here today to discuss your 19 (Pandora Exhibit Nos. 5008 and 5015 19 written direct testimony; is that correct? 20 were admitted into evidence.) 20 A. That's correct. 21 BY MR. LARSON: 21 Q. Okay. And that appears in the binder 22 Q. Mr. Herring, do you recognize Slide 1 22 in front of you, is the tab identified as Pandora 23 as the topics covered in your written direct 23 Exhibit 5007; is that right? 24 testimony? 24 A. Yes. 25 A. Yes. 25 Q. And can you just check for me on the

4197 4199

1 Q. Okay. And am I correct that you 1 JUDGE STRICKLER: The experiments that 2 testified to a number of these topics when you were 2 he's testifying to preceded the execution of the 3 here testifying about your written rebuttal 3 Merlin agreement? 4 testimony? 4 THE WITNESS: So I might have conflated 5 A. Yes. 5 two concepts here. So I was referring there to our 6 Q. Which topic on the slide will you be 6 ability to use our platform to connect artists with 7 focusing on today? 7 listeners. And, in that context, you know, create 8 A. The first bullet, the Pandora-Merlin 8 value and benefits for the labels and artists, you 9 agreement. 9 know, outside of just writing a royalty check. So 10 Q. Okay. Let's talk about that. 10 that's things like direct messaging and promoting 11 Did there come a time when Pandora 11 of tours and things that they can do on our 12 decided to explore the feasibility of direct 12 platform directly with listeners. 13 licensing with record labels as an alternative to 13 The experiments you're talking about, 14 securing rights to the statutory license? 14 one set of those was done in 2013. The second set 15 A. Well, yes. And at that time, actually, 15 was done in 2014, were ongoing when we actually 16 preceded my working at Pandora sometime in 2012, 16 signed -- the second set was ongoing, maybe the 17 but has been an initiative at Pandora since the -- 17 first section was done when we signed the deal with 18 you know, for my entire time at the company. 18 Merlin. 19 Q. And what was Pandora's rationale for 19 JUDGE STRICKLER: Was there any 20 considering direct licensing? 20 experiment undertaken before the Pandora-Merlin 21 A. Well, we believed that we would like to 21 agreement was executed? 22 have better partnerships with the music labels and 22 THE WITNESS: Yes. The steering 23 music publishers and we also believed that we could 23 experiments we did in 2013 were -- preceded the 24 do deals with the music labels specifically where 24 Merlin agreement. 25 we provided terms that might create lower prices 25 JUDGE STRICKLER: Preceded and were

4198 4200

1 for the, you know, lower royalty rates for the 1 completed before the execution of the Merlin 2 music that we played, and we were looking for a way 2 agreement? 3 to introduce price differentiation into our 3 THE WITNESS: Yes. Before we even 4 business model. 4 started negotiating. 5 Q. When you say, "price differentiation," 5 BY MR. LARSON: 6 is that the price competition that you mention in 6 Q. That leads directly into the next 7 Paragraph 23 of your testimony? 7 questions I have for you, Mr. Herring. 8 A. Yes. 8 First of all, can you just explain just 9 Q. Were there any other sort of 9 what were the steering experiments, as you referred 10 relationship benefits or relationships with artists 10 to them? 11 and record labels that were part of your direct 11 A. So there's really two sets of them. 12 licensing strategy? 12 One set was in 2013 where we instructed our 13 A. Yes. We have a pretty broad strategy 13 engineers and data scientists to -- to experiment 14 to build closer relationships with the music 14 with overspinning indie labels at various levels 15 industry with both labels and publishers. And we 15 and measure the effect on listening. So the number 16 have several initiatives that we wanted to 16 of times listeners returned, how long they 17 experiment with to roll out and test their 17 listened, and whether there was a negative effect 18 effectiveness, and we felt the best environment to 18 on retention rates of listeners over a period of 19 do that would be in a direct label -- in a direct 19 time. 20 relationship. 20 Q. And why did you run those experiments? 21 JUDGE STRICKLER: In that regards, you 21 A. Well, we wanted to see -- we believed 22 mentioned experiments. I want to understand the 22 that our platform is promotional. We wanted to see 23 time sequencing because we'll hear later from Mr. 23 if we could overspin above the natural rate indie 24 McBride with regard to experiments. 24 labels as a way of incentivizing lower payments for 25 THE WITNESS: Yes. 25 royalties.

4201 4203

1 Q. Okay. When you say, "overspin," what 1 first steering experiments, did Pandora begin 2 does it mean to steer towards a label or to 2 approaching record labels to attempt to negotiate 3 overspin them by, say, 40 percent? Can you just 3 direct licenses after that first set of steering 4 explain. 4 experiments? 5 A. Yeah. At a high level, it's just 5 A. Well, we've always been having 6 the -- there's a natural rate that we measure that 6 conversations with record labels, both majors and 7 a certain label's artists would spin on Pandora and 7 indies. But we got more serious about potential 8 that does vary somewhat, but it doesn't vary a lot. 8 terms, specifically with indie labels following 9 You know, outside of a new hit, a massive hit being 9 those steering experiments where we were confident 10 put into the system and being listened to a lot by 10 we could deliver on proposed terms. 11 listeners is just relatively steady. 11 Q. Okay. And how did you decide who to 12 What we do is we take that natural rate 12 approach? 13 and let's say for just illustrative purposes a 13 A. Well, you know, Pandora is about 35 14 label had represented one percent of our spins, if 14 percent of our spins generally are by indie labels 15 we were to overspin by 40 percent, we would, you 15 or not the three major labels, and so we have a 16 know, put extra weight on that label's songs in the 16 natural propensity to be attractive to the indie 17 algorithm such that they would spin 1.4 percent of 17 labels. Indie labels view Pandora positively and 18 our spins would be made up by that overspin label. 18 seemed more predisposed to do a direct deal and 19 JUDGE STRICKLER: You calculate an 19 to -- and to value the additional exposure on 20 actual rate based on the label's natural rate, not 20 Pandora. We just did -- although we had lots of 21 the art -- not an artist's natural rate? 21 conversations with the majors, that didn't seem -- 22 THE WITNESS: We do it both ways. We 22 you know didn't seem to be taking it seriously. 23 can do it both ways. 23 Their offers involved adding additional services 24 JUDGE STRICKLER: Which way did you do 24 like me to on-demand services to compete with 25 it in the experiments? 25 Spotify, et cetera, which we weren't -- was not on

4202 4204

1 THE WITNESS: We did it by label. 1 our strategic roadmap. 2 BY MR. LARSON: 2 Q. So, how did you decide to approach 3 Q. Okay. And what -- looking just at the 3 Merlin as opposed to other independent labels? 4 first set of steering experiments, what were the 4 A. Well, there's thousands of indie labels 5 results of that first set? 5 and one of the advantages Merlin gave to Pandora 6 A. The first set demonstrated that we 6 was the ability to negotiate with a single entity 7 could overspin indie labels and we specifically 7 and agree on a set of terms that then could be 8 targeted indie labels in the first set in 2013. 8 adopted by thousands of their member labels. So -- 9 CHIEF JUDGE BARNETT: Is any of this 9 and that was an opportunity to do a lot of deals 10 restricted? 10 with smaller labels all at once, rather than going 11 MR. LARSON: No. I believe we can stay 11 one by one through indie labels. 12 in public record. 12 Q. And just so the record is clear, can 13 CHIEF JUDGE BARNETT: Thank you. 13 you explain in a general level who Merlin is or 14 THE WITNESS: So I'll be careful. 14 what Merlin is? 15 You know, as much as 40 to 80 percent. 15 A. Yes. 16 So we -- there was a lot of leeway within to -- to 16 Merlin, generally, is a consortium that 17 spin indie labels much higher than their natural 17 negotiates or that prosecutes copyright violations 18 rate. 18 on behalf of their membership as well as negotiates 19 BY MR. LARSON: 19 licenses on be -- on, you know -- on behalf of 20 Q. And how did the second set of steering 20 their membership that have to be opted in by the 21 experiments differ from the first? 21 members. So they will negotiate a set of terms 22 A. The second set was set up to 22 with say YouTube or Pandora, and then their members 23 specifically measure our ability to over and under 23 can then, you know, voluntarily agree to operate 24 spin major labels by as much as 30 percent. 24 under those terms or not. 25 Q. Okay. So returning, again, to the 25 Q. Okay. And did Pandora make the first

4205 4207

1 sort of offer proposal in the Merlin negotiations? 1 opportunity. So as Pandora can drive its own 2 A. We did, yes. 2 monetization, that's how we've been able to drive 3 Q. And how did you arrive at the rate that 3 our gross profit up because we have, potentially, a 4 you initially proposed to Merlin? 4 fixed rate on a per song basis. 5 A. Well, with a little bit of art and 5 JUDGE STRICKLER: At the time that you 6 science. Our original proposal was a unitary rate 6 proposed the .0011 per-play fee, did you have an 7 of .0011, and that is a combination of, you know, 7 understanding of what you thought that would -- 8 experience and looking at our business model and 8 equilibrate to in terms of percentage of revenue 9 finding a balance between what we thought was a 9 for Pandora on a pro rata basis for the Merlin -- 10 rate that we could thrive as a business, we could 10 THE WITNESS: Yes. 11 still have the right investment levels based upon 11 JUDGE STRICKLER: -- labels that would 12 our monetization potential, and also pay a fair 12 sign on. I realize there's a lot of moving parts 13 royalty back to the copyright holders. 13 to it. 14 JUDGE STRICKLER: So the .0011 was 14 THE WITNESS: So you mean in a vacuum, 15 initially proposed by Pandora to Merlin? 15 or if there was, you know, a hundred percent of 16 THE WITNESS: Yes. 16 our -- 17 BY MR. LARSON: 17 JUDGE STRICKLER: What was your 18 Q. What role, if any, did the -- 18 projection? 19 CHIEF JUDGE BARNETT: I'm sorry. 19 THE WITNESS: So at the time we made 20 Mr. Herring, I'm curious, if you know, 20 the initial proposal, I believe the rates we were 21 did you propose this per-play rate because that was 21 paying were .0012, so it wouldn't have been a 22 what was extent under the CRB rules, or did you 22 significant decrease off the ad rate. But we did 23 consider other rate structures and then divide that 23 propose a unitary rate. 24 the per-play rate was the better of the others? 24 JUDGE STRICKLER: Well, what did you 25 THE WITNESS: So you mean as a 25 think that would result in in terms of percentage

4206 4208

1 structure versus a percent of revenue? 1 of revenue? 2 CHIEF JUDGE BARNETT: Correct. 2 THE WITNESS: Yeah. 3 THE WITNESS: Okay. Because it's a 3 It probably would have been slightly 4 different rate, obviously, than what the per-play 4 below -- if we were 55 percent, it would have been 5 rates were. There was a couple reasons why we went 5 like, you know, 52 percent of revenue, and that's 6 per play versus percentage of revenue. In general, 6 my approximate guess. We did the analysis, but the 7 it's because the percentage of revenue thresholds 7 point-in-time analysis isn't as important as how it 8 that we thought were fair and appropriate we 8 works over time when we do these sorts of deals. 9 wouldn't be -- you know, be hitting any time, you 9 So it's about projecting where we think our 10 know, as a -- would have already been a big drop in 10 monetization could go and whether that and how that 11 the existing revenue that we would have paid to the 11 per-play rate impact our ability to invest and grow 12 labels. We didn't think that -- at the time, I 12 the business. 13 mean in late 2013 or early 2014, you know, we're 13 CHIEF JUDGE BARNETT: I have another 14 paying 55 percent of revenue or so based upon on 14 question. 15 the Pureplay rate at the time, and that wasn't 15 THE WITNESS: Sure. 16 something that we would have agreed to and so that 16 CHIEF JUDGE BARNETT: Did you think 17 -- so the percent of revenues we thought were 17 Merlin just wouldn't go for the percentage of 18 appropriate would have been impossible at the time 18 revenue rate because of the shock factor, or was it 19 for Merlin to agree to, or, you know, would have 19 really substantially different from what you would 20 not been palatable. 20 be paying them per play? 21 When we looked at the Pureplay rate, 21 THE WITNESS: The percent of revenue 22 we're okay with Pureplay rates in the context of 22 rate historically, at least from our perspective, 23 it's a very defined cost and so that, you know, the 23 plays a different role. So the Pureplay rate 24 monetization becomes my problem. Right? I'm just 24 ensures that the labels get paid something as our 25 pay ing the fee to them, but it's also my 25 model develops, maybe is the best way of putting

4209 4211

1 that. If it was just a percentage of revenue 1 JUDGE STRICKLER: You used the word 2 basis, we would be accused of not driving revenue 2 "pure" to distinguish it from the greater-of 3 high up. We still get accused of that, but 3 formula? That's your use of the word "pure"? 4 generally speaking, that -- because there's a 4 THE WITNESS: No. I meant per play 5 pure -- per-play rate -- sorry -- that that 5 meaning just a cost per play song, not -- no 6 protects against that. Right? That monetization 6 reference to the -- to the settlement rates. 7 is important. 7 BY MR. LARSON: 8 On a -- we believe that long-term 8 Q. And another follow-up question, I think 9 percent of revenue actually aligns our interest a 9 you said that the per-play rates you initially 10 lot more directly. It's actually one of the 10 proposed would have resulted in a paying -- Pandora 11 reasons when Merlin proposed the 25 percent floor 11 paying a percentage of revenue somewhere in the 12 as a percent of revenue, we were okay with that 12 neighborhood of 50 percent; is that right? 13 level. Even though we didn't think during the term 13 A. I think at -- in the moment that we 14 of this agreement, at least in the first two years, 14 proposed it, that would have been correct. 15 likely in the next two years we would achieve a 15 Q. And how would that -- according to your 16 moment where we would be paying on the percent of 16 projections of monetization and the like, how would 17 revenue basis, we thought 25 percent of revenue was 17 that percent of revenue change over time in those 18 a, you know, an appropriate level of revenue share 18 per-play rates? 19 that where we could still invest aggressively and, 19 A. Well, I mean, that percent of revenue 20 you know, earn a return on our investment to pay 20 would have dropped alongside our ability to drive 21 off -- pay for the investing we made historically, 21 monetization. So in that same time over the period 22 but also pay a reasonable royalty to the copyright 22 of being front to late 2013 to now, our 23 holders and let them share in the upside. 23 monetization increased significantly where our 24 Now, one of the problems with the 24 percent of revenue paid today is in the 40s. If we 25 per-play rate is that if we -- if I'm monetizing at 25 were, you know, paying it at a -- at the .0011 cent

4210 4212

1 $200 revenue per thousand hours, you know, all that 1 per play -- per-play rate, you know, that would be 2 upside accrues maybe to us, and I would say it 2 probably closer to 40 percent. 3 sounds great for Pandora but in a world where we're 3 Q. Okay. And where do your projections 4 partners, that's not a great outcome. So it 4 show that rate going out over the next couple of 5 protects them in the event that there's significant 5 years? In other words, would it drop below 40 6 upside on the revenue side and so they get to 6 percent? 7 participate in that revenue upside. 7 A. Yeah. Well, I mean, we intend to 8 So finding that's part of the win-win 8 continue to invest in our monetization levels, you 9 in these negotiations. We started in one place 9 know. If it remained at .0011 for the -- you know, 10 that worked for us economically. We got there 10 for the next few years, we certainly would approach 11 eventually through a lot of give and take, but also 11 25 percent of revenue in terms of label payments. 12 by adding things that made it a win-win on both 12 JUDGE STRICKLER: Now we're talking 13 sides of the table. 13 about just Merlin? 14 BY MR. LARSON: 14 THE WITNESS: If that -- if that was 15 Q. A couple of follow-up questions, if I 15 the rate that we paid overall. Just Merlin. 16 could on that, Mr. Herring. 16 Because Merlin is a little over 6 percent of our 17 First of all, in the answers that you 17 spins, that rate by itself doesn't move. You know, 18 were giving in the colloquy with the judges, a 18 it's not enough to make a material difference to 19 number of times I think you referred to Pureplay 19 the percentage of revenue. You know, maybe a 20 rates. I just want to be clear for the transcript, 20 hundred basis points of 150 basis points, but not 21 is that -- you're referring to Pureplay rates or 21 5, 6 percentage points. 22 per-play rates? 22 BY MR. LARSON: 23 A. I was referring to the per-play rates 23 Q. What role, if any, did the upcoming Web 24 structure, not -- not the Pureplay rates that we 24 IV proceeding play in Pandora's decision to seek 25 operate under. 25 direct licenses?

4213 4215

1 A. Well, you know, any direct license we 1 BY MR. LARSON: 2 do would be, you know, in a -- it would be 2 Q. Was Pandora's ability to steer towards 3 generally known by both sides, could potentially be 3 Merlin a part of the negotiations? 4 served as a benchmark in these processes. But our 4 A. Yes. It was central to -- central to 5 initiative is to drive direct licensing as part of 5 the negotiations. Once we moved into a discussion 6 a much broader strategy that involves a lot more 6 off of a single unitary rate, our ability to steer 7 than just even domestic rates, but also our ability 7 was the way Pandora accomplished its economic 8 to have non-DMCA compliance features and 8 objectives under the deal. It also was the 9 functionality within -- within our service to 9 mechanism with which Merlin accomplished its 10 expanding international markets, you know, overseas 10 objectives under the deal. 11 and other markets. There is no equivalent of these 11 JUDGE STRICKLER: You said a moment ago 12 statutory rates. 12 in response to one of my questions that the 13 So if we want to expand outside of the 13 steering agreements preceded the Merlin 14 United States, which is a clearly stated goal of 14 negotiations, correct? 15 the business, we would need to do direct deals and 15 THE WITNESS: The steering experiments 16 work closely with label partners. So establishing 16 did, yes. 17 direct label relationships now and establishing 17 JUDGE STRICKLER: Steering experiments. 18 that partnership is important part of our long-term 18 So, did you go into negotiations with 19 strategy. 19 the goal of having steering in the contract? 20 Q. Were you involved personally in the 20 THE WITNESS: We went into the 21 negotiation of the Merlin license? 21 negotiations having -- knowing that if we were able 22 A. Yes. 22 to get pricing that was below what the statutory 23 Q. Okay. And over what period of time was 23 rates were or how we were paying everybody else, we 24 that license or agreement negotiated? 24 could use steering to bring our overall costs down. 25 A. I think the initial fall of 2013, 25 So if we had gotten just a single unitary rate, we

4214 4216

1 November or so of 2013, maybe December through the 1 would have then used steering to optimize our own 2 signing of August 2014. 2 internal cost structure by steering to a 3 Q. Okay. And how would you generally 3 lower-priced offering. It became a term of the 4 describe the nature of the negotiations, their 4 deal only after they, you know, were looking for a 5 tenor? 5 headline rate that was higher than the economic 6 A. So they were constructive. I think 6 rate. 7 they -- there was definitely a lot of back and 7 JUDGE STRICKLER: Correct me if I'm 8 forth. You know, both sides had very specific 8 wrong, you said that originally you proposed the 9 ideas about the -- what they were wanting to get 9 .0011, you meaning Pandora, correct? 10 out of the agreement, and certainly on the Merlin 10 THE WITNESS: Yes. 11 side they had lots of constituents to represent, 11 JUDGE STRICKLER: So when you proposed 12 and, you know, I'm assuming there was a lot of back 12 the .0011, it was without any steering, whatsoever? 13 and forth even within Merlin to understand -- make 13 THE WITNESS: Well, it was with the 14 sure we got to terms that everybody was comfortable 14 understanding that our pitch was is if you give us 15 with. 15 a lower rate, we will spin you more. 16 JUDGE STRICKLER: Were you the lead 16 JUDGE STRICKLER: Let's go back to the 17 negotiator? 17 number itself. When you proposed the .0011 -- 18 MR. LARSON: Chris Harrison was 18 THE WITNESS: Yes. 19 probably more the lead on the day-to-day basis, but 19 JUDGE STRICKLER: -- did you say 20 in terms of determining what terms we would accept 20 everyone would introduce the steering term, as 21 and not accept I was the lead from Pandora. 21 well, or you just said -- in other words, you said, 22 JUDGE STRICKLER: And by terms, you 22 here's the rate we propose and if Merlin had said 23 mean rates and terms? 23 deal on that rate, it would have been no steering 24 THE WITNESS: Rates and all the terms 24 in the agreement? 25 of the agreement, yes. 25 THE WITNESS: My recollection is the

4217 4219

1 pitch is this is the rate we propose. If you give 1 A. Yes. 2 us this rate, we will spin you -- we will have an 2 Q. And Mr. Van Arman, has his company 3 economic incentive to play your music more so that 3 opted into the Merlin deal? 4 you will get a defined benefit from it. Meaning 4 A. Yes. 5 greater exposure and promotion across Pandora's 80 5 Q. And Mr. Van Arman testifies that 6 million users. 6 Merlin, quote, "saw the appeal of gaining a first 7 JUDGE STRICKLER: Did you indicate you 7 mover or direct license advantage, particularly in 8 were going to put that in the contract, or you were 8 an environment where independent record companies 9 just saying it will give us an economic incentive 9 are typically not afforded the first opportunity to 10 to do so; we won't make it an obligation to do so? 10 take advantage of such a term." 11 THE WITNESS: I have to admit I don't 11 Is that consistent with your 12 remember what was in the original proposal, and 12 understanding of what Merlin was seeking in the 13 whether there was a commitment or not to it. It 13 negotiations? 14 quickly became a commitment. 14 A. Yes, it is. 15 JUDGE STRICKLER: I guess my more 15 Q. And just one more, do you understand 16 important point that I'm trying to pointe out about 16 that Beggars Group has opted into the Merlin 17 is when you went into the negotiations right from 17 agreement? 18 the get-go, did you want to have a steering term in 18 A. Yes. 19 there? Not whether you had the economic benefit of 19 Q. So Mr. Wheeler of Beggars Group claims 20 steering, but whether you wanted an economic -- 20 in his rebuttal testimony that the Merlin agreement 21 whether you wanted a contractual term to dictate 21 is not the result of free market negotiations; but, 22 steering? 22 rather, the result of an opportunity to experiment 23 THE WITNESS: I honestly don't 23 with direct licensing under the statutory rates. 24 remember. If in the first pitch -- within the 24 Do you agree with that statement? 25 second round of term sheets, it was a defined term 25 A. No.

4218 4220

1 that we were committing to. I don't remember, 1 Q. Why not? 2 honestly, if it was an actual in our initial term 2 A. Nothing in this agreement feels like an 3 sheet whether that was a commitment or not, I would 3 experiment. It was an agreement that involved not 4 have to say. 4 some small portion of their music or one or two 5 JUDGE STRICKLER: Thank you. 5 artists, but their entire catalog; a significant 6 BY MR. LARSON: 6 amount of time, a two-year time frame with two 7 Q. Mr. Herring, shifting gears just 7 additional years tapped on to that. It involved, 8 slightly, have you had a chance to review the 8 you know, significant economic terms and also 9 public version of Mr. Lexton's written rebuttal 9 promotional benefits in terms of additional 10 testimony? 10 exposure on our platform. You know, it felt like 11 A. Yes. 11 an arm's length negotiation for -- and a real 12 Q. Okay. And who is Mr. Lexton? 12 agreement to Pandora and certainly to Merlin, at 13 A. He is in charge of business affairs and 13 least the best we could tell. It certainly didn't 14 the lead negotiator for Merlin. 14 feel like any kind of experiment. 15 Q. Now, Mr. Lexton testifies that one of 15 BY MR. LARSON: 16 Merlin's objectives in its negotiations with 16 Q. Mr. Wheeler's quote and -- 17 Pandora was, quote, "to try to obtain as much value 17 JUDGE FEDER: Excuse me, Counsel. 18 as it could for its members that Pandora otherwise 18 MR. LARSON: Yes. 19 would not provide." 19 JUDGE FEDER: You said two years taxed 20 Is that consistent with your 20 on. Is that an option or has it been definitively 21 understanding of what Merlin was seeking in the 21 been extended to be an additional two years? 22 negotiations? 22 THE WITNESS: You know, there's -- each 23 A. Yes. 23 year is an option. The first year is at Merlin's 24 Q. Okay. And have you had the chance to 24 option and the second year is at mutual option. We 25 review Mr. Van Arman written rebuttal testimony? 25 both need to agree.

4221 4223

1 JUDGE FEDER: And I don't know if this 1 Q. Okay. And then I think you said a 2 gets into restricted information, let me know if it 2 third of those are in your system or something to 3 does, but how many of the indie labels have 3 that effect? 4 actually opted into the Merlin agreement? 4 A. Yes. 5 THE WITNESS: So there's about 15,000 5 Q. Can you explain what you mean by that? 6 labels of which about a third of which are 6 A. So if in -- a lot of this music isn't 7 currently in our system. So some of these labels 7 already on Pandora. Pandora, because we're not an 8 have one artist. These are lots of very small -- 8 on-demand service we don't have to have every song 9 JUDGE FEDER: So a third just gross 9 that exists. You will hear other on-demand -- 10 numbers or a third of repertoire? 10 you'll hear on-demand services claim to have 30 11 THE WITNESS: No. A third in terms of 11 million or 35 million songs. You know, we probably 12 gross numbers. I think it's -- it's -- I don't 12 have around 2 million, maybe something north of 2 13 know the exact total size of those 15,000. My 13 million songs on our database that we use. We like 14 guess is that it's, you know, well north of 50 14 to say we have the 2 million songs that matter, 15 percent of the repertoire. The largest labels are 15 that people want to hear. It doesn't mean that 16 the ones that have opted in. 16 somebody doesn't want to hear song number 15 17 JUDGE FEDER: Thank you. 17 million. It's just economically we don't need that 18 BY MR. LARSON: 18 to play a lean-back service. 19 Q. Why don't we turn now to the second 19 In this context, we are ingesting a lot 20 slide of your -- 20 of music that we hadn't previously ingested because 21 JUDGE STRICKLER: Just before you do 21 we're only focused on the 2 million. But as a -- 22 that, since you talked about who opted in and who 22 as part of our work with Merlin, we've been 23 did not opt in, the structure of the contract -- 23 ingesting a lot of music we previously hadn't 24 this may have come up previously and I don't recall 24 ingested and some labels that we previously hadn't 25 the answer. 25 worked with.

4222 4224

1 Did a late Merlin label have to 1 JUDGE STRICKLER: So 15,000 labels 2 affirmatively opt in or affirmatively have to opt 2 opted into the contract or were automatically opted 3 out? 3 in, as far as you know? 4 THE WITNESS: So it's my understanding 4 THE WITNESS: That's my understanding. 5 that the answer is both situations existed. That 5 JUDGE STRICKLER: How many either opted 6 there is two different levels of memberships or 6 out or otherwise are not in? 7 types of members within Merlin, and there's a 7 THE WITNESS: I don't know the answer 8 population of labels where Merlin can negotiate on 8 to that question. I don't know how many is in the 9 behalf of those labels and those labels have to opt 9 greater universe. I think it's like 20,000, but I 10 out if they don't want to accept the terms. And of 10 don't -- 11 that population, it's our understanding that 11 JUDGE STRICKLER: You think 20,000 is 12 greater than 90 percent of those labels did not opt 12 the total universe or -- 13 out. And that's the labels we talked -- you know, 13 THE WITNESS: Total universe, yes. 14 we'll talk about. They're the larger labels that 14 JUDGE STRICKLER: You think 15,000 out 15 we've already brought up are in that sort of core 15 of 20,000 opted in? 16 group. Then there's a much larger group of smaller 16 THE WITNESS: That's my understanding. 17 labels that aren't automatically opted in but can 17 BY MR. LARSON: 18 opt in voluntarily later if they choose to do so. 18 Q. Let's turn to the next slide of your 19 BY MR. LARSON: 19 deck, if we could, Mr. Herring. And because some 20 Q. I just want to clarify one thing you 20 of this material is restricted, I'll ask you not to 21 said in response to Judge Feder's question. First, 21 state the numbers out loud, if you can, if you're 22 you mentioned 15,000 labels in your answer. 22 capable of doing it that way. 23 Are those 15,000 labels who are, in 23 Does this slide reflect the main 24 fact, now covered by the Merlin license? 24 financial terms to which the parties agreed? 25 A. Yes. 25 A. It does.

4225 4227

1 Q. All right. And can you just walk us 1 2 through those basic terms? 2 3 A. So the royalty that we pay is a 3 4 greater-of structure between percent of revenue of 4 5 a prorated Merlin share -- prorated Merlin share of 5 6 our percent of revenue or a headline per-play rate 6 7 that has various discounts that can be applied to 7 8 it. One rate for advertising supported service and 8 9 another rate for subscription service. 9 10 Q. And what are those discounts, if you 10 11 would? 11 12 A. So we don't pay additional fees if we 12 13 steer member Merlin tracks up to about 13 14 12-and-a-half percent above their natural 14 15 performance rate. If we steer about that, we are 15 16 paying the sort of effective rate of 12-and-a-half 16 17 percent. 17 18 MR. KLAUS: Your Honor, I believe this 18 19 is -- I believe this is confidential information, 19 20 the percentages that are being discussed here in 20 21 terms of the spinning. I think that's been 21 22 designated as confidential and restricted by the 22 23 order. 23 24 MR. LARSON: Well -- so you're -- on 24 25 Merlin's behalf, you're making the objection? 25

4226 4324

1 MR. KLAUS: No, we're not -- we're 1 (THIS BEGINS PUBLIC SESSION) 2 making -- this has been designated by restricted 2 STEPHAN D. MCBRIDE, 3 and restricted by at least one party who's 3 being first duly sworn, to tell the truth, the 4 submitted something to the proceeding. I'm not 4 whole truth and nothing but the truth, testified as 5 sure that Pandora can -- they're not -- they're not 5 follows: 6 here to say anything. I'm not sure that Pandora 6 MR. MARKS: Good morning, Your Honors. 7 can -- it's appropriate for it to waive the 7 We've prepared some demonstratives that we'll be 8 restricted information of someone else who's 8 using during Dr. McBride's direct examination. We 9 designated as well. 9 have them on the screen and I believe you also have 10 MR. LARSON: We're happy to go to 10 been given copies of printouts, so whatever you 11 restricted session. We thought we could do it at 11 find more convenient. May I proceed? 12 this level and stay on the public record, but to be 12 CHIEF JUDGE BARNETT: Yes, please. 13 safe we can go to restricted. 13 DIRECT EXAMINATION BY COUNSEL FOR 14 CHIEF JUDGE BARNETT: Are you going to SOUNDEXCHANGE 15 have additional restricted questions? 14 BY MR. MARKS: 16 MR. LARSON: The next couple of slides 15 Q. Dr. McBride, will you please state your 17 I think will be. 16 full name for the record. 18 CHIEF JUDGE BARNETT: Okay. Ladies and 17 A. Stephan Douglas McBride. 19 gentlemen, if you are in the hearing room and you 18 Q. Where do you work? 20 have not signed a nondisclosure certificate, please 19 A. Pandora. 21 wait outside a couple of minutes. 20 Q. What is your job title? 22 (THIS ENDS PUBLIC SESSION) 21 A. Senior scientist, economics. 23 (RESTRICTED SESSION BOUND 22 Q. How long have you worked for Pandora? SEPARATELY) 23 A. Just over a year. 24 24 Q. Would you please identify for the 25 25 Judges any undergraduate and graduate degrees that

Before the UNITED STATES COPYRIGHT ROYALTY JUDGES THE LIBRARY OF CONGRESS Washington, D.C.

) In the Matter of ) ) DETERMINATION OF RATES AND TERMS ) Docket No.14-CRB-0001-WR FOR DIGITAL PERFORMANCE IN SOUND ) RECORDINGS AND EPHEMERAL ) RECORDINGS (WEB IV) ) )

WRITTEN DIRECT TESTIMONY OF STEPHAN MCBRIDE (On behalf of Pandora Media, Inc.) Introduction

1. My name is Stephan McBride. I am a researcher on the Science Team at Pandora

Media, Inc. ("Pandora"), Pandora's internal group of researchers tasked with designing studies and performing analyses to improve the listening experience ofthe service. My title is Senior

Scientist, Economics, a position I have held since joining the Company in April 2014.

2. I graduated from Stanford University with a Ph.D. in Economics in 2008. I also hold graduate degrees from Yale University and Queen's University. Prior to joining Pandora, I worked as a professional economist consulting companies on pricing, valuation, and intellectual property matters. I designed and analyzed numerous economics studies, including experimental studies, during this period. Including my experience during graduate studies, I have over a decade of experience performing and analyzing economic studies.

3. The primary focus of my role at Pandora is designing, executing, and analyzing studies and experiments that measure how Pandora's service (and changes to features ofthe service) impacts listener retention, pricing, and the broader music industry. In the performance of my duties, I regularly work with my Science Team colleagues, Pandora software engineers, and business managers to create and implement analyses that could produce guidance or answers to questions of business relevance to Pandora. For example, we recently investigated the effect on the hours listened and retention of listeners ofrecommending to listeners additional stations they may wish to create. Other experiments have addressed topics such as the effect of different compensation packages for Pandora sales personnel, the effectiveness of music recommendation based on computer-analysis ofthe music signal, the effect of push notifications, and effectiveness of targeted ads for Pandora One.

4. I submit this testimony to describe for the Copyright Royalty Judges the research methods used by the Science Team in general, and the specific application of these research methods to the design and analysis of two studies created to assess the impact of changes to the core Pandora music recommendation algorithms: what I refer to below as the "Steering

Experiments" and the "Music Sales Experiments." Although I undertook these two studies at the request of counsel in relation to this proceeding, they are materially equivalent to the types of studies that I and other members of the Science Team regularly perform in the course of our roles at Pandora. 1

Scientific Research at Pandora

5. The Pandora Science Team performs rigorous research and analyses to measure the effectiveness of Pandora features to improve the listening experience on Pandora. The

Science Team is comprised of 15 individuals, 13 of whom hold doctorate degrees in computer science, engineering, statistics, or economics from leading academic institutions, and who have considerable substantive expertise in areas related to Pandora's core mission.

1 For example, the Steering Experiments discussed here follow similar experiments conducted in 2013 at the request of management to inform important business decisions regarding entering into direct licenses.

2 6. The Science Team designs and executes analyses using the vast data available at

Pandora. We employ standard methods used by statisticians, computer scientists, and economists working in academia and at other technology companies. These methods can include traditional statistical methods such as linear regression or logistic regression, as well as machine learning approaches to analyze large quantities of data (a topic I address in more detail below).

7. The Science Team has primary responsibility for designing and analyzing controlled experiments, the most rigorous approach to data analysis. Controlled experiments are comparisons between groups of listeners, one of which randomly receives a manipulated experience (the "treated" group) and the other of which randomly receives the standard Pandora experience (the "control" group). When experiments are (1) randomized, (2) controlled, and (3) blind,2 they represent the "gold standard"3 for determining the causal impact of the manipulated experience. All Pandora experiments meet these gold-standard experimental requirements, ensuring that comparisons between the treated and control groups provide the most rigorous estimate ofthe effect of the changed experience that can be identified.4

8. Pandora has set up a framework for conducting experiments in which a Pandora

Scientist or Engineer intentionally changes one aspect of the Pandora experience for a sample

2 "Randomized" means assignment of listeners to treatment is ultimately random, in contrast to deterministic assignment by the researcher. "Controlled" means the primary outcome is a comparison between those receiving the exposure and those not to account for the 'placebo effect'. "Blind" means experimental subjects are unaware of their assignment to the treatment or control. In digital experimentation, blinding goes further by masking subjects from their participation in the experiment. Random assignment is critical to a well-designed experiment to remove researcher bias, to balance other factors that could affect the result (confounders), to ensure valid causal inference (whether A did or did not cause B, with what level of statistical confidence), and, most importantly, to ensure the general izability of results. 3 Such experiments, for example, are commonly used by the FDA to test adverse drug impacts, as well by courts. See Green, Michael D., D. Michal Freedman, and Leon Gordis, "Reference Guide on Epidemiology", in Reference Manual on Scientific Evidence, 3rd Edition, p. 555. 4 Currently, there are over 65 experiments operated by Pandora researchers.

3 group of listeners and then compares the effects to groups of listeners who did not experience the change. We refer to this approach as "A/B testing," where A is the control group and B is the group of listeners that receives the manipulated experience.

Pandora constructed its A/B framework to support controlled experimentation that takes advantage of Pandora's vast scale (76+ million monthly active users with over 800 million spins of music daily) and two-way communication with its users. Unlike terrestrial radio, we receive detailed information about our users' listening habits: listening time, retention, "thumbs up" and

"thumbs down," skips, station creation, subscription patterns, and the like. This has enabled, for the first time in a radio environment, the use of controlled experimentation to investigate, with sufficient power and reliability, how changes in programming and service features (among other things) affect listening behavior.

9. Such experimentation, when designed properly and executed correctly, is superior to observational studies because it directly addresses the research question at hand without additional (and untestable) assumptions. The external manipulation to create A/B comparison groups balances differences between the groups so that the comparison between treated and control groups reflects only the difference caused by the manipulation, and assures that other factors did not in fact cause the observed difference. Observational studies,. by comparison, attempt to translate an association (e.g., we see more purchases where there are more ads) into a cause (because there were more ads, and not because of any other factor, we saw more purchases).5

5 Consider a concrete example relevant to Pandora. Because we observe that listeners who create more personal Pandora stations listen more, we might ask whether Pandora should therefore encourage its users to create more stations (for example, by sending them emails highlighting that feature). It might be the case, however, that causation runs the opposite way: when users listen more frequently, they create more

4 10. This testing ability and the level of rigor with which such tests are conducted also allows Pandora to assess and improve the personalized Internet radio service we provide our listeners, the targeted advertising we use to assist our advertisers, and the revenue we provide to music creators- artists and their labels.

11. The distinction between research on Pandora using experimental methods and previous research on radio listening cannot be overstated. Consider a likely familiar problem: did service X (terrestrial radio/satellite radio/Pandora) increase exposure to a new artist ("New

Act")? For researchers analyzing the impact of, say, terrestrial radio, the two possible approaches would be to survey listeners ("Do you know New Act? And if so, where did you first hear about New Act?") or to compare listening of New Act across regions (arguing that increased listening was primarily attributable to their specific radio exposure in that region).

Neither approach, however, can control for other reasons- differential exposure to social media across interviewees/regions, recall biases, regional genre preference, or even the "home town" effect ofNew Act band members- that could explain differential familiarity. Accordingly, results from these analyses are typically anecdotal and not generalizable. In contrast, random assignment of exposure to New Act on Pandora across listeners (making the band's music available to some listeners but not others) accounts for these differences so that tracking, say, stations created for New Act or total spins for New Act (treatment v. control) provides a direct measure ofthe impact ofPandora's increased exposure.

The Steering Experiments

12. For the Steering Experiments, we were attempting to answer the question of whether increases or decreases in performances of sound recordings owned by a particular record stations. Only a controlled experiment in which Pandora encourages station creation for a treatment (B) group and compares resulting listening behavior to the control group (A) will answer the question.

5 company would have a measurable impact on a key listener metric, specifically average hours listened per registered user. 6 Secondarily, these experiments were also designed to test/substantiate Pandora's engineering capability to precisely manipulate the share of music played according to various criteria (here, the record company that owns the recordings).

13. To answer these questions, my colleagues and I intentionally manipulated the share of music played on the service for test groups of listeners based on the companies that own the sound recordings. Three ownership groups were considered in these experiments: Universal

Music Group ("UMG"), Sony Music ("Sony"), and Warner Music Group ("WMG"); all other sound recordings were classified as "Other." We conducted the experiments at the request of counsel and with the input of Professor Carl Shapiro, who provided detailed instructions

("Shapiro Directions") as described below.7 Although undertaken pursuant to Professor

Shapiro's direction, these experiments closely match previous experiments that the Science

Team ran between summer 2013 and January 2014 for business investigations in which the core music recommendation algorithms were manipulated to increase/decrease the spin share of music based on the ownership of the sound recording.

14. I was deeply involved in the design, execution, and preparation of results for these experiments. To the extent certain tasks related to the experiments were handled by my colleagues, I am well aware of those aspects and can speak to them. I provide an overview of the experimental design and analysis here; Technical Appendix A includes additional detail.

6 This is the most commonly studied measure we use for assessing changes to the core Pandora music recommendation algorithms. Almost all listening on Pandora is by users who are registered. 7 These instructions were followed in all material respects. The only exception: the experiments ended on September 3, 2014, two days later than directed, to ensure the final period had a full week's worth of data. A member of Professor Shapiro's team approved this extension prior to the originally scheduled end date.

6 Study Details

15. The Steering Experiments consisted of a group of 12 experiments, each defined

by a combination of a target ownership group (UMG/Sony/WMG) and a target deflection in share of spins (treatment group) as compared to spins that would occur according to the standard

Pandora music recommendation results (control group). The requested spin share deflections

(the "steering") were: -30%,-15%, +15%, and +30% for each ofthe three ownership groups manipulated. These percentage deflections comprise the experimental manipulations. These experiments started on June 4, 2014 and ended on September 3, 2014 (13 weeks).

16. The Steering Experiments operated through the above-described AlB Framework.

The experimental subjects were all Pandora listeners, each of whom was randomly assigned to

one of the 12 treatment groups, to the single control group, or were included in the portion of

listeners excluded from all experiments. The treatment groups for UMG each had 5% of

listeners, the treatment groups for Sony each had 7% of listeners, and the treatment groups for

WMG each had 8% oflisteners; the control group had 10% oflisteners.8

17. The Steering Experiments were implemented by manipulating the probability that

any song from the target music group would play on Pandora using a single manipulation factor.

The single manipulation factor applied to all situations- ignoring all information about listener,

sound recording, and station- and thus was a "naive" manipulation, producing the largest

reasonable estimate of listening impact; a targeted approach that steered based on knowledge of

8 To increase statistical reliability, the treatment group sizes varied across target music group to account for differences in share of music played on Pandora, having larger sample sizes for the smaller major music groups.

7 user characteristics or listening context would almost certainly yield a smaller listening impact.9

The manipulation factor was calculated nightly for each experiment based on the achieved difference in spin share of the target; the new factor applied to the subsequent day with daily updates in the factor until a value was achieved that yielded the desired level of steering.

18. The Steering Experiments used music ownership information previously collected for Pandora business purposes, with ownership combined into the three target music groups:

UMG, Sony, and WMG. Technical Appendix A provides additional detail. At all times during the experiments, the steering used the most current list of music ownership, and applied the most current list to assess steering accuracy for all previous weeks.

Analysis

19. The Shapiro Directions set forth that the experiments be analyzed using the standard listener metric that Pandora itself tracks when running experiments, namely, average hours listened per registered listener. The AlB Framework calculates this metric daily for 7-day periods, storing the information in databases. I personally queried these databases and provided the extracted results to Professor Shapiro's team. I also calculated average hours listened per registered listener for the entire experimentation period. To confirm the experimental manipulation, I calculated the spin share for the target music group for each experiment and provided these to Professor Shapiro's team.

9 For example, we might decide to reduce plays solely of"album cuts" from a given Sony album, but not the lead singles; or, we might not deflect Sony tracks on listener stations seeded with Mariah Carey, a prominent Sony recording artist. Additional factors that might be elements of the steering rules: decreased steering with duration of listening on a particular station, or greater steering for listeners who use Pandora less frequently, or number of thumbed up songs on a station (with greater steering for stations that have many available songs).

8 Results

20. Figure 1 presents a graphic illustration ofthe experimental manipulation. The figure presents the change in spin share for the target ownership group for each of the 12 treatment groups versus the control; the actual manipulations closely approximate the target steering. Two outliers are however evident: the first week during which the manipulation factor was adjusted most significantly on a daily basis (to obtain the desired increase/decrease in record company spin share), and the week including August 18, 19, and 20. For those days, a malfunction in the AlB Framework prevented the operation of the steering manipulation. Those outliers aside, this figure demonstrates that Pandora's play list algorithm can precisely manipulate the share of music played according to a specified criterion, as the change in the actual spin share closely matches the target spin share.

21. The experimental manipulations had very minor, and in most cases, statistically insignificant impacts on listening on Pandora; that is, Pandora is able to steer both toward and away from music of the three investigated music groups with minimal or no effect on the

Pandora listening experience. Figure 2 presents the change in average hours listened, weekly, between each ofthe 12 treatment groups and the contro1. 10 Notably, these results support that

10 The figure presents a grid, with rows for the three music group targets (UMG, Sony, WMG), and the columns for the four steering targets ( -3 0%, -15%, + 15%, +30% ). The effect estimate each week is the dot, with its 95% confidence interval the attached vertical line. 11 Results were statistically insignificant for WM G.

9 As shown in Table 1, the average effects on listening hours for the entire experimental period were even smaller (and less statistically significant). 12

22. I reiterate that these small listening impacts are likely larger than would occur if the steering had been more targeted- i.e., if we reduced or increased particular plays of the record company repertoire differently depending on context (listener characteristics, sound recording characteristics, and station characteristics).

Music Sales Experiments

23. The Music Sales Experiments ("MSEs") were controlled experiments designed to test whether performances of sound recordings on Pandora have a positive or negative impact on sales ofthose sound recordings. As described below, we intentionally manipulated the availability of certain recordings on Pandora in certain locations and measured the effect on sales of the same music. This work was performed at the request of counsel in this matter, but is related to business inquiries performed by Pandora researchers to examine Pandora's effect on the music industry. Consistent with Pandora practices, the Music Sales Experiments were broad collaborations. I participated in all facets of the design, execution, data merging, quality checking, and data analyses. Two of my Pandora colleagues were instrumental contributors.

Jonah Liebert, an experienced Pandora engineer, constructed and tested the code that turned off music for listeners based on geography. Oliver Bembom, another Pandora Senior Scientist who holds a Ph.D. in Statistics from UC Berkeley, and who has worked extensively in experimental design and analysis, performed much of the data analysis. I am aware of Jonah and Oliver's work on the experiments and can speak to it here.

12 Table 1 presents the average change in listening hours treatment v. control for the 12 experiments for the entire experimental period ( 13 weeks), and for the entire experimental period excluding the week during which the system malfunction occurred (week starting August 14, 2014).

10 Study Details

24. The Music Sales Experiments are a group of research experiments in which we

prevented the performance, for all listeners in randomly selected regions, of either (i) recordings new to Pandora ("New MSEs"); or (ii) catalog recordings long spinning on Pandora ("Catalog

MSEs"). The experimental manipulation was to not spin the referenced sound recordings on

Pandora during the experimental period. In contrast to the Steering Experiments, the manipulation was not a change to the core Pandora music recommendation algorithms, but rather a complete disabling of those recordings in the randomly selected geographic regions. For

listeners not treated with the experimental manipulation (the control group), music played according to standard processes by which Pandora creates playlists.

25. The experimental manipulation- not spinning on Pandora during the experimental period- occurred at the region level in this experiment rather than the listener

level. (Pandora lacks music sales data at the listener level.) To match the experimental manipulation to the available sales data, we segmented listeners according to the most granular region for which we could access weekly music sales data.

26. Table 2 presents a summary of the New MSEs and the Catalog MSEs. All experiments occurred between June and September 2014, and all were 8 weeks in duration, except for the second Catalog MSEs. 13 The New MSEs manipulated all tracks on albums that were new to Pandora, while those in the Catalog MSEs manipulated specific catalog songs. 14

13 The earliest New MSEs started on June 10, 2014; the last New MSEs ended September 3, 2014. The first Catalog MSEs started on July 16, 2014, and the second Catalog MSEs (explained in detail below) started on August 12,2014. All Catalog MSEs ended September 9, 2014. 14 The Catalog MSEs used particular tracks because catalog recordings are often available on multiple albums, requiring an infeasible escalation of the research if all such albums were included.

11 There were 1,215 experiments (814 New MSEs; 401 Catalog MSEs) with at least one unit sold during the experimentation period.

27.

15 When including SoundScan's

"extended DMAs" (DMAs not in the top 100), and "sub DMAs" (subdivisions of the five largest

DMAs), SoundScan tracks sales for 228 mutually-exclusive US regions. 16 All Pandora listeners with a zip code mapping to one of these DMA regions were included in the experiments. 17

28. The Music Sales Experiments were not operated through the AlB Framework because that framework lacks the capacity to randomize listeners based on location of residence.

Accordingly, my colleagues and I created an algorithm that randomized listeners to treatment and control groups based on their DMAs. We performed tests on the randomization to confirm the randomness of assignment. In addition, rather than apply a single randomization to all sound recordings in the experiments, we randomized the experimental manipulation for each sound

15 See, http://www.nielsen.com/intl-campaigns/us/dma-maps.html, accessed August 2014. 16 SoundScan tracks sales for 230 US regions. However, Pandora internal data combine two of the SoundScan DMAs (Tuscaloosa and Anniston, AL) into a single region. We discovered this difference after creating the randomization for use in the first MSEs and submitting for use in the Pandora music system. Accordingly, listeners in these two DMAs were jointly randomized to treatment or control, a change in experimental design. We removed all listeners in these DMAs from the experiments; producing an effective 228 US regions. 17 The vast majority of Pandora listeners in the US were included in the experiments. As of May 2014, 97.3% of all listeners had a Zip Code and 98.73% of these listeners were resident in one of the 228 SoundScan DMAs for which we have data.

12 recording tested (that is to say, the geographic locations oflisteners in the control and treatment groups varied for the different manipulated sound recordings). This approach helped to ensure that no predictable purchasing patterns, such as genre-preference differences in particular regions, would affect the results, making them more robust. In short, the experimental design of the Music Sales Experiments was randomized, controlled, and blind, consistent with everything in the A/B Framework we normally employ.

29. The study looked at aggregate music sales in each DMA for each track/album as reported by SoundScan for the experimentation period. SoundScan tracks unit sales (physical and digital) for weeks starting on Monday, along with location of sale for the majority of music sold in the U.S. 18 We decided to use SoundScan to measure sales as this is the industry standard

- for example, it is the data used by the recording industry itself to measure and report sales.

30.

To prepare unit sales for each experiment in each DMA, we converted single sales to album sales using SoundScan's standard definition of the "Track-equivalent album"

13 ("TEA"): one album equals 10 singles. TEA sales are the units we measure in the New MSEs?1

Track sales are the units we measure in the Catalog MSEs.

31. For the New MSEs, we pre-selected a week for the experiments to start, and identified all music that would newly spin on Pandora that week for inclusion in the experiments.

We had no advance information about the music that would be released that week when making the determination for the start week; we also did not inform any Pandora personnel involved in new music ingestion about the impending experiment. We know of no reason to deem the initial week exceptional.22 We later added additional rounds ofNew MSEs, applying the same "all comers" inclusion rules to subsequent weeks. At no point did we have advanced notice of what music was going to launch on Pandora on the selected weeks.

32. We applied two exclusion criteria to the "all comers" rule in the New MSEs to focus on albums (our unit of manipulation). First, we excluded follow-on singles- a new single from an unreleased album that had a previous single already spinning; the previous single was expected to comprise the bulk of spins. Second, in some instances, sound recordings that were on albums new to Pandora were not in fact actually "new." The same sound recording was included on another album already spinning on Pandora (for example, when a greatest hits album

we observed TEA sales smaller than zero. Such negative TEA

22 We devised rules for inclusion in the MSEs to exclude our own or other Pandora personnel from selecting the music, and thus introducing selection concerns. The rules effectively tied our hands.

14 is released that is just a compilation of previously released sound recordings). To ensure that the

New MSEs in fact only contained music new to Pandora, we excluded these tracks.23

33. For the Catalog MSEs, as with the New MSEs, we used inclusion rules that reduced researcher selection to guard against the introduction of bias. For each of the two

Catalog MSEs rounds, we randomly selected 200 songs that met our weekly spin and music sale specifications from the Rolling Stone Top 500 Songs (from April2011) and the Pitchfork 500

(based on songs released during the 1977-2006 period)- both lists of significant songs.24

34. These lists represent music deemed significant by music critics. We selected these lists so that we would be manipulating music that is broadly familiar, and that can be quickly accessed on genre stations on Pandora, among numerous other sources. These characteristics should, if anything, decrease the measured promotional effect of Pandora: because the music is familiar, one would assume it is harder for Pandora to introduce I re- introduce the music to listeners, thereby reducing the promotional effect. For these reasons, we deemed these lists a superior population to sample from to assess the Pandora effect on sales.

23 When a recording new to Pandora was a new version of a song by the same artist that was already available on Pandora, we made sure to include all versions of that song by that artist in the experiment. We did this so as to prevent the earlier track, an extremely close musicological substitute, from being played by the standard Pandora music recommendation algorithms for listeners precluded from hearing the manipulated track. 24 Inclusion criteria were minimum use on Pandora (spins), and minimum sales. See Technical Appendix B for additional detail. We excluded songs from consideration that did not meet inclusion criteria; we also excluded songs that we could not identify in either Pandora or SoundScan data. The minimum sales requirement explains why we focused on songs rather than albums for the catalog experiments: catalog albums- even those in the Rolling Stone Top 500 Albums list, for example- tend to have very limited sales, and sales of classic songs tend to be dispersed across multiple albums (notably, greatest hits albums). It was not feasible to remove and manipulate every album with any song from an album on the Rolling Stone Top 500 Albums list.

15 35. All MSEs were ended at their scheduled time, after eight-week experimentation periods for all experiments other than those in the second round of Catalog MSEs, for which, due to data limitations, we used a four-week experimentation period.

Analysis

36. The variability of music sales is well known, with some sound recordings selling tens or hundreds of thousands of tracks I albums per week, and others receiving no or negligible sales. Some ofthese differences in unit sales are predictable, for example based on the popularity of the artist as measured by sales of a recent album, or by spins on Pandora. Music sales are also highly varied across regions, some of which is also predictable.

37. Because music sales are so variable, both across tracks/albums and across regions, it is important for analyses ofthese sales to take advantage of known information to focus attention on the part that is not predictable- and in the case of experimental data, not predictable pre-experiment. Controlling for predictable sales patterns using pre-experimental data serves to remove much of the "noise" in music sales so that we can focus on the issue at hand- the effect that Pandora has on sales. Our analytic approach did exactly this.

38. Specifically, to estimate the effect of Pandora on music sales, we used a statistical approach that accounts for pre-experimental differences in population, sales of the experimented- on music, spins of that music on Pandora, and spins of all music on Pandora to provide an accurate (unbiased) estimate of the effect of Pandora while producing a small confidence interval to aid with inference. The specific statistical approach we used was targeted maximum likelihood estimation ("tMLE").25 This is a statistical method applied to measure the causal

25 Although tMLE is a relatively new approach (created in 2007), the approach has been used in multiple published studies in leading peer-reviewed academic journals and is increasingly used in biostatistics and

16 effect of experimental manipulations after accounting for the predictable variability of results based on pre-experimental data, and is known to produce more precise estimates than other approaches. As a check on the results, I also utilized an alternative approach (a so-called

"Poisson" regression); as described in Technical Appendix B, the results of this analysis align with those oftMLE, but with the theoretically expected wider confidence intervals.26 Both approaches are valid when a large number of observations have zero sales; an outcome that happens very often in the New MSEs. We use standard bootstrap methods to ensure valid inference, recognizing our experiments all use the same regions. We combined all experiments in the New MSEs, and separately all Catalog MSEs, to estimate the average promotional effect of Pandora across experiments. This estimate can reasonably have the interpretation of

Pandora's effect on aggregate sales of the manipulated music. 27

39. So as to avoid any concerns related to the particular statistical approach used, and so that our results would meet rigorous academic publication standards, we decided to use the tMLE approach prior to starting our experiments. Additional details about our analysis plan, including discussion of a small pilot experiment run to substantiate the engineering code prior to full experimentation, are contained in Technical Appendix B.

engineering. These publications, and our own prior experience with the estimator, substantiate that it produces very precise estimates of causal effects. 26 In Technical Appendix B, I present results for Poisson regression, and discuss limitations of other "standard" approaches- linear models or log-linear models when analyzing sales data characterized by tremendous skew in sales, and with a large number of observations (experiment- regions) lacking any sales. The Poisson regression model produces similar results to the tMLE results presented here. I performed another check on the resilience of the findings: I ran the tMLE analysis using only the top 200 experiments by sales; the results were similar. 27 The necessary assumption for this broader interpretation is that sales of a specific sound recording are not affected by manipulation of other music in the experiments.

17 40. We also supplemented the primary analyses by examining how Pandora's promotional impact varies with the extent ofPandora exposure; i.e., a kind of"dose-response" analysis.28 Specifically, we used the ratio of Pandora spins to SoundScan sales as our index of

Pandora exposure, calculating the average promotional effect of Pandora along a spectrum of exposure?9 By this measure, Pandora would be expected to have greater promotional impact for experiments with higher exposure ratios.

Results

41. I present results for the causal effect of performance of sound recordings on

Pandora. Positive estimates mean that music sales are greater when the music is spinning on

Pandora, that is, Pandora is promotional; negative estimates mean that Pandora substitutes for music sales. For each group of experiments, we measured the average impact of Pandora on sales overall, as well as the impact on the sales of major label recordings and, separately, on the sales of independent label recordings. In addition, I calculate average promotional impact on a per spin basis, separately for major label recordings and other recordings.

42. Table 3 presents the average estimated promotional effects of spinning on

Pandora versus not spinning on Pandora, for the New MSEs and the Catalog MSEs. As shown in the table, spinning on Pandora increases music sales by +2.31% for music new to Pandora,

28 Dose-response, as in, comparing different doses of Pandora exposure on sales. 29 We defined ranges of the spins to sales ratio, ranging from zero (including all experiments, as with the primary results), 1, 10, 25, and as high as 200; fewer experiments met the higher thresholds. We used contemporaneous sales and spins (2 times spins in control regions), that is, data during the experimental period, because most of the music in the New MSEs had, by design, no spins pre experiment. The thresholds are not mutually exclusive, so that all experiments that met the 200 spins per sale threshold were in all earlier effect estimates. For all qualifying experiments, we then recalculated the promotional impact of performances of sound recordings on Pandora.

18 and increases music sales by +2.66% for catalog music on Pandora. Both results are statistically significant.

43. Table 4 presents the average promotional effect of Pandora on music sales for both major label recordings and independent label recordings. As shown in the table, Pandora increases music sales for new music from major labels by a statistically significant +2.82%.30

For catalog music, Pandora has a promotional effect for major label recordings of +2.36% and a promotional impact on independent recordings of +3.85%.31 These results show that, even for music that is broadly familiar, playing on Pandora increases sales of music.

44. Figures 3a and 3b present the promotional impact of Pandora for experiments meeting minimum Pandora exposure thresholds, for all New MSEs and Catalog MSEs, and separately by major record company versus independent. These results provide additional evidence of a strong promotional impact of Pandora on music sales. For example, for the 409

New MSEs in which Pandora spins were at least 25 times the number of sales, Pandora is +5% promotional. For the 214 New MSEs in which Pandora spins were at least 150 times number of sales, Pandora is+ 15% promotional.32 Increasing promotional impacts are evident for the

Catalog MSEs, and for the New MSEs when separating Major from "Other". These results show that increased exposure on Pandora is associated with even greater promotional impact.

45. There are at least two reasons to look at the data in this fashion. First, I believe this result (i.e., increasing exposure) is consistent with past record label behavior with respect to

3° For all independent label New MSEs, there is no statistically significant evidence that demonstrates whether Pandora is promotional or substitutional. There is also no evidence from the New MSEs of a difference in the promotional impact that Pandora has on the independent label recordings versus the major label recordings. 31 The difference in promotional effect on catalog music is statistically significant (p-value: 0.002). 32 Note that these levels of Pandora exposure are not rare: over 50% of New MSEs had at least 25 spins per sale, and over 25% of New MSEs had over 150 spins per sale.

19 performances on terrestrial radio. As I understand it, record labels for decades have attempted

(whether through pecuniary or non-pecuniary payoffs) to influence terrestrial radio stations to play their music more frequently than the music of other labels. The record labels' attempts to increase spins relative to sales is consistent with this approach.

46. Second, I understand that it may be quite common for tracks to have significant sales before having significant spins on Pandora. For example, Pandora typically adds music to its listeners' playlists once per week on Tuesday. Because new music is often released on

Tuesday, it may be that new "hits" have been playing on other platforms for a week before they are playing on Pandora. In addition, because Pandora relies on listeners to increase (or decrease) the relative spins of music, even after a track is introduced to Pandora's listeners it may take time to spin at levels equivalent to spins on other platforms, which are more able to effect immediate change in the relative spins of a particular track. For example, Taylor Swift's most recent single,

"Shake It Off," was released on August 18th and immediately began playing on terrestrial radio stations across the country. In contrast, Pandora did not begin spinning "Shake It Off' until

August 25th. The below graph shows that "Shake It Off' was already spinning more than 10,000 times each day on terrestrial radio stations (likely reaching an audience of millions of listeners) before it began playing on Pandora.

20 TAYLOR SWIFT- Shake It Off Mediabase- All Stations (U.S.): Top 40- 12 Week Airplay Trend Plays -,- = Peak Week: 10/5/2014- 19247 Spins 2000n lW :#'/"'-~·""'' 175011- lr/';( 1500 11- --"""'"' wr#"';l'f'/' 1250 (} 1000I(} -#E,_,.:: 7500 I 5000 I 2500 / / 7t13t14 7127t14 8t1 Ot14 8/24t14 9f7t14 9/21 t14 10/5t14

It is logical that Pandora would not create a detectable impact on sales for songs that are playing heavily on alternative platforms while playing very little (relatively) on Pandora.

4 7. The above results reflect the impact of turning certain songs "on" or "off' on

Pandora (i.e., spinning or not spinning). Because in the normal course we spin different songs at different frequencies, turning Pandora "off' can mean a smaller or greater decrease in spins across recordings. To adjust for this fact, my colleagues and I normalized the results on a per- spin basis. Table 5 presents the results, in revenue per spin (with revenue measured in cents).

These results suggest that, on average for new music from major labels, a spin on Pandora increases music sales revenue by a statistically significant +0.16 cents. While not all ofthis revenue flows to content owners because of channel fees (iTune's take, for instance), this is a notable indirect transfer from Pandora to major labels and their artists. The estimated revenue on catalog music is smaller ( <0.0 1 cents per spin), but the promotional effect is statistically significant for both major and independent label recordings.

48. Table 6 presents results addressing any difference in promotional impact of

Pandora, on a per-spin basis, between major and independent labels. From a statistical

21 perspective, there is no evidence of a difference in promotional impact between majors and indies per spin.

49. The Music Sales Experiments confirm that the Pandora radio service is promotional of music sales- that is, music sales are higher when that music plays on Pandora.

As results from a well-designed gold standard experiment, they are generalizable from the specific music sampled to other new music and catalog music that spins on Pandora. We also present evidence that the promotional effect is greater for music with greater exposure on

Pandora. The robustness of the study design, and the rigor applied to the analysis, distinguish these results from any previous study of the promotional impact of a radio service.

22 >. c 0 (j)

c .....- ·- Q)Q) .... O> :::l c O>ctl ·-u..o..c

·······················--···r··············· '··~···· ...... J ...... ~ £0

0 ~ 0 ~ 0 v-0 ~ ~ I I (%) i0JlU08 "A lUawpadX3 'dnOJ8 ~!Snll\l Aq SU!dS U! a6UB48 McBride Figure 2

RESTRICTED-Subject to Protective Order in Docket No. 14-CRB-001-WR (2016-2010) (Web IV) OQ ?

0$'..(

ooL "'<

0$'

$'~

L' (]) l[) m 0 0 0 CJ) /1 '- ' (]) l 0... l[) OQ (/) 0 ? c 0 I 0... ,.... 0$'..( CJ) q "0 0 0 (/) .s::: > OoL (j) .....,(f) .._a> UUJ ,.... .s::: CUCJ) 0 "'< 1- 0 I a> E-~ ,.... 0$' co 0 U) -3: q .._ ~ (]) 0 $'~ a> :;::::;Z Cl.. 0 (j) ,....+ OL c E 0 ·c.. 0 U) e 0 -- L' 0.. v ! m _, 0 (") + ~ :::J OQ u:0> ? 0$'..(

ooL "'<

0$'

$'~

OL'

0->'.t

Oo.t

$'< if) Q) ..0

..(

(]) U") 0 ro c:i 0 (f) II ._ I (]) a. '1' U")' Oo (/) 0 ~ c c:i a. I 0->'.t (f) 0 "'0 c:i 0 (/) ..c > (/) Oo.t (/) .... w ....(1,) (.)({) $'< ..c ro~ 0 f- a. c:i (1,) 0') I E ~ OS' Cii 0 0 (/) 0 ..__ (]) ro 0 > ...... $'~ (1,) ·- ro c. (/) oo + 0..( c ~ E 0 ·c. ._0 0 (/) c:i ..( a. v $ ..c ('I") + 0 (]) Q) ._ ::I ::J ii'i 0') !'" Oo u:: c. ~ 0->'.t

Oo.t

$'<

:;:( OS'

$'~

0..(

..(

0 r .... I...... r

RESTRICTED-Subject to Protective Order in Docket No. 14-CRB-001-WR (2016-2010) (Web IV) Table 2. Summary of Music Sales Experiments

New Music Sales Experiments Catalog Music Sales Experiments Round 1 Round 2 Round3 Round 1 Round 2

Start Date June 10, 2014 June 24, 2014 July 8, 2014 July 16,2014 August 12, 2014

End Date August 5, 2014 August 19, 2014 September 3, 2014 September 9, 2014 September 9, 2014

Duration of Experiment 8 weeks 8 weeks 8 weeks 8 weeks 4 weeks

Sound Recordings Manipulated Album Album Album Song Song

Music Selection Music New to Pandora, Music New to Pandora, Music New to Pandora, Rolling Stone Top 500 Songs Pitchfork 500 Songs Week of Week of Week of (Pre April 20 II), (1977-2006) June 10,2014 June 24, 2014 July 8, 2014 random selection

Inclusion Criteria Sound recordings new to Pandora during selection week, plus all other 2:1000 spins in June 2014 2:700 spins week prior to tracks on the same album, and all earlier versions of the same sound 2:25 sales previous week experiment recording with at least one sale in the experimentation period 2:25 sales previous week

Exclusion Criteria Subsequent lead singles (a new single with the same album id as an already Song not identified in Inclusion in Round I of spinning single); greatest hits I compilations (albums with a new track that Pandora or SoundScan data Catalog MSEs, song not is on another, already spinning album) found in Pandora or SoundScan data

Number of Experiments 326 269 217 200 201

Number of Manipulated Sound Recordings 2395 1454 1082 505 451

Average Daily Spins 6750 5432 3567 25176 17568

Notes: I. Experiment counts only include experiments with sales. 2. Average daily spins calculated as the ratio of spins during the experiment divided by the number of experiments divided by number of days in the experimentation period. Table 3. Music Sales Experiments Results - Effect of Performance on Pandora

New Music Sales Experiments (N=814) Catalog Music Sales Experiments (N=401) Effect 95% CI P-Value Effect 95% CI P-Value

Effect of Performance on Pandora on Music Sales 2.31% [0.52%- 4.79%] 0.012 2.66% [2.26%- 3.14%] <0.0005

Notes: !. See Written Statement and Technical Appendix B for description of New MSEs and Catalog MSEs. 2. Estimates from targeted maximum likelihood estimation on aggregate sales during experimentation period, using pre-experimental baseline factors across experiment and region. See Technical Appendix B for additional detail. 3. Standard errors bootstrapped using 1999 replicates. Table 4. Music Sales Experiments Results - Effect of Performance on Pandora by Music Ownership Group

New Music Sales Experiments Catalog Music Sales Experiments Effect 95% Cl P-Value Number Effect 95% CI P-Value Number

Effect of Performance on Pandora on Music Sales Major Recording Label 2.82% [0.84%- 5.44%] 0.007 124 2.36% [1.95%- 2.85%] <0.0005 294 Other Recording Label 0.62% [-2.86%- 4.88%] 0.682 690 3.85% [2.9%- 4.87%] <0.0005 107

Notes: I. See Written Statement and Technical Appendix B for description of New MSEs and Catalog MSEs, and identification of music ownership group. 2. Estimates from targeted maximum likelihood estimation on aggregate sales during experimentation period, using pre-experimental baseline factors across experiment and region. See Technical Appendix B for additional detail. 3. Standard errors bootstrapped using 1999 replicates. Table 5. Music Sales Experiments Results - Revenue to Labels and Artists from Per-Spin Promotion Effect

New Music Sales Experiments Catalog Music Sales Experiments Effect 95% CI P-Value Effect 95% CI P-Value

Per-Spin Effect of Performance on Pandora Major Recording Label 0.167 ¢I spin [0.05- 0.32] 0.008 0.006 ¢ I spin [0.005- 0.008] <0.0005 Other Recording Label 0.048 ¢I spin [-0.22- 0.36] 0.727 0.008 ¢I spin [0.006- 0.01] <0.0005

Notes: I. Results are cents per spin, assuming that track-equivalent units sell for $1. 2. See Written Statement and Technical Appendix B for description of New MSEs and Catalog MSEs, and identification of music ownership group. 3. Estimates from targeted maximum likelihood estimation on aggregate sales during experimentation period, using pre-experimental baseline factors across experiment and region. See Technical Appendix B for additional detail. 4. Standard errors bootstrapped using 1999 replicates. Table 6. Music Sales Experiments Results - Difference in Per-Spin Promotional Effect

New Music Sales Experiments Catalog Music Sales Experiments Effect 95% Cl P-Value Effect 95% CI P-Value

Difference in Per-Spin Effect of Performance on Pandora Major vs. Other Recording Label 0.12 ¢I spin [-0.18- 0.44] 0.409 -0.002 ¢ I spin [-0.003 - 0.001] 0.113

Notes: l. Results are cents per spin, assuming that track-equivalent units sell for $1. 2. See Written Statement and Technical Appendix B for description of New MSEs and Catalog MSEs, and identification of music ownership group. 3. Estimates from targeted maximum likelihood estimation on aggregate sales during experimentation period, using pre-experimental baseline factors across experiment and region. See Technical Appendix B for additional detail. 4. Standard errors bootstrapped using 1999 replicates. Table 7. Music Sales Experiments Results- Effect of Performance on Pandora (Poisson Regression)

Effect of Performance on Pandora New Music Sales Experiments Catalog Music Sales Experiments on Music Sales Effect 95% CI P-Value Effect 95% CI P-Value

All Experiments 2.00% [-0.02%- 4.49%] 0.063 2.71% [1.78%- 3.71 %] <0.0005

Major Recording Label Experiments 2.55% [-0.32%- 5.56%] 0.072 2.37% [1.25%- 3.5%] 0.001 Other Recording Label Experiments 0.30% [-3.54%- 4.29%] 0.855 3.37% [0.91%- 5.13%] 0.002

Notes: 1. See Written Statement and Technical Appendix B for description of New MSEs and Catalog MSEs, and identification of music ownership group. 2. Estimates from Poisson regression pooling all New MSEs I Catalog MSEs. See Technical Appendix B for additional detail. 3. Standard errors bootstrapped using 1999 replicates. Table 8. Music Sales Experiments Results- Effect of Performance on Pandora (Top 50 Experiments by Sales)

Effect of Performance on Pandora New Music Sales Experiments Catalog Music Sales Experiments on Music Sales Effect 95%CI P-Value Number Effect 95%CI P-Value Number

All Experiments 2.22% [0.43%- 4.61 %] 0.018 100 3.07% [2.56%- 3.66%] <0.0005 100

Major Recording Label Experiments 2.70% [0.52%- 5.21%] 0.015 50 2.77% [2.16%- 3.52%] <0.0005 50 Other Recording Label Experiments 0.46% [-3.1%- 4.8%] 0.797 50 3.73% [2.81%- 4.75%] <0.0005 50

Notes: 1. See Written Statement and Technical Appendix B for description of New MSEs and Catalog MSEs, and identification of music ownership group. 2. Estimates from targeted maximum likelihood estimation on aggregate sales during experimentation period, controlling for pre-experimental factors across experiment and region. See Technical Appendix B for additional detaiL 3. Standard errors bootstrapped using 1999 replicates. Before the UNITED STATES COPYRIGHT ROYALTY JUDGES THE LIBRARY OF CONGRESS Washington, D.C.

In re ) ) DETERMINATION OF ROYALTY ) Docket No. 14-CRB-0001-WR (2016-2020) RATES AND TERMS FOR ) EPHEMERAL RECORDING AND ) DIGITAL PERFORMANCE OF ) SOUND RECORDINGS (WEB IV) ) ) ·------DECLARATION OF STEPHAN MCBRIDE

I, Stephan McBride, declare under penalty of perjury that the statements contained in my

Written Direct Testimony in the above-captioned proceeding are true and correct to the best of .;~ my knowledge, information, and belief. Executed this? d~y of October 2014 in Oakland,

California. TECHNICAL APPENDIX A: The Steering Experiments Overview

Experimental Manipulation: the share of music played on Pandora's service based on

the companies that own the sound recordings. Professor Shapiro requested specific "steering"

targets: -30%,-15%, +15%, and +30%, applied to each ofthe experimental targets, yielding 12

total experiments: 4 steering targets x 3 experimental targets. The manipulation used was na"ive

- a common constant factor affecting the likelihood a song owned by the target would play, with

no consideration ofthe listener, sound recording, or station characteristics.

Experimental Target: Music owned by one ofthree music ownership groups: Universal

Music Group ("UMG"), Sony Music ("Sony"), or Warner Music Group ("WMG"). Sound

recording ownership was determined using publicly available information acquired for a

business purpose not related to any litigation or this proceeding. The ownership information is

updated weekly.

Experimental Subjects: All Pandora listeners were eligible for inclusion in the

experiments. Professor Shapiro directed the size of the treatment group for each of the

experiments: a 5% sample of listeners for each UMG experiment, 7% sample of listeners for

each Sony experiment, and 8% sample of listeners for each WMG experiment. The control

group consists of a 10% random sample of listeners. All listeners are blind to their assignment.

Experimental Design: Listener-level randomization of assignment to treatment or

control. The experiment was operated through the Pandora A/B Framework.

Experimental Outcome: Change in average hours listened per registered listener

between each treatment group and the control group, calculated weekly and over the entire experimentation period. Experiment Timeline: All experiments started on Wednesday, June 4, 2014; all experiments ended on Wednesday, September 3, 2014.

Study Details

1. The Steering Experiments consisted of a group of 12 experiments, each defined by a combination of a target ownership group (UMG/Sony/WMG) and a target deflection in share of spins (treatment group) as compared to spins that would occur according to the standard

Pandora music recommendation results (control group). The requested spin share deflections

(the "steering") were: -30%, -15%, + 15%, and +30% for each of the three ownership groups manipulated.

2. The experiments all started at 3 p.m. on June 4, 2014 (a Wednesday), and all experiments ended at midnight September 4, 2014 (a Wednesday, consistent with analysis of weekly data). Consistent with Pandora practice, we tracked results using full weeks, given predictable patterns in listening across days. I excluded the data for June 4, 2014 from analysis as a partial day (3pm to 12am). Weeks ran from Thursday just after midnight to the following

Wednesday at midnight, given the day the experiments started.

3. The Steering Experiments were implemented by manipulating the probability that any song from the target music group would play on Pandora. The experiment used a single manipulation factor that modified the likelihood that a sound recording owned by the target music group played, either increasing the likelihood (steering toward) or decreasing the likelihood (steering away from). The manipulation factor was originally set according to experience, and then tuned daily by comparing achieved steering versus target. The manipulation factor increased ifthe steering was insufficient, and decreased if the steering was

2 greater than target. The frequency of the manipulation factor tuning was decreased during the experiment, given noise in daily spin differences (day of week effects). 1

Ownership Information

4. The Steering Experiments used ownership information, previously collected for

Pandora business purposes. Pandora undertook a costly, lengthy, and structured program to match tracks to music groups who hold at least a partial interest in the sound recording, as

Pandora had not previously collected ownership data. Pandora matched songs to music groups by matching songs to music labels and labels to music groups. Pandora used music group label lists, internal expertise and industry-recognized external sources to execute this merging process.2 Pandora matched sound recordings to three music groups: Universal Music Group

("UMG"), Sony, and Warner Music Group ("WMG").

5. After completing the above analyses of labels and ownership information for catalog music, Pandora invested in collecting and retaining label information for each sound recording newly added to the music database. The Curation Team manually enters the ownership group (not the distributor) when ingesting new music to Pandora, using several sources to ascertain the information.3 The ownership data are updated weekly in the Pandora

1 The frequency oftuning was also reduced to tamp an observed pattern of alternating days of increasing then decreasing changes in the factor. 2 Sources included: http://www. universalmusic.com/labels, http://en. wikipedia.org/wiki/List_of_ Universal_ Music_ Group _Ia bels, http://www. universalmusic.com/labels, http://en. wikipedia.org/wiki!List_ of_ Sony_ Music_ Entertain ment_labels#Current_labels, and http://en.wikipedia.org/wiki/List_of_ Warner_ Music_ Group_labels#Main _Labels. 3 Sources include: Amazon.com from digital download music, and searches on CDDB I Amazon for physical CDs purchased. The Curation Team enters information from liner notes on physically ingested music if unable to identify information from searchers.

3 databases; each week, the steering manipulation used the then current ownership data. All results rely on the now current ownership.

Analysis

6. The Shapiro Directions stipulate that the Steering Experiments be analyzed using the standard metric for analyzing retention, namely hours per registered listener. This is the primary listener metric tracked and analyzed for internal business inquiries. The A/B

Framework calculates this metric daily for 7-day periods, storing the information in databases.4 personally queried these databases and provided the extracted results to Professor Shapiro's team, including measures ofprecision (standard deviation) and number of registered listeners per experiment. The standard deviation is the standard deviation in weekly listening across all registered listeners in that experiment.

7. Professor Shapiro's team requested that I provide average hours per registered listener estimates for two periods- the entire experimental period (13 weeks), and the entire experimental period excluding the week starting August 14, 2014, due to a systematic malfunction in the AlB Framework (no steering occurred for three days). I calculated these requested

8. The only other data requested pertained to the spin shares for the ownership groups. For this request, I computed the spin share for the target ownership group for the 12 experiments using total spins each week. I performed no further processing ofthese metrics.

4 The A/B Framework uses 7-day look back periods. The Shapiro team requested that results be presented using the first day of the week, rather than the last. I executed this change.

4 Results

9. Results are presented by week, consistent with Pandora's standard practice given substantial variation in listening across days, for each week in the experimental period.

Statistical tests compare each treatment group to the control group; given the millions of observations, the tests use the normal distribution (based on a central limit theorem), again consistent with standard Pandora practice when analyzing results from the AlB Framework.

5 PUBLIC VERSION

TECHNICAL APPENDIX B: Music Sales Experiments Overview

Experimental Manipulation: not spinning a sound recording for all Pandora listeners in randomly selected regions. For simplicity, the results are presented for the effect of spinning on

Pandora, rather than the effect of not spinning on Pandora.

Experimental Target: the "New MSEs" included music new to Pandora. The manipulated music included all albums that would have started to play on Pandora, but for the experiment, along with any earlier sound recordings on the same album that had already been spinning (to include lead singles with the album). The "Catalog MSEs" included music long spinning on Pandora, randomly selected from lists of significant music (Rolling Stone Top 500 and Pitchfork 500) that met minimum Pandora spin and SoundScan sale thresholds.

Experimental Subjects: All Pandora listeners with a Zip Code contained in one of the

228 SoundScan designated market areas (DMAs) with sufficient data (sales and zip code match).

All listeners were blind to treatment.

Experimental Design: Region-level randomization of assignment to treatment or control applied to each experiment for all 228 SoundScan DMAs. Experimental design meets all requirements for "gold standard" design.

Experimental Outcome: Music sales during the experimentation period. We used

SoundScan's own standard approach to combine album and track sales- track-equivalent albums

("TEAs"), where an album counts as 10 track sales.

Experiment Timeline: The earliest experiment started on June 10, 2014, and the final experiment ended on September 9, 2014. All experiments were eight weeks in length, except for the second round of Catalog MSEs (random sample from the Pitchjhrk 500), which were four weeks in length. Analysis: We selected targeted maximum likelihood estimation ("tMLE") to be our statistical approach prior to experimentation. We use tMLE to estimate the average promotional effect of Pandora on sales by pooling data across the experiments. This yields an estimate that approximately weights each experiment by its total sales. This estimated effect can be interpreted as the effect of Pandora on aggregate sales of all the experiments- that is, comparing aggregate sales where all the music is available to the scenario where none of the music is available- when each experimental manipulation does not affect sales of the other manipulated sound recordings. The tMLE approach uses pre-experimental information to account for substantial variation in music sales across regions within an experiment and across experiments that is predictable. Theoretical results for this estimator establish that it is efficient- that is, has the smallest confidence interval - for the average promotional effect investigated.

Inference: We compute studentized bootstrap confidence intervals. The bootstrap simulation is based on a resampling of regions (DMAs) as the independent experimental unit.

We use 1999 replicates in our presented results.

Study Details

I. The Music Sales Experiments ("MSEs") are a group of research experiments in which we prevented the performance for all listeners in randomly selected regions of either (i) recordings new to Pandora ("New MSEs"); or (ii) catalog recordings long spinning on Pandora

("Catalog MSEs"). The experimental manipulation was to not spin the referenced sound recordings in randomly selected regions on Pandora during the experimental period; the manipulation occurred at the region level in this experiment rather than the listener level.

2. We used SoundScan 'designed market areas' ("DMAs"), for which sales data would be available, to define the regions for experimentation. Including "sub DMAs"

2 (subdivisions of the five largest DMAs), SoundScan tracks sales for 230 mutually-exclusive U.S, regions. For each experiment, we randomized DMAs to the treatment group or the control group. The vast majority of Pandora listeners have provided Pandora with a zip code of residence; these zip codes were used to link listeners to DMAs based on a pre-existing Pandora database that maps zip codes to SoundScan DMAs (developed in collaboration with SoundScan previously). However, it turns out that this file combined two DMAs (Tuscaloosa and Anniston,

AL into a single region). Accordingly, our randomization code jointly assigned these two spatially proximate DMAs either to the treatment or control group. We discovered this after submitting our randomization code to the core Pandora music databases; we therefore elected to remove these two DMAs from the experiments. Accordingly, we have 228 eflective regions for all experiments.

3. We randomized assignment for each DMA for each experiment using a randomization algorithm external to the Pandora A/B Framework. We tested randomization by comparing the mean and variance of the randomized assignment to the theoretical results given an equal probability of assignment (as used in the actual experiments). These tests confirmed that the randomization code was operating appropriately.

3 5.

We computed track-equivalent album

("TEA") sales, SoundScan's own measure, to combine album and single sales, counting an album sale as the equivalent of 10 track sales.

Music Inclusion

7. For each of the New MSEs rounds, we pre-selected a week for the experiments to start, and identified all albums that would newly spin on Pandora that week for potential

4 manipulation. We had no advance information about the music that would be released that week when selecting the start week.

8. The weekly approach to the New MSEs matches the weekly data available from

Sound Scan, and the ingestion process for new music at Pandora- music new to Pandora is made available for spinning on 'l'uesday each week, from sound recordings approved by Sunday at midnight. Our "all comers" inclusion rule applied to the newly approved albums (multiple tracks with the same album_id approved that week), or single tracks that did not share an album_id with any other previously spinning track. Conversely, we excluded single releases that shared an album_id with previously spinning singles, deeming follow-on singles peculiar. In particular, we were concerned that the subsequently released single- the new music- was unlikely to constitute the primary driver of music purchasing behavior during the experimentation period.

As a manipulation at the album level, we decided to include all albums with lead singles that were already spinning on Pandora (in part because albums with lead singles are more popular, and we wanted to ensure our inclusion criteria captured these market entry dynamics), but not music that involved sequential singles as the most recent new music would be unlikely to comprise the maturity of listening during the experimental period.

9. In some instances, particular sound recordings on the list were not truly new to

Pandora, as Pandora was already spinning a prior version of the sound recording (example: a remastered version of a catalog song). To address this situation, we performed the following steps to identify our final set of music in the New MSEs:

• Combine all new sound recordings by album

• Identify all sound recordings on the album, including those that are already spinning. Exclude the album ifthe newly approved music is a single, and previously spinning singles had the same album id

5 • If the new music is an album, include all tracks on the album in the experiment (to ensure that we integrated lead singles into the album analysis)4

• Check whether any specific sound recording on the album was already spinning on Pandora on a different album. If so, that album and all of its specific sound recordings were excluded from the experiment. We did this to exclude albums of Greatest Hits or partial compilation albums.5

10. These inclusion criteria yield a reasonable definition ofmusic that is truly new to

Pandora for the purpose of analyzing music sales at the album level. However, we went further by identifying previous versions ofthe songs already spinning on Pandora, and included these in the experiments as well. That is, we turned off all versions of the selected music in the randomly selected DMAs. We performed this integration to prevent the earlier track, an extremely close musicological substitute, from being played by the standard Pandora music recommendation algorithms for listeners precluded from hearing the manipulated new sound recording.

11. However, at ingestion of new music, Pandora does not record an 'album' identifier in all cases, in part because the album itself may not be available at the release of its

"lead" single or singles, or there is no album- only a single is released. To distinguish these cases and to help ensure that new music that is likely from the same album was manipulated in the same regions, Pandora researchers strived to combine multiple, newly approved songs by the same lead artist into one experiment even when they did not share an album id, and included any subsequent releases of new music by the same artist during the experimental period in the same

4 By integrating all songs from the album, we include some music that was already spinning on Pandora; these songs did spin but were turned off in the experimental regions for the duration of the experiment. The entirely new to Pandora music never played on Pandora prior to the ingestion. 5 To perform these steps, we used several data fields, including album id and song identity (an indicator of an artist-song, rather than specific sound recording). Experiments with different song identities were not excluded.

6 experiment. We monitored the newly approved music each Monday to match against any existing experiments; when we found music matching the inclusion rules, we included it in the earlier experiment prior to it spinning. We undettook these detailed steps to produce, as much as feasible, a clean separation of new albums for other music.

12. Thus, the New Music Sales Experiments were manipulations ofthe availability of all albums new to Pandora (including any lead singles), as well as previous versions ofthe same song by the same artist. We obtained sales data only for the new to Pandora album; any effect of the experiment on earlier albums was not assessed.

13. For the Catalog MSEs, we used third party lists of significant music from various publications to select the music for manipulation. We did this to guard against researcher selection concerns. Because these lists include familiar music, we believed that we were selecting music that, if anything, would be harder for Pandora to promote (producing smaller promotional effects). We ran two rounds of Catalog MSEs, each intended to have 200 songs, sampling first from the Rolling Stone Top 500 Songs (songs released before April 20 II) and in the second round, selecting from the Pitc~fork 500 (based on songs released between 1977 and

2006).

14. Unlike in the New MSEs, we manipulated songs in the Catalog MSEs rather than albums, as we observed limited sales for the top 10 albums included in the Rolling Stone Top 500 albums list,6 a result likely attributable to increased purchases of greatest hits albums. The limited sales, and the dispersed sales across numerous albums, necessitated the change to manipulation by song rather than album. The alternative approach- including all albums with any song from the Rolling Stone Top 500 album list- was infeasible.

6 We only investigated sales tor the top I 0 albums on the Rolling Stones Top 500 Albums list.

7 15. We defined inclusion criteria for the Catalog MSEs prior to starting the experiments; we planned on randomly selecting songs for experimental manipulation from those meeting the criteria. We defined two minimum thresholds of continued use of the music-

7 minimum spins on Pandora and minimum sales in the week preceding the experiment launch - and exclusion of music in any previous MSE.

For the music in the Rolling Stone Top 500 list,

487 songs met the inclusion criteria; for music on the Pitc~fork 500, 20 I songs met the inclusion criteria. We used a random number generator to select 200 songs for inclusion in the experiment from the eligible songs. Specifically, we used the random number generator to create a uniform random number tor each eligible song; we selected the 200 songs with the highest random numbers generated in this manner. We had intended to apply the same methodology to the second round of Catalog MSEs, but when only 201 songs were eligible, we included all songs in the experiment.

I 6. In the Catalog MSEs, we disabled all versions of the included songs, meaning the same song and same artist (studio versions, live versions, remastered versions, etc.) .•

7 Specifically, for the first round of Catalog MSEs, a song was required to have at least 1000 spins on Pandora during June 2014, and at least 25 unit sales for the week ending July 6, 2014, to be included. For the second round of Catalog MSEs, we used a modified spin threshold (2':700 spins in the preceding week), a decision informed by the spin distribution of music on the Pitchfork 500 list, and the shorter experimentation period (so greater Pandora exposure was deemed helpful a priori to detect the causal effect (2':700 per week maps to ~3000 spins per month); we applied the same 2':25 sales units in the preceding week criterion. I

8 As with the New MSEs, we monitored the newly approved music list weekly for any new versions to include in the experiment.

17. The first Catalog MSEs started on July 16, 2014 and lasted for 8 weeks. The second Catalog MSEs started on August 12, 2014 and lasted for 4 weeks. All Catalog MSEs ended on September 9, 2014.

Analysis

18. Prior to performing any analyses, we decided to estimate the causal effect of performance on Pandora by grouping experiments, rather than providing estimates for all experiments individually(> 1200 effect estimates). We undertook this large number of experiments because of the predictable variability in sales for any small number of experiments; grouping ofthe New MSEs and, separately, the Catalog MSEs was designed to help reduce this

''noise''.

19. We also decided prior to analysis to separate our experiments according to Major

(UMG/Sony/WMG) or "Other" (not UMG/Sony/WMG), using the same underlying ownership data used in the Steering Experiments, to assess any differential effect across ownership group.

20. The classic experimental estimator is the difference in means (sales in our case) between the treatment and control groups; it is classic because it is unbiased (on average, gives the true estimate) and easy to understand. This approach ignores all information about the predictability of sales, thus making it harder to discern the true effect of the experimental manipulation because of higher uncertainty around the estimated effect. In the present context, this imprecise estimator would not take advantage of predictable patterns, resulting in wide confidence intervals, and thus hampering the evaluation of the causal effect of performances on

Pandora.

9 21. Prior to experimentation, we decided to use targeted maximum likelihood estimation ("tMLE"), a statistical approach that is unbiased, just like the classic estimator,9 but one that theory establishes is more efficient (that is, produces smaller confidence intervals). For these reasons, tMLE is used to analyze experimental data (perform causal inference ). 10 The tMLE achieves the efficiency gains by taking advantage of pre-experimental information in a regression fhmework to account for the predictable differences in sales. A second advantage of the tMLE approach is that it remains unbiased and consistent for the causal effect even if the regression model is not correctly specified.

22. There is a direct relationship between the precision of causal estimates, and how accurately we can predict sales from pre-experimental data; the more accurate are sales predictions, the more precise will be the causal effect estimate. The tMLE allows the use of powerful, automated statistical learning methods- methods that are standard in statistics- that flexibly adapt to the data to yield higher prediction accuracy than is typically achieved by fixed regression models specified, a priori, by a researcher. In particular, we use Lasso regression to select, in an automated way, those features that are most predictive of sales for each experiment.

This is a standard algorithm, and one that removes feature determination from the researchers.

This is in contrast to assuming the relevance of particular factors in predicting sales. For these reasons, prior to starting our experiments, we deemed the tMLE approach the best approach to test the research hypothesis.

9 Further, the tMLE is unbiased for the same reason as for the classic estimator --randomized treatment assignment. 10 Although tMLE is a new approach (created in 2007), the approach has been used in several published studies in leading peer-reviewed academic journals and it is increasingly used in biostatistics analyses of experiments, engineering, and finance. These publications, and our own prior experience with the estimator, substantiate that it produces very precise estimates of causal effects.

10 23. Targeted MLE builds on the classical maximum likelihood approaches, but is targeted for the parameter of interest (the average promotional effect, in our case). Traditional maximum likelihood approaches (including linear regressions) are aimed at obtaining an efficient estimate of the entire data generating distribution (in our case: sales in each region) and therefore chooses a bias-variance tradeoff appropriate for this goal. The purpose of our inquiry is not to predict sales for all228 regions for all 1200 experiments with sales. Quite distinctly, our area of interest is to estimate the average promotional effect of Pandora across these experiments (separately for New MSEs and Catalog MSEs), which is a much lower- dimensional parameter of the entire data generating distribution than traditional methods focus on. The method is "targeted," in that it concentrates on this lower-dimensional parameter.

24. Targeted MLE consists ofthree steps. In the first step, we use Lasso regression to obtain predictions of sales for each experiment and each DMA. In the second step, the targeting step, we update these predictions using a Poisson regression of sales on an indicator for the treatment, using the prediction from the first step as an offset. Poisson regression is the standard model for "count data" (non-negative integer outcomes like sales), 11 and especially in the presence of a high frequency of zeros. 12 In the final step of the tMLE approach, we use the updated ("targeted") predictions for each experiment-DMA ofwhat sales would have been in the presence or absence of Pandora- and then calculate the percent increase in predicted sales in the presence of Pandora.

11 Technically, our data are not "count data'' because of the track-equivalent album adjustment, making multiples of I 0 far more frequent than would randomly arise, though they are non-negative integer data. 12 Although classic Poisson regression calculates confidence intervals under the assumption that the (conditional) mean of sales equals the (conditional) variance of sales, this assumption is not relied on, either in this use, or later when I present Poisson regression results (because confidence intervals are estimated using bootstrap methods).

11 25. We directly calculate the average promotional effect for New MSEs and Catalog

MSEs by pooling all experimental data. This approach is advantaged because it includes more experimental evidence than alternative approaches would, and with the assumption of no effect on sales across the experiments, the effect estimate has the interpretation of Pandora ·s effect on aggregate sales of the music. Further, this approach is comparable in design to a pooled Poisson regression (performed below as a robustness check). We considered three indirect approaches to averaging the experimental estimates: unweighted, weighted by sales, and weighted by inverse of the variance. The unweighted estimate is highly variable; the latter two approaches yield effect estimates that are close to our chosen approach. However, they have the distinct disadvantage that they cannot include experiments in which either the set of treatment regions or the set of control regions had no sales.

26. The tMLE procedure provides analytic confidence intervals. We compared these confidence intervals with those produced by bootstrap simulation and found that they are in close agreement. We performed the bootstrap simulation by treating the DMA as the independent unit of analysis, sampling a single set of228 DMAs (with replacement) for each simulation. Consistent with treating the DMA as the unit of analysis, the data set of a given bootstrap replicate then consists of the data for all experiments corresponding to the sample

DMAs. For each such bootstrap replicate we then re-computed the average promotional effect of

Pandora. Given that the tMLE provides an analytic standard error for each of the bootstrap samples, we can compute studentized bootstrap confidence intervals using the tMLE standard error; studentized bootstrap confidence intervals have been shown to perform reliably. We elected to present the slightly larger studentized bootstrap confidence intervals rather than the analytic ones to be conservative.

12 27. We perform our estimation using software packages for the R programming language, a standard language used in data analysis. We perform our estimation using functions in baseR, and use several ofthe well-regarded additional packages. 13

28. To establish the resilience of the findings, I also calculated results fi·om a pooled

Poisson regression excluding the feature selection (Lasso) or targeting steps. Theory suggests that the reduced prediction accuracy of this approach should result in lower precision, but similar point estimates as both Poisson regression and tMLE are unbiased and appropriate for data with frequent zero sales. Table 7 presents results from the pooled Poisson regression without targeting. Consistent with theory, the confidence intervals are larger than for tMLE; the estimated effects are however very close, providing additional evidence of the appropriate specification used in the tMLE. The point estimates are +2.00% for New MSEs, and +2.71% for

Catalog MSEs for this approach; the Catalog MSEs are highly significant. These pooled Poisson results also suggest a large +3.37% promotional effect for independent catalog music.

29. If playing music on Pandora has a causal effect on sales, one might then expect to see a dose-response relationship, according to which there is a greater effect on sales in cases where there is more substantial exposure on Pandora. In the context ofthe large variation in sales and spin counts across experiments, a sensible metric for Pandora exposure is the ratio of

Pandora spins to sales. Playing a song 10,000 times a day on Pandora might generate a noticeable lift in sales for a small artist with low baseline sales (say, 10 units per week), while similar spins on Pandora might have a lesser lift ifthe baseline sales are instead 500 units per week. While we did not explicitly randomize exposure (quantity of spins) across experiments, as we used an "on" I "off" treatment, we can explore this relationship by looking at the estimated

13 R packages used include: boot package for studentized confidence intervals, and glmnet package for lasso regression; Poisson regression was carried out using the glm function included in 'base' R.

13 promotional effect along a spectrum of different exposure levels. Figures 3a and 3b present the results of this analysis. We tind strong evidence of increasing promotional effect from increased

Pandora exposure, for both new music and catalog music.

30. One concern would be that the inclusion of many experiments in the New MSEs with very few sales materially affects the results- a variant of the concern that there are too many zeros (which makes linear and log-linear models tremendously challenged in this context).

1 analyze the top 50 sales experiments for Major I Indie New MSEs and top 50 sales experiments for Major I Indie Catalog MSEs to remove the smaller sales experiments. Table 8 presents the results ofthis analysis; these results again reveal a statistically significant promotional effect of

Pandora tor all New MSEs, major label New MSEs, and all variants ofthe Catalog MSEs. The effect tor major label catalog experiments was +2.77%, versus +3.73% for independent label

Catalog MSEs.

31. Given this difference in promotional impact on Majors and Indies for catalog music, I undertook an additional inquiry to value this difference. Even with the lower promotional effect, I anticipated an increased revenue effect for Majors given higher sales, and possibly, higher retail music prices. To adjust for this expectation, my colleagues and I performed the calculation on a per-spin basis. We estimated the incremental number of TEA unit sales, across all DMAs, that Pandora was responsible for, using tMLE as before. We then divided the incremental number of unit sales by twice the number of spins in the control DMAs

(given 50% chance of spinning in the experimental design). We again used bootstrap simulations to compute confidence intervals. I also compute whether there was a difference in the promotional effect between major and independent label recordings; statistically speaking, there was no evidence of a difference.

14 Pre Experiment

32. We performed standard statistical analyses and ran an experimental pilot prior to commencing the experiments to ensure the sound operation of our code and to inform our sample size. Specifically, we performed three checks prior to experimentation. First, we performed

(statistical) power calculations to inform sample size by calculating the variance of sales for 8 researcher-selected songs, all of which were popular on Pandora and had regular sales. 14 No experimental manipulation occurred, rather, we analyzed the variability in the sales data for these songs to predict the number of experiments necessary to find a statistically significant effect of

Pandora. These calculations informed the large scale of the experiments.

33. Second, we performed two "pilot" experiments from April23, 2014, to July 22,

2014, by turning off one popular song and an entire album of medium popularity on Pandora. 15

We ran these pilots to ensure that our code operated as intended for the experiments, but also to ensure the code did not impact any other songs or the general Pandora music recommendation algorithms. Beyond the desired effect (not spinning in treated areas), we found no additional effect. These experiments were not intended to inform the selection of particular music for the experiments or to inform the design of the analyses. 16

34. Once the selection criteria for the New MSEs was set as "all comers" on the pre- selected week, we revised the interface for entering the experimental code into the music

14 The included songs: Get Lucky (Daft Punk), Blurred Lines (Robin Thicke), Wake Me Up (Avicii), Roar (Katy Perry), I'm Diflerent (2 Chainz), Dirt Road Diary (Luke Bryan), V.S.O.P. (K. Michelle) and Out of My League (Fitz and the Tantrums). 15 The song was This is How we Roll (Florida Georgia Line). The album was Kylie Minogue's 13 track album ''Kiss Me Once." 16 Neither the song nor the album manipulated in the pilot meets the inclusion criteria for any experiment we ran. Although we never intended to analyze these pilots for the promotional effect of Pandora, we did so at one point to test our code for computing the tMLE. The results were similar to those observed in the MSEs, but less precise as expected.

15 databases. The prior code, used in the pilots, was not able to scale to hundreds and potentially thousands of experiments. The team created an interface to easily create the random assignment of each experiment to all regions, but also to generate the required code to disable the playing of the music.

35. We ran standard engineering tests to confirm that the code would not undermine the Pandora listening experience.

Results

36. The primary results are in the written statement. I reiterate here that all results are presented for the promotional impact of a performance on Pandora, that is, the estimated effects are the causal effect of spinning on Pandora. Positive estimates mean that music sales are greater when the music is available and spinning on Pandora.17

37. The Music Sales Experiments confirm that the Pandora radio service is net promotional of music sales- that is, music sales are higher when the music plays on Pandora.

As experimental results, they are generalizable from the specific music sampled to other new music and catalog music that spins on Pandora. We also present evidence that increased exposure on Pandora further enhances the promotional effect. The robustness of the study design, and the rigor applied to the analysis, distinguish these results from any previous study of the promotional impact of a radio service.

17 The results are identical to the results of not spinning on Pandora. That is, the estimated effects will be the negative effect of not spinning on Pandora. Effect sizes are necessarily identical. We present the results in the affirmative for ease of interpretation.

16 Capital Reporting Company Day 16 In Re: Determination of Royalty Rates (Public) 05-18-2015

4184 Before the

UNITED STATES COPYRIGHT ROYALTY BOARD

Library of Congress

Washington, D.C.

PUBLIC SESSION

Washington, D.C.

Monday, May 18, 2015

The hearing in the above-entitled matter was convened at 9:12 a.m.

BEFORE COPYRIGHT ROYALTY JUDGES:

SUZANNE M. BARNETT, CHIEF JUDGE

DAVID R. STRICKLER, JUDGE

JESSE FEDER, JUDGE

4225 4227

4226 4324

4325 4327

1 you have. 1 CHIEF JUDGE BARNETT: 5020 is admitted. 2 A. Yes. I hold a Bachelor of Arts Honors 2 (Pandora Exhibit 5020 was admitted into 3 in economics from Simon Fraser University in 3 evidence.) 4 Canada. I hold a Master of Arts in economics from 4 BY MR. MARKS: 5 Queen's University also in Canada. I hold a 5 Q. Dr. McBride, do you work with any 6 Masters of Study in Law from Yale Law School and I 6 particular group or team at Pandora? 7 hold a doctor in economics from Stanford 7 A. I do. 8 University. 8 Q. What is that team called? 9 Q. And do you have any peer-reviewed 9 A. I work within the science team. It is 10 publications in scholarly journals? 10 a group of now around 20 researchers who are tasked 11 A. I do. 11 with designing, executing and analyzing research, 12 Q. Before joining Pandora, what type of 12 primarily using controlled experiments to 13 work did you do? 13 understand the relationships and the causal 14 A. I worked as a professional economist. 14 relationships between features and aspects of our 15 Q. Where did you do that? 15 service. 16 A. For the six previous years, I worked at 16 Q. And your written direct testimony uses 17 Analysis Group, an economic consulting firm, where 17 the term gold standard in reference to Pandora's 18 -- for clients, I undertook numerous analyses, 18 experimental research. 19 economic analyses, where I designed and executed 19 Could you please explain what the gold 20 and analyzed numerous analyses. 20 standard requires? 21 Q. What type of work do you do at Pandora 21 A. Yes. The gold standard requires three 22 now? 22 things: Randomized, controlled and blind. 23 A. Similar work. I am tasked with 23 Q. And what does it mean for an experiment 24 designing, executing and analyzing analyses, 24 to be randomized and why is that important to your 25 including experimental analyses. 25 work?

4326 4328

1 Q. Dr. McBride, did you prepare written 1 A. An experiment is where the researcher 2 direct testimony in connection with this 2 modifies the experience of users and -- with 3 proceeding? 3 randomized, exactly who gets that modified 4 A. Yes, I did. 4 experience, listening, a modified experience on 5 Q. I place before you in the white binder 5 Pandora's service is randomized so we are able to 6 on your desk, a document that has been marked for 6 balance all of the unobservable distinctions 7 identification as Pandora Exhibit 5020. 7 between people on our service so that half are in a 8 A. Okay. 8 treatment group and half are not in a treatment 9 Q. Do you recognize this document? 9 group. 10 A. I do. 10 Q. What does it mean for an experiment to 11 Q. What is it? 11 be controlled and why is that important to your 12 A. It's the testimony I submitted in this 12 work? 13 matter. 13 A. Controlled means you randomize some 14 Q. If you could please turn to the last 14 into treatment and some -- those not into 15 page before the appendices. 15 treatment, into controlled, and that allows you to 16 A. I'm there. 16 compare for randomly selected or assigned 17 Q. Is that your signature? 17 individuals, the experience between those who have 18 A. It is. 18 a slightly modified experience entirely caused by 19 Q. And are the attached figures, tables 19 the researcher, and those in the control who have 20 and appendices materials that you referenced in 20 no difference, and that allows you to compare 21 your written testimony? 21 treatment to control, so that you have a 22 A. Yes. 22 counterfactual estimate of what would otherwise 23 MR. MARKS: I offer Pandora Exhibit 23 have occurred absent only the treatment introduced 24 5020 into evidence. 24 by the researcher. 25 MR. CHOUDHURY: No objection. 25 Q. What does it mean for an experiment to

4329 4331

1 be blind and why is that important to you? 1 report back the requested results to Professor 2 A. The final prong, blind, means that any 2 Shapiro. 3 individual who is in the experiment is unaware of 3 Q. Go to the next slide. 4 whether or not they are in a treatment group or 4 If you could please describe for the 5 control. It's important because if individuals are 5 judges in general terms what Dr. Shapiro asked you 6 aware that they are in an experiment and know which 6 to do. 7 arm they are on, they can modify their behavior in 7 A. For design, he asked that we use AB 8 ways that would hamper the inference, the causal 8 framework so that we use Pandora's normal 9 inference that gold standard experimental research 9 experimental procedures. For the experiment and 10 allows. 10 manipulation, the researcher modified experience, 11 Q. Are you familiar with Pandora's AB 11 what we did was we changed the share of spins 12 framework for conducting experiments? 12 available for an outcome that would be available to 13 A. Yes, I am. 13 the listeners in the treatment groups based on the 14 Q. What is the AB framework? 14 ownership of those sound recordings and then get 15 A. The AB framework is Pandora's internal 15 four specific requests in deflection for steering, 16 name for the architecture that modifies our base 16 changing the spin share. 17 code, that allows us to produce gold standard 17 He asked for negative 30 percent, 18 research. The randomized controlled is blind by 18 negative 15 percent, plus 15 and plus 30, so four 19 randomly assigning some of our listeners to a 19 treatment objectives for the spin share deflection. 20 treatment group that have a modified listening 20 JUDGE STRICKLER: Do you know how 21 experiencing and the controlled group that allows 21 Dr. Shapiro selected the four, negative -- minus 22 the counterfactual comparison for causal analysis. 22 15, minus 30 and plus 15 and plus 30? 23 Q. Have you conducted any sets of 23 THE WITNESS: I don't, Your Honor. 24 experiments in connection with this proceeding? 24 JUDGE STRICKLER: I guess we will have 25 A. Yes, I have. 25 to ask him.

4330 4332

1 Q. So could you just briefly identify what 1 THE WITNESS: And then he applied -- he 2 sets of experiments that you conducted in 2 requested at the directions -- Professor Shapiro 3 connection with this proceeding? 3 requested that those four treatments or steering 4 A. Yes. Two sets of experiments. What I 4 targets be applied to the three major ownership 5 refer to as the steering experiments and also the 5 groups, UMG, which -- Universal, Sony and then 6 second set are what I refer to as the music sales 6 Warner Music Group or Warner. 7 experiments. 7 BY MR. MARKS: 8 Q. Let's start with the steering 8 Q. And what was the time frame of the 9 experiments. 9 steering experiments? 10 What questions were you attempting to 10 A. They ran from June 4, 2014 through 11 answer with that research? 11 September 3, 2014. 12 A. That research tried to analyze the 12 Q. And are additional details of your 13 effect of changing or modifying the spins of 13 methodology set forth in your written direct 14 listeners in treatment group based on the ownership 14 testimony? 15 of those sound recordings. 15 A. Yes, they are. 16 Q. Did Professor Shapiro have a role in 16 Q. Turn to the next slide. 17 designing the steering experiments? 17 Is this slide, Figure 1, from your 18 A. Yes, he did. 18 written direct testimony? 19 Q. What was Professor Shapiro's role? 19 A. Yes, it is. 20 A. Professor Shapiro provided us detailed 20 Q. What does this show? 21 instructions on the implementation and design of 21 A. This figure shows the difference 22 this set of experiments. 22 between the actual spin share in each of the 12 23 Q. And what was your role? 23 treatment groups versus control. So the previous 24 A. My role was to take those instructions 24 slide mentioned that Professor Shapiro directed us 25 and implement them in the AB framework and then 25 to modify the spin share by plus 30, plus 15, which

4333 4335

1 are the values -- I'm sorry, Your Honor, the left Y 1 CHIEF JUDGE BARNETT: Please be seated. 2 axis label is not showing there but the left -- the 2 Mr. Marks, shall we have another run at 3 Y axis is the change in spins by music groups, so 3 it? 4 the steering quantity in percentage terms or 4 MR. MARKS: Please, Your Honor. 5 percentage point terms, and then -- so the 5 And I would just want to note for the 6 30 percentage terms is plus 30 percent, plus 15 6 record -- I realize that we were in closed session 7 percent, minus 15 and minus 30. 7 for the portion of Dr. McBride's testimony that was 8 What this shows is that our engineers 8 conducted before lunch. That does not need to be 9 were able to modify the actual spin distribution of 9 restricted. 10 listeners in those treatment groups to closely 10 And so for purposes of the final 11 approximate the requested target values that 11 transcript, that can be part of the public record. 12 Professor Shapiro provided. 12 CHIEF JUDGE BARNETT: Great. Thank 13 Q. It looks like there is an aberrational 13 you. 14 point in the middle of August in Figure 1. 14 And for those of you who missed it, too 15 What is the cause of that? 15 bad. 16 A. There was a brief malfunction in the AB 16 Go ahead, Mr. Marks. 17 framework which prevented the spin shares closely 17 BY MR. MARKS: 18 matching the actual requested targeted shares. 18 Q. Dr. McBride, you'll recall that right 19 Q. And in your written direct testimony, 19 before the break we were talking about the steering 20 do you present the results both including that week 20 experiments and just about to discuss what the 21 and then separately excluding that week? 21 results of the steering experiments were. 22 A. Yes, I do. 22 Do you recall that? 23 Q. What is the metric that you used to 23 A. There was quite a cliffhanger. Yes. 24 measure the effect on the listening experience? 24 So behind the judges are the primary 25 A. Average hours per listeners. 25 findings. The primary -- the first primary finding

4334 4336

1 Q. Why did you use that metric? 1 was that -- 2 A. It is Pandora's standard measure of 2 Q. Let me just stop you before you get 3 listening behavior on the service for all business 3 into it. 4 -- for business purposes. 4 Is this slide a simplified version of 5 Q. And what do the results of the steering 5 Table 1 from your written direct testimony? 6 experiments show? 6 A. That's correct. Yes. 7 MR. POMERANTZ: Your Honor, I'm sorry, 7 Q. Okay. And does this report the results 8 I had a request from our client. If this is 8 of the steering experiments? 9 public, we would just want to let the public come 9 A. Yes, it does. 10 in. 10 Q. And what did you fund? 11 MR. MARKS: I apologize. I didn't 11 A. We found that we were able to steer 12 realize we were in closed session. The entire 12 either plus or minus 15 percent without 13 direct examine will be public, I'm sorry. 13 statistically significant impact on listening. No 14 JUDGE STRICKLER: Shall we start all 14 listener effect. 15 over again? 15 We found in two instances a negative 30 16 MR. POMERANTZ: No, we object to that. 16 percent Universal and a plus 30 for Sony. Small -- 17 THE WITNESS: This particular table is 17 small average effect over the period on listening 18 restricted in my written direct testimony. 18 retention. 19 MR. MARKS: That's fine. It's fine to 19 Q. And are you aware that Dr. Rubinfeld 20 disclose. They can come in. 20 critiques certain aspects of your written direct 21 CHIEF JUDGE BARNETT: It's noon. We're 21 testimony in his written rebuttal testimony? 22 going to take our noon break. Thank you. 22 A. I am aware of that. 23 You provided us a convenient breaking 23 Q. And what is your response to his 24 point. We will drop back ten. 24 contention that the loss of listenership tends to 25 (A short recess was taken.) 25 grow as the steering experiments continue and his

4337 4339

1 suggestion that the average effects you observed 1 available. 2 may understate the loss of listenership over a 2 We had a second set of experiments, 3 longer period? 3 which I refer to as the catalog music sales 4 A. So this exhibit here is Professor 4 experiments. And these included music that had 5 Rubinfeld's own Exhibit 13-A. I have added the 5 been long available on Pandora and numerous other 6 dotted line. And these represent data that I 6 places, in some case years and decades. 7 included in my written direct testimony. He has 7 And so in that experiment, again, we 8 added the black simple average. 8 took that music and randomly prevented our -- some 9 And my interpretation of all of these 9 of our listeners in the treatment group not to be 10 data, and from my experience in reviewing these 10 able to access that music. 11 data, is that there is no trend or any evidence of 11 So then we have a separation between 12 declining listenership over the period of the 12 the treated group -- as we defined it, for 13 experiment after the midpoint. So I would -- I 13 simplicity -- it doesn't change the results, but we 14 would disagree with that point. 14 define treatment as having the music on Pandora and 15 And that -- so logically I have 15 the control is not having the music on Pandora. 16 additional problems with this. First, he presents 16 So we had a causal -- we had a 17 a simple average, which is an average of an 17 separation produced by this experimental 18 average, which is not an appropriate method when 18 manipulation. 19 you have different portions of the sample in each 19 Q. And why did you distinguish between new 20 of the groups. 20 releases or new music and catalog music? 21 And then, finally, he doesn't disclose 21 A. We thought there were distinctions in 22 or acknowledge on this exhibit that he's combining 22 expected sales of new music versus catalog and also 23 numerous examples that were insignificant 23 potential differences in the general availability, 24 statistically with significant results and not 24 marketing of new music versus catalog. 25 reporting that. 25 Q. How does Pandora obtain geographic

4338 4340

1 Q. And do you think that's a correct 1 information about the location of its listeners? 2 methodological approach? 2 A. Yes. Upon registration, our U.S. 3 A. I find it irregular. 3 listeners are required to provide us a ZIP code. 4 Q. All right. So turning back to the sets 4 And that has been case for at least seven years. 5 of experiments that you mentioned, there's another 5 Those listeners, upon discretion, sometimes at our 6 set of experiments that you conducted; is that 6 prompt, are encouraged to update their ZIP code as 7 right? 7 well. 8 A. That is correct. 8 Q. And do, in fact, Pandora listeners 9 Q. And what were the music sales 9 update their ZIP codes from time to time? 10 experiments designed to test? 10 A. Yes, they do. 11 A. The purpose of the music sales 11 Q. Why did you perform the analysis by 12 experiments was to test Pandora's causal impact on 12 region? 13 whether Pandora promotes music sales or substitutes 13 A. Yes. So with the steering experiments, 14 for music sales on net. 14 we had individual level outcome. We had hours of 15 Q. And if you could -- how did you go 15 listening for every listener on Pandora. 16 about answering that question? 16 With the music sales experiment, the 17 A. Yes. So in that experiment, the 17 equivalent would be we would need individual 18 manipulation imposed by the researcher, myself and 18 listeners' purchase data. And at Pandora we -- we 19 my colleagues, was to disable, prevent from playing 19 aren't -- we do not have access to individual 20 on Pandora, particular music. 20 listener purchase data. 21 We had two groups of experiments. We 21 As a consequence, we had to go find a 22 had new music, which was music that had never 22 source that had the most granular music sales data 23 played on Pandora. So when it was about to start 23 that we could find. 24 playing for all of our listeners, we intercepted it 24 Q. And where did Pandora obtain the weekly 25 in the engineering code and randomized where it was 25 sales data that it used?

4341 4343

1 A. We used Nielsen SoundScan weekly data. 1 randomized. This is the actual availability of one 2 Q. Why did you use Nielsen SoundScan? 2 of the experiments included in our research. The 3 A. We understood in discussion that this 3 dark green are the randomly assigned areas where 4 was perceived to be the best, most complete, 4 the music would be available for one particular 5 comprehensive data available at a granular level of 5 experiment. And all of the light green areas were 6 music sales for physical albums and digital 6 the randomly assigned areas where the music was not 7 downloads in the United States. 7 available. 8 Q. And did the Nielsen data that you used 8 So at a high level, we are comparing 9 track both physical CD sales and digital downloads? 9 the sales in the areas that were randomly 10 A. That's correct. 10 available -- sorry -- where the music was randomly 11 Q. And what is the geographical unit by 11 available, the total sales in the green areas, 12 which Nielsen tracks sales? 12 compared to the sales in the light green areas. 13 A. The most granular region that Nielsen 13 And notably, our experimental design 14 will make their data available on music purchases 14 had separate randomization for all of our 15 is called a designated market area, which in my -- 15 experiments. In outcome we had over 1,200 of these 16 which is also sometimes referred to as a DMA. 16 music availability experiments to explore our 17 There are -- in our study, we had 228 of these. 17 causal impact on music sales. 18 Q. And were the experiments done at the 18 And so, for any of the areas that are 19 level of individual tracks, albums or some 19 dark green here, on average they would be light 20 combination? 20 green half the time and dark green so that each of 21 A. For the new music experiments, we 21 these were randomized so that we capture and 22 manipulated the availability of an album. So all 22 control for any regional differences in, for 23 tracks on an album were either available to a 23 example, genre preference. 24 listener to play at our normal Pandora algorithm or 24 Q. And is there -- do you have any reason 25 unavailable, and none of the tracks on that album 25 to believe that the starting week that you used for

4342 4344

1 would be available. 1 the music sales experiments was exceptional in some 2 For catalog music sales experiment, we 2 way? 3 used songs. We found that albums of catalog music, 3 A. I have no reason for that, no. 4 because of golden his, greatest his, and remastered 4 Q. And do you have any reason to believe 5 versions, sales of catalog music were too 5 that the ending week was exceptional in some way? 6 fragmented across albums to be able to analyze at 6 A. No. No, I don't. I would also 7 the album level. 7 actually note that, because we ran so many 8 Q. Does your written direct testimony 8 experiments, we also had staggered start periods 9 describe in greater detail how you determined which 9 and staggered end periods so that they weren't 10 new music and which catalog music to turn off in 10 entirely overlapping. 11 your experiment? 11 So that also was able to address any 12 A. Yes, it does. 12 concerns that could be had regarding exceptional 13 Q. Let me -- 13 natures, the periods that we ran the experiments. 14 A. Sorry to -- 14 Q. And are the additional details of your 15 Q. There we go. 15 methodology for the music sales experiments set 16 If you could just explain what this 16 forth in your written direct testimony? 17 slide reflects. 17 A. Yes, they are. 18 A. Yes. This is a map of the United 18 Q. Did the music sales experiments reflect 19 States. All of the internal boundaries represent 19 an average promotional effect? 20 the boundaries of Nielsen DMAs -- sorry -- 20 A. Yes. 21 designated market areas. So all of those internal 21 Q. And is that what's shown on this slide? 22 boundaries are each of the 228 designated market 22 A. This slide here -- I apologize. It 23 areas. 23 says Table 4. This is actually Table 3 from my 24 The dark color -- I'm sorry. 24 written direct testimony. 25 The dark color and light color were 25 And what this shows is that, on

4345 4347

1 average, music being available on Pandora, Pandora 1 For new music we had 814 different 2 causes that music -- or sales of that music to rise 2 music albums that we were able to manipulate the 3 by 2.3 percent on average for new music. And 3 availability and then were able to assess the 4 because the music is available on Pandora, we cause 4 sales. 5 catalog music sales on average to be 2.66 percent 5 And what these broadly show is that, as 6 higher. 6 you look at areas -- or for experiments by which 7 The confidence interval, the values in 7 Pandora represents more exposure, then we see a 8 the second column, show the range where the -- the 8 higher promotional effect. 9 range by which we can be very confident the true 9 And so the scale along the X axis is 10 causal estimate of how much Pandora in this case 10 it's a threshold that says, you know, there are at 11 promotes music sales-wise. Both of these are 11 least one spin per sale. 12 highly statistically significant. 12 At a high level, when you think about 13 Q. And did you also calculate the results 13 it, if a particular song has 10,000 spins on 14 separately by label type? 14 Pandora, that could be a lot or a little exposure. 15 A. Yes, we did. This figure here presents 15 So if we find that that particular 16 a simplified version of Table 4 in my written 16 sound recording had two sales, 10,000 spins on 17 direct testimony. 17 Pandora might represent substantial exposure. 18 Starting with the first column -- oh, I 18 Conversely, if that album or sound 19 apologize, Your Honors. The -- some of the letters 19 recording had 50,000 sales, then I would see that 20 are cut off. It says "New Music" -- it says "New 20 10,000 spins on Pandora is not substantial 21 Music" in the top table. 21 exposure. 22 So we find for new music -- when 22 The way we thought about how much does 23 looking by the ownership of the music that was 23 Pandora represent share of exposure is by the ratio 24 included in the experiments, we found that Pandora 24 on the bottom, which is the ratio of spins on 25 causes music sales for music released on major 25 Pandora for the music divided by sales that we

4346 4348

1 labels to rise by a higher amount 2.82 percent, a 1 observed from Nielsen. So one measure of relative 2 very highly significant figure. 2 exposure. 3 You'll note that I did not bold the 3 And what we find is, as Pandora 4 estimate -- the positive estimate of 0.62 for new 4 represents more and more relative exposure, we see 5 music on other labels. One of the -- that estimate 5 higher promotional effects. So in the right panel, 6 wasn't statistically significant. 6 you'll note that the first couple estimates, 7 For catalog music sales, when you 7 because they're gray and they cover -- across the 8 separate by the ownership of the music, we again 8 zero dash line, it means they're not statistically 9 see extremely -- or positive promotional estimates 9 significant. 10 that Pandora causes increase in these sales at very 10 But when you look at the large share of 11 high levels of statistical significance. 11 experiments that were new music released on other 12 So we are confident that these are 12 labels, we actually observed extremely strong, 13 promotional effects. 13 highly statistically significant promotional 14 Q. Did you find any evidence that the 14 effect. 15 extent of exposure on Pandora for a given sound 15 And the last value on your right, left, 16 recording impacts the size of the promotional 16 depending upon how you see it, we estimate a 25 17 effect? 17 percent net promotional effect of Pandora for the 18 A. Yes, we did find that. 18 subset of experiments where Pandora represents 200 19 And this next figure will need 19 spins per one sale. 20 explanation. Broadly it's Figure 3(a) in my 20 Q. And this slide is the promotional 21 written direct testimony. It includes three 21 impact by exposure for new music. 22 panels. 22 A. Yes. 23 Focusing first on the left panel, these 23 Q. Did you also have a figure of 24 are estimated promotional effects for new music for 24 promotional impact by exposure for catalog music? 25 different subsets of the experiments. 25 A. Yes, we did. This figure that's now

4349 4351

1 shown as Figure 3(b) from my written direct 1 in any way affected by this criticism because all 2 testimony. And the same pattern holds. 2 of the music included in the catalog music sales 3 As you look at the subset of 3 had sales. So that argument is simply not 4 experiments where Pandora represented a larger and 4 applicable to all of these promotional estimates. 5 larger share of the exposure, how much the 5 Q. What other responses do you have to the 6 listeners in that region were able to access or 6 criticism by Dr. Blackburn that you should not have 7 experience that music, Pandora continuously, we 7 excluded experiments with zero sales analysis? 8 estimate, is a higher promotional effect on the 8 A. Most notably, he criticized this 9 sales of that same music. 9 research for excluding experiments for which the 10 Q. Are you aware that Dr. Blackburn has 10 vast majority have no data available. So -- 11 critiqued your music sales experiments? 11 Q. Does that mean there are no sales? 12 A. Yes, I am. 12 A. We don't know. No one is able to know 13 Q. Are you aware that Dr. Blackburn has 13 whether or not this music had sales. For the vast 14 criticized you for not using data related to the 14 majority of what he's criticizing, neither he nor I 15 buy button on Pandora's Website to measure 15 have any access to data to say whether or not there 16 promotional impact? 16 were sales and Pandora promoted them or there were 17 A. Yes. 17 sales and Pandora suppressed them. It's simply 18 Q. How do you respond to that criticism? 18 impossible as a researcher to know. 19 A. My first response is it's observational 19 An additional reason that I would note 20 insight. It's correlation; it's not causation. 20 why I think it's a misplaced criticism is the 21 All of the results that I've presented today are 21 strong promotional estimates that we find for new 22 causal estimates. They come from an experiment. 22 music where there are some experiments with zero 23 Thinking about the buy button observational, 23 sales depends on 28.7 million units of sales. And 24 that -- that's simply correlation. 24 Professor -- or Dr. Blackburn is focusing on a 25 The second point is it presumes that 25 couple of cases where some of them had zero.

4350 4352

1 the only way a radio services would be able to 1 Even if they had some units of sales 2 promote music sales is through a direct channel of 2 and, for argument, if Pandora substituted for all 3 this button. I find that unconvincing, in part 3 of them, it would not materially affect net 4 because terrestrial radio does not have a buy 4 promotional effects that I show in my report. 5 button, and my general sense is that the industry 5 Q. For some of the experiments that 6 accepts that terrestrial radio is promotional even 6 Dr. Blackburn criticizes you for excluding, were 7 absent any form of direct channel for purchase. 7 there spins on Pandora? 8 And then finally, I will note that the 8 A. Right. Another thing that he -- there 9 availability of the data for use of the buy button 9 were not. For seven of the experiments that he 10 on our service is not broadly available, in part 10 criticizes us for excluding, the music that was 11 of -- because of privacy and security reasons for 11 included in the experiments never played on our 12 our service and our listeners. 12 service. 13 Q. Did Dr. Blackburn assume that the buy 13 Of course we could not be promotional 14 button data would be more available to you than 14 for music that we did not use. So accordingly, 15 they are? 15 reasonably as researchers, we did not include that 16 A. That's correct. 16 in our analysis. 17 Q. And are you aware that Dr. Blackburn 17 Q. Are you aware that Dr. Blackburn has 18 has criticized you for excluding experiments with 18 criticized your experiments because of potential 19 zero sales from your analysis? 19 inaccuracies in the data Pandora maintains about 20 A. Yes. 20 the geographic location of its listeners? 21 Q. Does that criticism affect in any way 21 A. Yes. 22 your catalog music sales experiments? 22 Q. How do you respond to that criticism? 23 A. No. All of the results that are 23 A. It makes a profound error regarding 24 currently on the image from Figure 3(b) for the 24 normal interpretation of experimental methods. 25 catalog music sales experiments, none of those are 25 Conceptually we're comparing the

4353 4355

1 outcomes for the region that's treated to the 1 Q. As you know, I represent SoundExchange. 2 outcomes that the regions are controlled. 2 A. Yes. I know that. 3 If you accept, as Professor Blackburn 3 Q. So first of all, just by way of 4 suggests, that there is some inaccuracies in our 4 background, the two studies you report in your 5 ZIP code data, what we're actually estimating the 5 testimony, you've been working on them since your 6 promotional effect on is a smaller spread between 6 first day at Pandora, correct? 7 the treated group and the control. 7 A. Very close. One I started -- on my 8 So if anything, if you were to accept 8 first day I started thinking about. The other one, 9 his claim, the true promotional effect that I -- is 9 my first week. 10 actually higher, is larger than I report in my 10 Q. And when you interviewed for your 11 estimate. It's the very common finding in 11 position at Pandora, you interviewed with Mr. 12 experimental methods. It's called dilution or 12 Harrison, right? 13 watering down of a treatment effect. 13 A. That's correct. 14 And so the primary implication is in no 14 Q. And he's not in the science department 15 way to affect the validity of the results but 15 of Pandora, right? 16 instead actually to suggest helpfully that the 16 A. He is not. 17 promotional effect that we find is, in fact, higher 17 Q. He is in-house counsel at Pandora, 18 than number that I report in my testimony. 18 right? 19 Q. And are you aware that Dr. Blackburn 19 A. He is that. He is also vice president 20 has singled out the 90210 area code -- excuse me -- 20 of business affairs. 21 ZIP code as having particular inaccuracies as a 21 Q. And looking back to your first day, 22 result of Pandora users self-reporting that they 22 after you completed your human resource paperwork, 23 live in Beverly Hills, California? 23 what was the first thing you did at Pandora? 24 A. Yes. 24 A. After that I met with my boss for an 25 Q. How do you respond to Dr. Blackburn's 25 orientation meeting.

4354 4356

1 criticism on the 90210 Zip code point? 1 Q. Right. 2 A. I have several thoughts, I think. The 2 And that's Mr. Bieschke? 3 first is this is a single region out of 228. And 3 A. Yes, it is. 4 so it is unlikely to materially affect the results. 4 Q. And you discussed with him there the 5 Second, and I think importantly, the 5 music sales experiments, correct? 6 experimental design of our research randomizes each 6 A. That is correct. 7 region, sometimes to treatment, to sometimes 7 Q. And between the time you were hired by 8 control. So contrary to any assumption in this 8 Pandora and the time you submitted your written 9 material things, sometimes it could be driving 9 direct testimony, is it correct that the only 10 sales; sometimes it could not. It's simply not -- 10 studies or experiments you worked on were the two 11 it's balanced because of the experimental design. 11 sets of experiments in your testimony plus a study 12 And then finally, to the previous 12 on how ad load affects listener retention? 13 point, this is a -- this could be an example of 13 A. That's not true. 14 inaccuracies which suggest a spreading, the 14 Q. Okay. What was the other study or 15 difference between treatment and control, so that, 15 experiment? 16 if you were to accept his argument, in fact, the 16 A. The other two were ad quality and the 17 true promotional effect would be higher than I 17 effect of ad repetition on listening patterns. I 18 report in my testimony. 18 did not work on the ad load experiment. 19 MR. MARKS: No further questions, Your 19 Q. And all of those were at the direction 20 Honor. 20 of counsel, correct? 21 CROSS-EXAMINATION BY COUNSEL FOR 21 A. That's not correct. SOUNDEXCHANGE 22 Q. Which wasn't? 22 BY MR. CHOUDHURY: 23 A. The two that I added were not at the 23 Q. Good afternoon, Dr. McBride. Good to 24 direction of counsel. 24 see you again. 25 Q. Okay. And at the time you were deposed 25 A. Thank you. Likewise.

4357 4359

1 though, all of the studies or experiments you had 1 Q. And when you worked at Analysis Group, 2 executed since coming to Pandora were at the 2 none of your projects there involved the music 3 direction of counsel, correct? 3 industry, correct? 4 A. Sir, could you ask the question again? 4 A. That's my recollection, yes. 5 Q. Sure. 5 Q. You never published an academic paper 6 At the time you were deposed -- 6 on the music industry, correct? 7 A. Yes. 7 A. That's correct. 8 Q. And as you might imagine, I'm harkening 8 Q. You never published an analysis of 9 back to our deposition. 9 promotion or substitution, right? 10 A. Yes. 10 A. That's correct. 11 Q. At the time you were deposed, all of 11 Q. And besides this experiment, you've 12 the studies or experiments you had executed since 12 never studied the reason that consumers purchase 13 coming to Pandora were at the direction of counsel, 13 music, correct? 14 correct? 14 A. That's correct. 15 A. I don't think exactly correct. The 15 Q. And besides this experiment, you've not 16 only experiments I had run as of our deposition 16 studied the reasons for increases or decreases in 17 were at the direction of counsel. 17 sales across the music industry, right? 18 Q. And that was counsel for this 18 A. That's correct. 19 litigation, right? 19 Q. And that includes regional variations, 20 A. Yes. 20 correct? 21 Q. At the time I took your deposition, 21 A. It would, yes. 22 it's true that you had no research to report that 22 Q. And that includes seasonal variations, 23 was not at the direction of counsel, correct? 23 too, right? 24 A. That's correct. 24 A. I agree. 25 JUDGE STRICKLER: Just so it's clear, 25 Q. Okay. So before we get into the

4358 4360

1 when you say you had no research to report that 1 experiments themselves, I just want to understand 2 wasn't at the direction of counsel, in terms of 2 the scope of the testimony you're presenting to the 3 topic or in terms of results? 3 judges. 4 THE WITNESS: In terms of results. So 4 So your testimony isn't studying 5 the two studies that I reminded counsel about were 5 whether Pandora is a substitute or promotional for 6 exploratory studies. And when I became more 6 revenue from interactive streaming services, right? 7 involved in executing these experiments, those were 7 A. I agree. 8 performed and finished by colleagues on the science 8 Q. And you agree that your experiments 9 team. 9 don't say anything about how promotion interactive 10 JUDGE STRICKLER: Was counsel involved 10 services are, right? 11 in managing the results? 11 A. I agree with that. 12 THE WITNESS: No. 12 Q. Or terrestrial radio? 13 JUDGE STRICKLER: That was my question. 13 A. I agree. And my studies weren't 14 THE WITNESS: Sorry, Your Honor. 14 designed to test any of those. 15 BY MR. CHOUDHURY: 15 Q. Or simulcasters? 16 Q. I'd like to begin by discussing the 16 A. I agree. 17 music sales experiments. 17 Q. Or any other Webcasters besides 18 Your assignment there was to test 18 Pandora, correct? 19 whether performances of sound recordings on Pandora 19 A. I agree. 20 have a positive or negative impact on sales of the 20 Q. And you'd also agree that your 21 sound recordings, correct? 21 experiments don't say anything about how 22 A. That's correct. 22 promotional Pandora is as compared to an 23 Q. And you never worked in the music 23 interactive service, right? 24 industry, right? 24 A. Studies don't test that. I agree. 25 A. I have not. 25 Q. Or as compared to terrestrial radio,

4361 4363

1 right? 1 Q. Okay. And, in fact -- well, did anyone 2 A. I agree with that. 2 -- did counsel tell you that you couldn't 3 Q. Or as compared to simulcasters? 3 investigate that assumption? 4 A. I agree. 4 A. They did not. 5 Q. Or as compared to any other Webcaster, 5 Q. Okay. So you were free to, correct? 6 right? 6 A. Yes. 7 A. I agree. 7 Q. But you chose not to. 8 Q. And you'd also agree that your 8 A. I don't know if it's feasible, but I -- 9 experiments don't say anything about whether 9 I did not do that. 10 Pandora promotes subscription revenue to any other 10 Q. And that would mean that your 11 digital service, right? 11 experiment doesn't tell us anything about whether 12 A. We didn't have those data. So 12 Pandora increases total music sales for the 13 certainly we weren't able the study that, no. 13 industry, right? 14 Q. Or to concert revenue to the music 14 A. I don't fully agree with that 15 industry, correct? 15 interpretation. The study results provide insights 16 A. I'm working on it. 16 on that broader interpretation. This is a 17 Q. But that's not in your testimony. 17 necessary assumption for the exact numbers to 18 A. It is not. 18 apply. 19 Q. And, in fact, your experiments don't 19 Q. Well, Dr. McBride, this isn't the first 20 say anything about any revenue source besides sales 20 time you've been asked that question, is it? 21 of CDs and downloads, right? 21 A. It's not. 22 A. And the direct royalties that we pay 22 Q. Okay. So you, in fact, provided 23 for the music that we're playing in the 23 deposition testimony in the BMI rate setting 24 experiments, yes. 24 proceeding, correct? 25 Q. Sure. 25 A. I did.

4362 4364

1 So you tested the royalties in your 1 Q. And you -- in that proceeding, we're 2 experiments? 2 talking about these same music sales experiments, 3 A. The royalties certainly went down when 3 correct? 4 we turned the music off. 4 A. That's correct. 5 Q. I see. 5 Q. So I'd ask you to turn to Tab 3. If 6 So that's what you mean when you say 6 you can turn to Page 14 of that document. 7 you tested the royalties? 7 JUDGE STRICKLER: That'd big Page 14? 8 A. I didn't mean to suggest I tested the 8 MR. CHOUDHURY: It'd be big Page 14. 9 royalties. 9 BY MR. CHOUDHURY: 10 Q. Okay. Now, you acknowledge in Footnote 10 Q. If you could look at the -- 11 27 of your testimony that to draw a broader 11 MR. MARKS: I just want the record to 12 interpretation of Pandora's effect on aggregate 12 be clear. Counsel identified -- he referred to 13 sales -- oh, and, please, by all means, take a 13 deposition testimony. Tab 3 is not deposition 14 look. This is on Page 17. 14 testimony. 15 A. I'm good. 15 MR. CHOUDHURY: Thank you for that 16 Q. Okay. So you suggest that to draw a 16 correction. 17 broader interpretation of Pandora's effect on 17 BY MR. CHOUDHURY: 18 aggregate sales of manipulated music, you would 18 Q. In fact, this is your trial testimony 19 need to assume that sales of a specific sound 19 from the BMI hearing; isn't that correct? 20 recording are not affected by manipulation of other 20 A. You'll have to give my a second. Yes. 21 music in the experiment, correct? 21 It appears to be. 22 A. That is what it says, yes. 22 Q. Okay. So if you look at the -- on 23 Q. And you did nothing to investigate 23 page -- big Page 14, you see Page 1922, correct? 24 whether that assumption was correct, right? 24 A. Yes, I do. 25 A. I did not. 25 Q. And you see -- you're -- and this is --

4365 4367

1 this is your testimony, correct? 1 promotional effect for the music study does provide 2 CHIEF JUDGE BARNETT: Are you asking 2 insight on the general promotional effect." 3 this -- 3 I would say that directly pertains to 4 MR. CHOUDHURY: I'm just using it 4 your interpretation of my Footnote 27. 5 for -- 5 JUDGE STRICKLER: That begs the 6 CHIEF JUDGE BARNETT: To refresh 6 question: What insights? 7 recollection? 7 THE WITNESS: Right. So I would assert 8 MR. CHOUDHURY: -- impeachment and to 8 that it's the only -- it is, to my knowledge, the 9 refresh his recollection. 9 only gold standard evidence of promotional or 10 CHIEF JUDGE BARNETT: Okay. Go ahead. 10 substitutional effect of radio. And it suggests 11 THE WITNESS: Yes. That is my 11 that, because music is on Pandora, not correlation 12 testimony. 12 but cause, Pandora causes sales to go up. 13 CHIEF JUDGE BARNETT: I'm sorry. Just 13 JUDGE STRICKLER: Sales of a specific 14 so the record is clear, this is testimony that was 14 song that's been manipulated. 15 given in the U.S. District Court for the Southern 15 THE WITNESS: That's true. 16 District of New York on March 4, 2015, before the 16 The impact of the response from the 17 Honorable Louis Stanton. 17 exposure effect, where I show figures 3(a) and 18 MR. CHOUDHURY: Thank you, Your Honor. 18 3(b), say increased exposure on Pandora causes or 19 BY MR. CHOUDHURY: 19 is associated with higher promotional effect, which 20 Q. So if you could look at Line 11. The 20 my -- 21 question you're asked is: "But you would agree 21 JUDGE STRICKLER: With regard to that 22 that your experiment doesn't say anything about 22 specific song. 23 whether Pandora increases total music sales, 23 THE WITNESS: True. But because it's 24 right?" 24 inducing higher sales of that song, the core issue 25 Your answer there is: "The results are 25 is where are those sales coming from? And it is

4366 4368

1 specific to the music that is included in the 1 unknown broadly. But some of those, because of the 2 experiments, which I agree is not all of the music 2 response from the increased exposure effect, it 3 that is available on Pandora." 3 suggests that it does induce pure promotional 4 And then the question is: "Right. If 4 effect of increased sales overall. 5 Pandora didn't exist, you're not testifying that 5 BY MR. CHOUDHURY: 6 your music sales say anything about whether there 6 Q. Now, the answer you just gave, that is 7 would be more music sales or less music sales in 7 specific to the song that is being manipulated, 8 the absence of Pandora, right?" 8 correct? 9 And you answer: "It's somewhere in 9 A. Is it not. 10 between. It provides insights on that, but it does 10 Q. In fact, that's what your Footnote 27 11 not directly address the question." 11 assumption is about, right? 12 A. I said -- 12 That if we wanted to know Pandora's 13 Q. Is that correct? 13 effect on aggregate sales of manipulated music, we 14 A. -- exactly that. I would also direct 14 need to assume that the sales of a specific sound 15 you to Line 1, which says: "That is certainly true 15 recording are not affected by the manipulation of 16 prior to the period of experimentation. But the 16 other music in the experiments, correct? 17 promotional effect for the music study does provide 17 A. That's not correct. 18 insight on the general promotional effect." 18 For -- what Footnote 27 says is, for 19 CHIEF JUDGE BARNETT: Dr. McBride, I'm 19 the exact value reported in my study derived from 20 going to ask you to read that again at normal 20 admittedly a sample of it -- of music, for that 21 speed. 21 exact value to be the aggregate promotional value 22 THE WITNESS: I apologize. So I was 22 does require this assumption. 23 reading you Line 1 from Page 1922, which says, 23 I'm not asserting that it is the true 24 answer to a question: "That is certainly true 24 promotional effect overall. I am noting that it 25 prior to the period of experimentation. But the 25 provides insights on the interpretation of is

4369 4371

1 Pandora promotional in general on music sales. 1 A. That's correct. 2 JUDGE STRICKLER: So you don't know 2 Q. And then you compared the sales for 3 whether the net promotional effect could be lower 3 where DMAs were being -- where the song was being 4 because it might just -- the sales of the 4 spun to where it wasn't, right? 5 manipulated song may be offset to some extent by 5 A. Correct. 6 fewer sales of a nonmanipulated song. 6 Q. And that's because -- and you didn't 7 THE WITNESS: I agree. 7 have listener level data, right? 8 JUDGE STRICKLER: And is it -- your 8 A. We wanted it, but we don't have it. 9 study doesn't address whether or not the reduction 9 Q. You don't have it, so you couldn't use 10 in sales in the nonmanipulated song might swamp the 10 it. 11 promotional effect on -- of the manipulated song. 11 A. Could not use it. 12 THE WITNESS: I agree, Your Honor. It 12 Q. Right. 13 doesn't directly test that. 13 So consumers inside a DMA who are not 14 My interpretation of the dose response 14 subscribers to Pandora -- I mean let's say that I 15 effect, the increased exposure of music on Pandora 15 am not a Pandora subscriber. I'm in one of your 16 being associated with higher promotional effect. 16 DMAs. I buy music. 17 So if you're producing more exposure on this one 17 Are my purchases of music still 18 but less on this, that does produce a difference in 18 counted? 19 exposure which would be suggestive of an overall 19 A. Yes. 20 effect on informing music sales. 20 Q. Okay. And consumers who are Pandora 21 Because, if you're giving this 21 users but didn't use Pandora during the weeks in 22 particular song ten more spins, but you're giving 22 which your experiments were conducted, were their 23 one extra spin to ten other songs, there's a 23 purchases counted as well? 24 difference in quantity of exposure. And my 24 A. Yes. 25 interpretation is that that provides insights on 25 Q. Okay. And consumers who are Pandora

4370 4372

1 the effect broadly. 1 users would have never found the songs, either 2 But -- so my meaning in Footnote 27 is 2 because they had -- just have different tastes, 3 simply, for the exact values I report to be Pandora 3 they would have never used the songs manipulated. 4 causes a broader music effect of the exact figures, 4 Were their purchases counted? 5 requires this assumption. And I did not test that 5 A. Yes. 6 assumption. I think it provides insight. 6 Q. Okay. And it is true, right, that, 7 JUDGE STRICKLER: Thank you. 7 with respect to, for example, consumers who are not 8 BY MR. CHOUDHURY: 8 Pandora users, you couldn't control which of them 9 Q. And just to be clear, your insights 9 received the treatment versus the control in terms 10 there, you're not providing a statistically 10 of the manipulation of the sound recording, right? 11 significant value to the panel, right, on those 11 A. False. 12 insights? 12 Q. But you have no ability to, in fact, 13 A. It's not a test. It's not a 13 control whether someone hears a song on Pandora or 14 statistical test. It's an interpretation. 14 not if they're not actually a Pandora user. 15 Q. And you're not, in fact, providing any 15 A. I think that's a wrong interpretation. 16 numerical value on the net promotional effect; 16 So I agree -- 17 you're saying that this is your interpretation from 17 Q. Let's start with my question. 18 what you're seeing, correct? 18 Do you have the ability to control a 19 A. That's correct. 19 user who is not a Pandora user with respect to 20 Q. Now, this experiment, I think we talked 20 whether they're going to hear the sound recording 21 about in your direct examination, happened at the 21 you're manipulating or not? 22 regional level. 22 A. They have the exact same experience as 23 You picked what we call DMAs, and you 23 everyone in our control. 24 chose on a either specific song or specific album 24 Q. Which is to say that they're not on 25 level two stops spinning Pandora on them, correct? 25 Pandora at all.

4373 4375

1 A. So they're not exposed to the music. 1 randomization and the design of the study, it's 2 Q. But some of them are inside your 2 part of the reason we're able to control for that 3 treatment DMAs, correct? 3 difference. And that's a core difference between 4 A. That's correct. 4 observational and correlation. 5 Q. And some of them are inside your 5 In causation, because we're balancing 6 control DMAs, correct? 6 all of those differences, is this music on 7 A. That's correct. 7 terrestrial radio here or not, because it's 8 Q. And their purchases are being counted 8 entirely balanced, we're able to account for that. 9 toward the sales you call the first spin 9 CHIEF JUDGE BARNETT: Okay. Thank you. 10 performance of Pandora, correct? 10 BY MR. CHOUDHURY: 11 A. Yes. 11 Q. And so just to follow up on that, 12 Q. Now, let's talk about consumers who 12 consumers who, for example, may have seen a TV show 13 were Pandora users. 13 like "The Voice," and then decided to go buy their 14 A. Yes. 14 artist music, those purchases were counted in the 15 CHIEF JUDGE BARNETT: Before we go 15 purchases you called the Pandora spin -- you know, 16 there, Dr. McBride, was there any attempt to 16 the effect of Pandora's performances, right? 17 compare the sound recordings that Pandora was 17 A. We did. We included all sales from 18 leaving in -- 18 those regions. 19 THE WITNESS: Yes. 19 Q. And to follow up on your answer to 20 CHIEF JUDGE BARNETT: -- or -- that's 20 Judge Barnett, that's because you assumed that, 21 the scientific term -- that Pandora was spinning 21 through your randomization, half the time those 22 with contemporaneous activity on either other 22 performances would end up, you know, in a 23 streaming services or terrestrial radio? 23 controlled DMA and half in a treatment DMA? 24 For instance, if Taylor Perry or Katy 24 A. That's not an assumption. That's 25 Swift or any of those were hot on terrestrial radio 25 testable. So for that to be true, all we had to do

4374 4376

1 at the same time that Pandora was using that 1 was test whether or not we successfully randomized. 2 particular sound recording, you wouldn't know what 2 It is the exact same requirement for any clinical 3 Pandora's effect was, correct? 3 trial to be valid. Did we randomize correctly. 4 THE WITNESS: We would. So the direct 4 And we tested that. 5 answer to your question is no, we didn't directly 5 Q. Did you randomize for how many consumer 6 control for that. But it's included in the 6 purchases occurred as a result of a consumer 7 experimental design. 7 listening to Taylor Perry when she was hot on the 8 So because we're randomizing, we're 8 radio? 9 accounting for some of those differences in -- 9 A. I really don't know what that means. 10 perhaps Taylor Swift is really popular in 10 Q. Well, let me ask you. You say that 11 Nashville. And so randomly, because part of 11 randomization is the way you can test for whether 12 Nashville might be in the treatment, and part of 12 you've controlled for these circumstances. 13 Nashville is in the control, we're comparing really 13 So my question to you sir, is -- 14 like to like. And they might actually have access 14 MR. MARKS: My objection is just that 15 to the exact same radio experience. 15 it mischaracterizes the witness's testimony. I 16 And so because we're randomizing and 16 believe he was characterizing what he said. I 17 running an experiment and we're manipulating the 17 don't think it was an accurate or fair 18 experiment, so running a real experiment, we can 18 characterization. 19 actually control for all of those things. 19 CHIEF JUDGE BARNETT: Would you 20 And so, if we had had those data and 20 rephrase. 21 been able to control for it, it's my contention 21 MR. CHOUDHURY: I'll rephrase, yes. 22 that our -- our confidence intervals would be 22 BY MR. CHOUDHURY: 23 tighter. We would have a more exact estimate and 23 Q. Did your randomization -- did your 24 higher statistical significance. 24 study of your randomization specifically look at 25 So we didn't do it. But because of the 25 whether a consumer purchased music after watching a

4377 4379

1 television show of their favorite artist? 1 Let him finish the answer. 2 A. No. It seems quite unrelated. 2 THE WITNESS: They did not affect or 3 Q. And did your study specifically test 3 interact in any way with that one hold-out group. 4 whether a assumer purchased music after using an 4 The music sales experiments operated 5 interactive service? 5 outside of the AB framework because we're not able 6 A. No. 6 to randomize at the individual level because we 7 Q. Did your study test whether a consumer 7 didn't have individual-level transaction data. 8 purchased music after attending a concert for their 8 And so because they were randomized at 9 favorite artist? 9 the region, and we were getting music sales from 10 A. No. 10 off platform, all U.S. listeners with a ZIP code, 11 Q. And did your study test whether a 11 so 95 percent of all of our listeners, were 12 consumer purchased music after watching a music 12 randomized. So they -- including people in our AB 13 video? 13 framework hold-out group. 14 A. No. 14 BY MR. CHOUDHURY: 15 Q. And did your study test whether a 15 Q. And so you agree with me that the 16 consumer purchased music while they were 16 subjects of the music sales experiment could have 17 simultaneously subject to other experiments on 17 been subject simultaneously to other experiments 18 Pandora? 18 from Pandora, correct? 19 A. No. 19 A. Yes. 20 Q. And you agree with me that Pandora is 20 Q. Okay. Now, you determined the 21 constantly running experiments on its users, right? 21 treatment versus the control group -- I think we -- 22 A. I agree we run lots of experiments for 22 by looking at a user's ZIP code, correct? 23 research purposes. 23 A. That's correct. 24 Q. And so the subjects of your experiment 24 Q. And that's the ZIP code they either 25 could have been subject to up to 50 different 25 provided to Pandora when signing up or subsequently

4378 4380

1 experiments at any one time, right? Because that's 1 updated, correct? 2 -- well, let me ask you. 2 A. That's correct. 3 Dr. McBride, how many experiments does 3 Q. And I think you testified that Pandora 4 Pandora, on average, run on its listeners at any 4 can sometimes prompt users to update their ZIP 5 given time? 5 code, correct? 6 A. 40 to 60. 6 A. We do do that sometimes. 7 Q. And so any given consumer on Pandora 7 Q. But you didn't prompt any users to 8 could be subject to 40 to 60 different experiments 8 update their ZIP code as part of the music sales 9 in a given time. 9 experiment, right? 10 A. That's not correct. 10 A. I did not. 11 Q. And -- well, let me see. 11 Q. And so, if they hadn't updated their 12 What is the hold-out group on Pandora? 12 ZIP code, and they moved, you would be using the 13 A. There's many hold-out groups. 13 incorrect ZIP code, right? 14 Q. But how many -- is there a hold-out 14 A. This goes to the dilution point I went 15 group that is exclusively excluded from all 15 about earlier. But yes, I agree. They would be in 16 experiments? 16 a control region if they were assigned to 17 A. Yes. There's one of those. 17 treatment, and -- 18 Q. And that is about what percentage? 18 Q. Just so we make it very clear. I think 19 A. It's 1 percent of our listeners. 19 we talked about this in your deposition. 20 Q. Okay. But your experiments were not 20 If I was a user who lived in Washington 21 subject to the hold-out group, right? 21 State, and then I moved to Washington, D.C., but I 22 A. The steering experiments operated 22 didn't update my ZIP code, your experiment would 23 through the AB framework. So yes, they did not -- 23 have called me a Washington State user, correct? 24 Q. Just to be clear -- 24 A. That's correct. 25 JUDGE STRICKLER: They did not what? 25 JUDGE STRICKLER: And sometimes he

4381 4383

1 would have shown up in the treated group, and 1 was predictive information about past sales in the 2 sometimes he would have shown up in the untreated 2 DMA, correct? 3 group? 3 A. Yes. Absolutely. Certainly 4 THE WITNESS: Yes. Because of the 4 pre-experiment. 5 regional variations across all of the experiments. 5 Q. Right. 6 BY MR. CHOUDHURY: 6 And you know that albums have 7 Q. And you're aware that -- or are you 7 lifecycles of sales, right? 8 aware that, among younger demographics, like 18 to 8 A. Yes. 9 25, those individuals frequently move, correct? 9 Q. They tend to sell more upon release and 10 A. Yes. 10 less in the weeks that follow? 11 Q. And you didn't do anything to 11 A. Yes. 12 particularly update the ZIP codes of that 12 Q. And so, if a song was new to Pandora, 13 demographic, correct? 13 that would be at a particular point in its 14 A. I didn't update the ZIP code data, no. 14 lifecycle, correct? 15 Q. And you matched up the sales by looking 15 A. All over the lifecycle. We have new 16 at sales within a DMA, right, from Nielsen 16 music on Pandora that is very new to the world. 17 SoundScan? 17 And we ingest and have new music on the service 18 A. That's correct. 18 that is a little older, yes. 19 Q. And how does Nielsen SoundScan 19 Q. And I know you haven't studied 20 determine what sales occur in a DMA? 20 seasonality. 21 A. They get data from retailers from where 21 Did you account for seasonality in this 22 the sale occurred. 22 experiment? 23 Q. On where the sale occurred, correct? 23 A. We did not. 24 A. Correct. 24 Q. And all of the music sales experiments 25 Q. So if I drive to another DMA -- for 25 occurred over the summer, correct?

4382 4384

1 example, I live in one Nashville DMA; I drive to 1 A. Starting at different periods. But 2 the other one -- you would have reported the sale 2 yes, they all occurred over the summer of 2014. 3 where it occurred, correct? 3 Q. And one thing you did find out is that 4 A. Correct. 4 sales releases are actually higher in the music 5 Q. Even if my registered ZIP code with 5 industry in the summer, correct? 6 Pandora was in the first DMA -- 6 A. Yes. 7 A. Correct. 7 Q. But you didn't do anything to control 8 Q. -- correct? 8 for any possible seasonality effect? 9 And that would have been a mismatch in 9 A. So those are different. But I agree 10 your data between where the sale occurred and where 10 that I didn't account for that. So -- 11 the user was, as far as Pandora is concerned, 11 JUDGE STRICKLER: So you said that's 12 experiencing the treatment. 12 different. 13 A. So they would experience that treatment 13 What's different? 14 anywhere. But I would agree that we would allocate 14 THE WITNESS: Right. So seasonality I 15 them to a region that they were not resident and 15 interpret as differences in purchasing patterns 16 they would make purchases elsewhere, in -- in your 16 rather than more music being available and you 17 situation. 17 purchasing at the same rate, but there's a greater 18 Q. And you tried to use predictive 18 quantity. And because we're calculating a 19 information about sales in the DMA; is that right? 19 percentage, it's -- 20 A. Yes. 20 BY MR. CHOUDHURY: 21 Q. But that predictive information 21 Q. In terms of what a sale meant for your 22 wouldn't necessarily measure the sales of a song 22 study, you didn't include physical singles, right? 23 that was new to Pandora, correct? 23 A. We did not. 24 A. I -- can you explain that? 24 Would you like to know why? 25 Q. Your predictive information about a DMA 25 Q. Well, because it's not in the Nielsen

4385 4387

1 SoundScan data, correct? 1 manipulating it, we're driving our promotional 2 A. Sorry. Say that again. 2 effect down. 3 Q. It isn't reported to you by Nielsen 3 Q. But, of course, you're still counting 4 SoundScan, correct? 4 those sales as sales towards the total sales that 5 A. No. Incorrect. Nielsen's 5 you're measuring, correct? 6 representatives said it's not worth it, that there 6 A. Correct. 7 are 1,000 units of physical sales per week, and 7 Q. Okay. And you chose to end the 8 since we were getting around 5 million total sales 8 experiments after eight weeks and then, in one 9 a week -- 9 instance, after four weeks, correct? 10 Q. See, but you -- and you made an 10 A. We designed our experiments before 11 independent judgment, too, that you didn't think it 11 starting them. And we set periods of 12 was worth it to include it in the study, correct? 12 experimentation, which were, in all cases, eight 13 A. We decided not to include that 13 weeks with the exception of the final ones, which I 14 additional form that where Pandora could be 14 agree were four weeks. 15 promotional. 15 Q. And did you decide, prior to starting 16 Q. Let's understand how -- if an album is 16 experiments, that the final one would be four 17 purchased, how that counted in your study. 17 weeks? 18 So if an album was purchased, you used 18 A. Yes. 19 what you called track equivalent album; you counted 19 Q. And why did you want to have one 20 that as ten sales, correct? 20 four-week experiment and the rest of them eight 21 A. So yes, it counted as ten units, which 21 weeks? 22 is Nielsen's measure of the average number of 22 A. We were required to finish our research 23 tracks per album. 23 prior to the filing of evidence in this case. 24 Q. Right. 24 JUDGE STRICKLER: Required by whom? 25 And you used what Nielsen said, right? 25 THE WITNESS: Pandora broadly. We --

4386 4388

1 A. We used their standard measure of units 1 part of it was we had a limited access to data from 2 of sale, yes. 2 Nielsen SoundScan, which had an end date from our 3 Q. And in the new music sales experiments, 3 original agreement. So we had to meet that. 4 let's understand how an experiment's created. 4 That end date was based originally on 5 So a song comes into Pandora, Song A, 5 needing to finish our research by the end of the 6 let's call it, correct? Right? 6 period. 7 A. Okay. 7 JUDGE STRICKLER: The period meaning -- 8 Q. And it meets your inclusion criteria. 8 being defined as what? 9 And so then the album that song A is on 9 THE WITNESS: The period of research 10 becomes an experiment, correct? 10 where we could get access to music sales data from 11 A. That's correct. 11 Nielsen. And so our agreement with Nielsen I think 12 Q. Now, if there is Song B on the same 12 ended September 13. 13 album, and I purchase Song B, do you count that as 13 JUDGE STRICKLER: And the period -- the 14 a sale toward that experiment? 14 end date was a function, in part, of when evidence 15 A. Yes. That was one unit of sale. 15 needed to be submitted in this proceeding? 16 Q. Even if Song A is the one that is 16 THE WITNESS: Yes. 17 manipulated in the treatment and control? 17 JUDGE STRICKLER: Thank you. 18 A. No. All of tracks on that album, A, B, 18 BY MR. CHOUDHURY: 19 C, through let's say F, would also be included. 19 Q. And just to follow up on that, you knew 20 All of those would be turned off for a person in 20 that the margin of error in your study would 21 the control. 21 decrease if you extended the experiment out to 12 22 Q. But if Song A is the new-to-Pandora, 22 weeks or 16 weeks, correct? 23 how do we know that Song B was even ever going to 23 A. Yes. We did many studies before 24 play on Pandora? 24 starting experiments to design our experiments. 25 A. So we don't. That said, if we're not 25 Q. Right.

4389 4391

1 But you chose eight weeks and for 1 statistically significant result, correct? 2 weeks, correct? 2 A. Not at 5 percent. I agree. 3 A. We analyzed the data that you might 3 Q. And, in fact, if we look at the Poisson 4 present would show that we would get precise 4 regression, you find the possibility of 5 estimates of a promotional effect, if there be one, 5 substitution within the confidence interval of all 6 using eight weeks' data. We made the judgment to 6 experiments, majors and others, correct? 7 use eight weeks. 7 A. I presume you're talking about new 8 Q. Now, Dr. McBride, I want to take a look 8 music and not -- 9 at some of the results that we were looking at. So 9 Q. Yes. The new music. 10 if we can start with Table 4 in your testimony. 10 A. Oh, okay. Yes. That's correct. 11 And if you -- if you want to look at that. 11 Q. Now, you made the decision to use the 12 CHIEF JUDGE BARNETT: Can you give us a 12 TMLE methodology as your estimator, correct? 13 page reference, please. 13 A. My colleagues and I, as researchers, 14 MR. CHOUDHURY: I wish I could. It's 14 prior to experimentation, before we had any data, 15 in the middle of his written direct testimony, but 15 we did select that as our estimator, yes. 16 it's not actually page numbered. 16 Q. So you -- but you had personally never 17 JUDGE STRICKLER: Figure or a Table 4? 17 used TMLE, right? 18 MR. CHOUDHURY: So it's -- I want to 18 A. I had not. 19 look at Table 4. And so this is between Table 3 19 Q. And also that's true with the respect 20 and 5. 20 to your six years at Analysis Group; you never used 21 This is -- so there's the testimony, 21 TMLE there, right? 22 then the figures, then the tables. And this is the 22 A. That's correct. 23 one that says "Music Sales Experiment Effective 23 JUDGE STRICKLER: Excuse me. 24 Performance on Pandora By Music Ownership Group." 24 You said you decided to use TMLE before 25 BY MR. CHOUDHURY: 25 you ever had any of the data.

4390 4392

1 Q. Now, I think we discussed this in your 1 THE WITNESS: That's correct. 2 direct, that you did not find a statistically 2 JUDGE STRICKLER: Is there 3 significant effect on what you label here as other 3 documentation to that effect? 4 recording label, correct? 4 THE WITNESS: We didn't write that 5 A. I agree. 5 down. What we can say and is -- has been produced 6 Q. And if we look at -- within the 6 is that we used it to analyze the pilot prior to 7 confidence interval there, noting that it's not 7 having any data. 8 statistically significant result, am I reading this 8 And what I would say more broadly is my 9 correct that, when it says negative 2.86, that 9 colleague, Oliver Bembom, who assisted with this 10 would be the possibility of substitution? 10 research, his adviser was the creator of this 11 A. Correct. 11 method. He used it numerous times at previous 12 Q. And if we look at the confidence 12 work. 13 intervals for majors where you found a 13 And for numerous statistical 14 statistically significant result, within the 14 properties, specifically contrary to the Poisson 15 confidence interval there's still a possible effect 15 regression, it's proven mathematically to produce 16 of less than 1 percent, correct? 16 the smallest confidence interval, the most precise 17 A. That's the lower range, but yes. 17 estimate for promotion or substitutional effect. 18 Q. So now if we could turn to Table 7. 18 JUDGE STRICKLER: So you used the TMLE 19 So now, if we look at Table 7 with 19 when you did the pilot? 20 respect to the new music experiments, am I correct 20 THE WITNESS: Yes, we did. 21 that this is the same experiments, but you're using 21 JUDGE STRICKLER: Is there 22 a Poisson regression as your estimator instead of 22 documentation of that in the record? 23 TMLE, correct? 23 THE WITNESS: Yes, there is. 24 A. Yes. 24 JUDGE STRICKLER: In your report? 25 Q. Okay. And here you do not find a 25 THE WITNESS: It was in the materials

4393 4395

1 that were provided as exhibits last night. 1 certainly the implication. But that's because of a 2 MR. MARKS: And, Your Honor, if I could 2 core statistical property, which is that it is 3 clarify one question. 3 efficient. It has the smallest confidence 4 I think it's possible the witness may 4 interval. 5 be confusing your question about whether or not 5 Q. Right. 6 something in the record with whether or not it's 6 And so the answer to my question is it 7 reflected in information that was produced during 7 is most likely to generate a statistically 8 the discovery phase. 8 significant result, right? 9 JUDGE STRICKLER: Fair enough. So let 9 A. If there is, in fact, a true effect. 10 me ask counsel is the use of TMLE in the pilot 10 Q. And if the music sales experiment found 11 project something that's been in evidence in this 11 a statistically significant result that was not 12 case now? 12 promotional, that would be helpful to the evidence 13 MR. MARKS: I don't believe so beyond 13 Pandora presents in this proceeding, correct? 14 his testimony. 14 A. That's seems logical. 15 MR. CHOUDHURY: And we -- I think, when 15 Q. And you knew that these studies were 16 we get to the restricted section, we'll get -- 16 done at the direction of counsel, correct? 17 we'll look at that document. 17 A. Yes. 18 JUDGE STRICKLER: Okay. We'll wait. 18 Q. And that, if they were helpful, they 19 BY MR. CHOUDHURY: 19 were likely to be offered in this proceeding, 20 Q. So just with respect to TMLE, you 20 correct? 21 didn't discover TMLE through your college, Oliver 21 A. Yes. 22 Bembom, right? 22 Q. And when you were conducting the study, 23 A. That's correct. 23 you were assuming that, if they did not -- if they 24 Q. How did you discover it? 24 were not helpful, they would not be used in this 25 A. I ran into a professor who mentioned 25 proceeding, correct?

4394 4396

1 it. 1 A. So I assumed that they would not be 2 Q. And that was at the airport, and this 2 presented by Pandora. But I have no sense of the 3 is the Berkeley statistics professor? 3 discovery process, whether or not they would be 4 A. He's a Berkeley professor, I believe of 4 provide by yourselves. 5 econometrics, but... 5 Q. So when you were making the decision to 6 Q. And you did take a graduate statistics 6 use TMLE, you knew it was most likely to produce a 7 course just a year or two ago, correct? 7 statistically significant result, correct? 8 A. I took two courses, one of which I 8 A. It satisfies -- it has that 9 audited. A data analysis, empirical analysis 9 implication. 10 course using a particular statistics language from 10 Q. And that result, if net promotional, 11 Stanford. And I also took another course through 11 would be helpful to Pandora in this proceeding, 12 an online source. 12 correct? 13 Q. And neither of those courses refer to 13 A. Among other settings, yes. 14 TMLE, correct? 14 Q. And if net substitutional, you were 15 A. They do not cover that, no. 15 assuming that Pandora would not present that study 16 Q. And in your work at Analysis Group, did 16 in this proceeding, correct? 17 you know anyone else who applied TMLE? 17 A. I -- yes. 18 A. Not to my knowledge. 18 Q. And the steering experiments, they did 19 Q. Now, I think a second ago, in response 19 not use the TMLE methodology as an estimator, 20 to Judge Strickler's question, you said that TMLE 20 right? 21 has a very small confidence number, right? 21 A. That's correct. 22 A. That's a math proof, yes. 22 Q. And those were done at Dr. Shapiro's 23 Q. And so, therefore, it's most likely to 23 instruction, correct? 24 generate a statistically significant result, right? 24 A. That's correct. 25 A. I mean yes and no. I mean that is 25 Q. And he is not a full-time Pandora

4397 4399

1 employee, right? 1 Q. And so, when a song is sold, a label or 2 A. He is not. 2 artist would actually receive 70 cents, not a 3 Q. And at his instruction, you used a 3 dollar, correct? 4 difference in unconditional means analysis, 4 A. I agree. 5 correct? 5 Q. And so if that's true, then Table 5 is 6 A. Yes, we did. 6 not an accurate statement of the revenue to labels 7 Q. If you had used a difference in 7 and artists, right? 8 unconditional means analysis for the music sales 8 A. I think the title is not as clear as I 9 experiments, you would agree with me that your 9 would like. This estimate of retail price, and I 10 results were much more likely to have not been 10 agree -- so this is gross, and then there is a net 11 statistically significant, correct? 11 implication. 12 A. Correct. 12 Q. Right. 13 Q. And even though you knew that, you 13 So you agree with me that this does not 14 didn't attempt to use analysis under this 14 reflect the revenue to labels and artists. 15 difference in unconditional means, which I think at 15 A. Yes. 16 least you can tell us is not -- is the more 16 Q. And you didn't consult with anybody 17 conventional estimator, correct? 17 about whether $1 a track was the appropriate 18 A. It's also -- it is more conventional, 18 assumption, right? 19 and it's also inappropriate in this setting. 19 A. I didn't when I submitted it. 20 Q. I'd like to take a look at Table 5. 20 Q. And you didn't consult with any 21 A. Yes. 21 industry sources either, right? 22 Q. Okay. So here is where you attempt to 22 A. Not when I submitted this. 23 put a per-spin effect on the revenue to labels and 23 Q. At the time of your study, you also 24 artists, correct? 24 weren't aware of the average album price in the 25 A. Yes. 25 industry, correct?

4398 4400

1 Q. And what you see here is that, using 1 A. That's correct. 2 your own numbers, the per-spin effect on catalog is 2 Q. And you didn't look at how catalog 3 much smaller than the per-spin effect you found on 3 albums are priced in the industry, correct? 4 the new music sales experiments, right? 4 A. That's correct. 5 A. That's direct. 5 Q. And you aren't aware of whether 6 Q. And there, in fact, is not a 6 download signals could be priced at a retail level 7 statistically significant effect when we look at 7 lower than a dollar, correct? 8 other recording labels under new music experiments, 8 A. I knew that to be true. 9 right? 9 MR. CHOUDHURY: Okay. Mr. Nichols, I'm 10 A. I agree. 10 going to ask you if you can pull up the first 11 Q. Now, I want to draw your attention to 11 demonstrative, please. And yeah, we've also handed 12 Footnote 1. Here it says: "Results are cents per 12 out the demonstrative slides there in the folder. 13 spin assuming that track equivalent units sell for 13 BY MR. CHOUDHURY: 14 $1," correct? 14 Q. Dr. McBride, do you recognize this? 15 A. Yes. 15 A. I do. 16 Q. And this assumption is important to 16 Q. Okay. What is this? 17 your calculation in this table, correct? 17 A. This is Apple iTunes, which was 18 A. Yes. 18 accessed by Philip recently. And it shows some of 19 Q. Okay. And that's not -- $1 is not the 19 the recent songs that were available for sale for 20 revenue that a label or artist receives from a sale 20 digital download from iTunes and also shows their 21 of a single, correct? 21 prices. 22 A. Correct. 22 Q. And you would agree with me that at 23 Q. Because a label or artist receives 23 least some of the downloads are sold at 69 cents, 24 approximately 70 percent of the sale, correct? 24 correct? 25 A. That's my understanding. 25 A. I see that, yes.

4401 4403

1 Q. And do you also see in the top left 1 numbers on the table would be exactly lower. 2 where it says: "New album plus 7.99 albums"? 2 Q. And if you looked at Table 6, Table 6 3 A. Yes, I do. 3 is dependent on your results from Table 5, correct? 4 Q. So you'd agree with me that some albums 4 A. It is. 5 are priced at 7.99, correct? 5 Q. So if those were incorrect, so too 6 A. I interpret it that way, yes. 6 would be Table 6, correct? 7 Q. And so, if we assume that an album was 7 A. So the effect estimate in revenue terms 8 ten songs, that would actually be less than a 8 would be affected by the price assumption. The 9 dollar a song, correct? 9 confidence interval and the P value, whether or not 10 A. That would be. 10 the results are statistically significant, are not 11 MR. CHOUDHURY: And if we can go to the 11 dependent on the assumed price. 12 second slide. 12 Q. So thank you for that clarification. 13 BY MR. CHOUDHURY: 13 If we look at Footnote 1 on Table 6, 14 Q. Dr. McBride, here we're looking at the 14 are you making the same assumption in Footnote 1, 15 B.B. King album, which I think if we got lower on 15 Table 6, that you made in Footnote 1 of Table 5? 16 the first demonstrative, you'd see is actually on 16 A. Yes. 17 the top 10 right now. 17 Q. Now, if we can look at Figure 3(a). So 18 Do you see that it's a 7.99 buy? 18 this is a figure. So we're now pretables. 19 A. I see that it's priced at -- oh, sorry. 19 A. Sorry for that. 20 Yes. I see exactly those words. 20 CHIEF JUDGE BARNETT: Do you have a 21 Q. And that is, in fact -- and I'm going 21 page number for that one? 22 to represent to you it's, in fact, a top 10 album, 22 MR. CHOUDHURY: You know, it -- the 23 correct? 23 figures are right before the tables. 24 A. I'll accept that representation. 24 CHIEF JUDGE BARNETT: Oh, thank you. 25 Q. So even new albums or albums that are 25 MR. CHOUDHURY: So I can give you

4402 4404

1 -- let's not say new albums. Let's withdraw that. 1 the -- between Figure 2 and Figure 3(b). 2 Even albums that are high on the charts 2 THE WITNESS: I apologize for not 3 can be priced below $10, correct? 3 having sequential numbering. 4 A. Yes. 4 JUDGE STRICKLER: 3(a)? 5 Q. Just so we're clear, you didn't have 5 MR. CHOUDHURY: 3(a). 6 anyone look at the price of the albums in your 6 BY MR. CHOUDHURY: 7 experiment at the time of the experiment, correct? 7 Q. And this was a figure that we looked at 8 A. Correct. 8 during your direct examination, correct. 9 Q. And you didn't have anyone look at the 9 A. Correct. 10 price of the digital singles in your experiment at 10 Q. Now, you would agree with me that, on 11 the time of the experiment, correct? 11 all three of your graphs, at the 50-spin-per-sale 12 A. That's correct. 12 threshold, you find a statistically insignificant 13 Q. That was true for both the new music 13 result, correct? 14 sales and the catalog music sales experiments, 14 A. That's correct. 15 right? 15 Q. So just so we're clear, it's not a 16 A. That's correct. 16 linear dosage; it's not that, the more spins per 17 Q. And you haven't done any analysis to 17 sale, it'll automatically increase, correct? 18 see how sensitive the results you're reporting in 18 A. I don't think that interpretation can 19 Table 5 are to your assumption in Footnote 1, 19 be made from a single data point. A linear 20 correct? 20 relationship can still have noise around it. 21 A. It's straight math. So any changes in 21 So if you drew a line, that could still 22 that assumption would directly scale the numbers. 22 fall within the upper confidence bands of those 23 So if the number were 25 percent higher, all of the 23 points. So I agree that those points are 24 numbers in the table would be exactly 25 percent 24 insignificant. 25 larger. If it was 25 percent lower, all the 25 Q. And, Dr. McBride, do you have any

4405 4407

1 explanation for why you get a statistically 1 alternative platforms, correct? 2 insignificant result at every -- on all three 2 A. That's not what I said. What I said 3 charts at the 50-spin-per-sale threshold but not 3 was, when that's true and it's spinning very little 4 necessarily at thresholds lower than that? 4 on Pandora. 5 A. I think it's -- the data are speaking. 5 So it stands to reason to me, if it has 6 These are the results. 6 essentially no exposure on our service, I don't -- 7 Q. Okay. If we can look at Page 21 of 7 I find it unlikely that we would be promotional for 8 your written direct testimony. 8 that. I'm sure we cannot defect it when we have 9 A. I'm there. 9 such a minuscule effect relatively. 10 Q. So there you write: "It's logical that 10 Q. Well, the logical inference you're 11 Pandora would not create a detectible impact on 11 drawing is that the availability of a song on an 12 sales for songs that are playing heavily on 12 alternative platform would affect the possible 13 alternative platforms while playing very little 13 promotional effect on Pandora, correct? 14 relatively on Pandora," right? 14 A. I didn't say that. 15 And we used Taylor Swift as an example, 15 Q. Okay. I just want to be clear. I 16 so of course that got my attention, right? 16 think we talked about this earlier. 17 So what we're talking about here is the 17 You testified about these experiments 18 dosing effect, correct? 18 in the BMI rate court proceeding, correct? 19 A. So -- largely, yes. 19 A. That's correct. 20 Q. And what you're saying here is that, if 20 Q. And you're aware that Judge Stanton had 21 a song is spinning on an alternative platform at a 21 issued a decision in that proceeding, correct? 22 very high level, for example, it's getting a lot of 22 A. I'm told that's the case, yes. 23 terrestrial spins, and it's not spinning that much 23 Q. Do you know whether Judge Stanton 24 on Pandora, you wouldn't actually see a very large 24 relied on the music sales experiments as part of 25 promotional effect, right? 25 their opinion?

4406 4408

1 A. That stands to reason. 1 MR. MARKS: I'm just going to object to 2 Q. And so, as long as a song is spinning 2 the -- just caution the witness -- I don't know 3 on an alternative platform at a high level, you 3 what the witness -- if he has an understanding 4 wouldn't actually see the large promotional effect 4 other than as a result of conversation of counsel. 5 that you're reporting in your study, correct? 5 But I just want to be clear it's not going to be 6 A. Sorry. Could you repeat that, please. 6 intruding on conversations that Dr. McBride may 7 Q. Sure. 7 have had with counsel. 8 Here we're looking at Paragraph 46. 8 CHIEF JUDGE BARNETT: If you can answer 9 And what you're explaining there, I believe, but 9 the question without revealing a conversation 10 correct me if I'm wrong, is that, if a song is 10 you've had with counsel, then you may answer. 11 spinning on an alternative platform at a high rate, 11 JUDGE STRICKLER: Also, just -- didn't 12 but it's spinning relatively little at Pandora, you 12 we hear this morning that the decision is 13 wouldn't expect to see a promotional effect, right? 13 embargoed? So if it's embargoed, even if he knows, 14 A. I -- if it's spinning at a low rate, I 14 why would we violate the embargo and let him say 15 agree. 15 something that he knows otherwise is -- 16 Q. And so the promotional effect of 16 MR. CHOUDHURY: Your Honor, I'll 17 Pandora is actually dependent in some sense on the 17 withdraw the question. 18 availability of the song on alternative platforms, 18 And, in fact, at that point, we should 19 correct? 19 move to restricted session for the rest. 20 A. So our ability to estimate it does not, 20 CHIEF JUDGE BARNETT: Okay. We will be 21 but the estimate amount might -- might depend on 21 closing the hearing room now to anyone who has not 22 whether it's -- 22 signed the nondisclosure certification. 23 Q. Right. 23 MR. CHOUDHURY: I actually have a few 24 You would estimate a lower promotional 24 more public questions, but -- so people can hang 25 effect if a song was spinning heavily on 25 around while I'm asking them if they want.

4409 4411

1 THE WITNESS: I'll stay. 1 2 MR. CHOUDHURY: Great. Me too. 2 3 BY MR. CHOUDHURY: 3 4 Q. Dr. McBride, I want you to assume for a 4 5 second that, in your experiment, 90 out of a 5 6 hundred listeners stopped buying music because of 6 7 what they heard of Pandora. They just -- they 7 8 decided to walk away. And ten out of the hundred 8 9 users decided to buy more music because of what 9 10 they heard on Pandora. 10 11 Does your experiment show the effect on 11 12 the 90 users who walked away? 12 13 JUDGE STRICKLER: When you say walk 13 14 away, you mean walk away from purchasing? 14 15 MR. CHOUDHURY: From purchasing. 15 16 THE WITNESS: I think so. Because they 16 17 would be randomized between treatment and control. 17 18 BY MR. CHOUDHURY: 18 19 Q. So if someone chooses not to purchase, 19 20 in this example the 90, how would we see that in 20 21 your results? 21 22 A. So sorry. You're saying that 90 22 23 percent have chosen not to purchase anything. We 23 24 see a difference in whether or not there are sales. 24 25 So your hypothesizing that Pandora 25

4410 4468

1 caused them to stop purchasing? 1 (THIS BEGINS PUBLIC SESSION) 2 Q. For example, they thought that Pandora 2 MR. RICH: Thank you, Your Honors. 3 was satisfying enough that it was net 3 REBUTTAL EXAMINATION BY COUNSEL FOR 4 substitutional for them. PANDORA 5 A. I'll have to think about it more. My 4 (CONTINUED) 6 general sense is that it's an extreme hypothetical. 5 BY MR. RICH: 7 If they're randomized, we would not see any 6 Q. Welcome back, Professor Shapiro. 8 promotional or substitutional effect. 7 A. Thank you. 9 Q. And for any individual user who decided 8 Q. I wish I could say it feels like 10 not to purchase because of what they heard on 9 yesterday, but honestly, it doesn't. 11 Pandora, how would we see that substitution effect 10 We left off -- still covering on your 12 reflected in your results? 11 prior examination, a series of topics drawn from 13 A. That would not directly be observable. 12 your original rebuttal testimony. 14 We would observe other effects. 13 Do you recall that? 15 MR. CHOUDHURY: Okay. Now let's move 14 A. I do. 16 into a few restricted questions about some of the 15 Q. And just to recenter where we are, if 17 documents. 16 you will take a look at the first -- pardon me, at 18 (THIS ENDS PUBLIC SESSION) 17 the second document in the demonstrative which 19 (RESTRICTED SESSION BOUND 18 begins: "Seven major topics." SEPARATELY) 19 Do you have that? 20 20 A. Yes. 21 21 Q. We covered Topics 1 through 4 as of the 22 22 time we interrupted your examination, and 23 23 therefore, Your Honors, we're going to pick up with 24 24 Topics 5 through 7 beginning this afternoon. 25 25 So Topic 5, Professor Shapiro, relates

Before the UNITED STATES COPYRIGHT ROYALTY JUDGES THK LIBRARY OF CONGRESS Washington, D.C.

) In re ) ) DETERMINATION OF ROYALTY ) Docket No. 14-CRB-0001-WR (2016-2020) RATES AND TERMS FOR ) EPHEMERAL RECORDING AND ) DIGITAL PERFORMANCE OF ) SOUND RECORDINGS (PXB IV) ) )

WRITTEN DIRECT TESTIMONY OF TIMOTHY WESTERGREN (On behalf of Pandora Media, Inc.) Introduction

1. My name is Tim Westergren. I am the Pounder ofPandora Media, Inc.

("Pandora" or "the Company"). I also sit on Pandora's Board ofDirectors. The primary focus ofmy current role at Pandora is guiding the overall strategy and vision for the Company. I regularly work with technology partners, distribution partners, advertisers, and investors to help ensure that Pandora continues to grow and realize our mission ofunleashing the power ofmusic to support artists and provide a wonderful music listening experience for our audience. I

&equently travel across the United States to promote Pandora, connect with Pandora listeners, and refine the strategy for Pandora's future.

2. I graduated &om Stanford University in 1988 with a degree in political science.

I While at Stanford, I also studied music theory, composition, and recording arts at Stanford's

Center for Computer Research in Music and Acoustics. I am a life-long musician and trained jazz pianist. I have more than twenty years of experience in the music industry in areas ranging

&om production and audio engineering to film scoring and live performance. 3. As this is the first proceeding before the Copyright Royalty Board ("CRB") in

which Pandora has participated, I submit this testimony to describe for the Copyright Royalty

Judges the founding ofPandora and some key milestones in its early history, including the

invention ofthe Music Genome Project ("MGP"), the patented music recommendation tool that

forms the core ofPandora's service. Perhaps more importantly, I submit this testimony to

elucidate the guiding principle that drove Pandora's creation and subsequent success: to devise a

personalized Internet radio platform that exposes listeners to music they will love, and gives

artists the opportunity to have their music discovered by fans who might not otherwise have

learned about it. It is a deeply held, personal passion ofmine, and one to which Pandora has

been committed from its inception. It is this commitment to creating a symbiotic relationship

between listeners and artists, combined with our hard work and ingenuity in developing a truly

unique product, that has made us the leader in Internet radio in the United States today.

The Develo ment of the Music Genome Pro'ect

4. After college, following years ofpractice, composing and performing, I formed a rock band called Yellowwood Junction, for which I wrote original songs and played keyboards.

Our band released a record and toured the Pacific Northwest, but we struggled to make ends meet. I eventually realized that it was unlikely that I would be able to support myself as a member of the band, so I explored other avenues and began to compose soundtracks for commercials and small independent films.

5. During my work as a film composer, I found that many directors lacked the vocabulary or frame of reference to articulate the type of music they wanted for their films. To help solve this problem, I developed a form ofmusical interview that allowed me to identify the preferences of a given director in musicological terms by letting the directors listen to certain songs and then provide their feedback. I translated that feedback into musicological information,

which I then used to compose music that reflected their preferences.

6. As I became increasingly proficient at mapping taste in this way, I came up with

the idea of creating a taxonomy for music recommendation. Basically, the vision was to use the

Internet as a platform to deliver a smart recommendation service that could help listeners

discover music that they would love, but not through the typical "people who bought this also

bought that" format of other collaborative filtering-based services, which are essentially popularity contests that favor already-established artists. Rather„ this method would be based on musical similarity, without regard to popularity — a level playing field for all music. My idea was that once a song's characteristics (or "genes") were mapped, and expressed as numerical values, the recommendation tool could use mathematical algorithms to identify other songs with similar musical "DNA" to those a user already knew and liked. This idea grew into what is now known as the Music Genome Project, which is a completely popularity-blind method of making music recommendations.

7. In 2000, I partnered with Will Glaser and Jon Kraft to start Savage Beast

Technologies ("Savage Beast"), the company that would later become Pandora. I was in charge ofthe musical taxonomy, the lynchpin of Savage Beast's product. Shortly after founding Savage

Beast, I hired Dr. Nolan Gasser, a musicologist from Stanford, to help develop and refine the

MGP. Dr. Gasser and I developed the "pop/rock" "genome," which we then further refined into five separate genomes: Pop/Rock, Jazz, Hip-Hop/Electronica, World Music, and Classical.

These genomes formed the taxonomical structure subsequently used to map each genre of music for the MGP. 8. Each genome contained a set ofhundreds of individual "genes" or traits typically

present in that genre ofmusic, including very granular details on instrumentation, tempo, form,

melody, harmonic structure and lyrical content ofthe works. We also developed a standardized

process of analyzing each recording (listening and assigning a score to each gene based on its

role in the work), and trained a team ofmusic analysts to begin building the MGP's musical

catalog. The first entry into the MGP was an analysis ofthe musicological traits that make up

ABBA's "Dancing Queen," which was completed on pencil and paper and subsequently entered

into an Excel spreadsheet.

Savable Beast Technologies and the Transition to Internet Radio

Originally, Savage Beast's business plan was to license the MGP technology as

a recommendation tool to other companies. As an independent musician, I was very interested in

the problem ofhow to connect artists with potential listeners. The Internet created an

environment where every musician could have broad distribution and direct access to listeners, but the volume ofavailable content made navigation and discovery ofnew music a big problem for listeners. I saw the MGP as a way to solve that problem by helping to introduce listeners to new songs and artists who they might otherwise not have encountered, based on what they already liked.

10. Our initial strategy led us to market the MGP as a tool that music retailers and music websites could use to drive new music sales and consumption. Savage Beast's first product was a set ofweb tools that allowed customers to integrate our system into their websites through application programming interfaces ("APIs"). We also developed sofbvare for Internet- enabled kiosks located in "brick-and-mortar" retailers; customers could listen to music at these kiosks, and, with the assistance ofthe MGP, discover new music that shared characteristics of the songs they enjoyed. We marketed this technology to large music retailers such as Tower

Records, Best Buy, and Borders. These brick-and-mortar retailers used the software via web-

enabled kiosks to promote music sales. According to our retail partners at the time, these

systems had a demonstrated success rate: for example, Borders informed us that its use of kiosks

resulted in a 15% "lift" rate (i.e., stores that used the kiosks sold 15% more music than stores

that did not).

11. Between 2000 and 2003, Savage Beast struggled financially. The CD market was

on the decline, Internet portals and retailers were struggling, and retail stores were increasingly

unwilling or unable to invest in listening kiosks. We soon exhausted our initial investment and

resorted to salary deferral, as well as the founding team taking on substantial personal debt to keep the business alive.

12. In 2004, after literally hundreds ofpitches to investors, Savage Beast finally raised a second round of financing from a consortium ofventure capital firms. With those funds, we paid back our employees, who had been working without pay for some time,'nd hired a talented new executive team. Together, we began the process of considering strategic options for the future ofthe business.

13. Surveying the digital music landscape, we began to wonder whether the company's most valuable asset — the MGP — might be better employed outside the music retail business. In September 2004, we put together a presentation deck ofpotential new strategies, including a page, well-known among long-time Pandora employees, with a single, simple line:

"one-click custom radio."

14. We recognized that throughout the digital transition, radio had remained a very robust business, and that radio listening was increasingly shifting to the Internet. Although other companies were already operating in the online radio space at that time, and were growing, i we

felt that Savage Beast could leverage its core technology to develop a radio product that was

significantly simpler and easier to use, and could create far more compelling and personalized

song selections. We also believed that transitioning to radio would stay true to the company's

core purpose of connecting listeners with new music while helping artists find their audiences.

and earn income from the use of their music.

15. Savage Beast made the decision to abandon the business-to-business retail

strategy it had been pursuing and shift to a consumer-based Internet radio model. We repurposed the recommendation technology into a playlist engine, renamed the company Pandora Media, and set about creating a consumer-facing product and brand.

16. In building the new product, we modeled Pandora on the best oftraditional radio: a highly promotional form ofmusic consumption that drove discovery, which led to further paid consumption through concerts and record sales, to the benefit ofthe entire music ecosystem.

Pandora also had at least one major advantage over terrestrial radio: using the MGP, we could introduce listeners to a broader range of songs and artists they did not previously know, artists that rarely, if ever, received airplay on terrestrial radio. Pandora could bring invaluable exposure to talented, but otherwise little-known, artists. In short, our mission was to connect artists with the audience they deserved — a role we felt was not being adequately filled by radio of any kind.

Our listeners frequently send us feedback that this is precisely their experience when they listen to Pandora.

'y 2005, tech giants like America Online and Yahoo! had entered the Internet radio market with AOL Music / Spinner.corn and Launchcast, respectively, and Microsoft's Internet radio product already had fairly substantial audiences. Many terrestrial radio stations were also simulcasting their over-the-air programming through branded websites. 17. In addition to providing the invaluable service of connecting audiences and artists,

we set out to build a legal business that would, through a delightful listening experience and the

payment of royalties, offer a constructive alternative to piracy in the digital music space, thereby

further supporting a healthy music ecosystem.

18. Pandora Radio launched in the fall of 2005. We began with a subscription model

that allowed for ten free listening hours, after which users were required to subscribe at a rate of

$36 per year. That model proved to be flawed, as listeners were generally unwilling to pay for

the service. Instead, they would listen to the free ten hours and then never subscribe.

Recognizing this reality of consumer behavior, in November 2005 we launched a free, ad-

supported version of Pandora Radio and began to hire a sales team to sell advertising. (Mike

Herring, Pandora's Chief Financial Officer, will separately explain Pandora's contribution to creating, from the ground up, a dynamic and growing audio advertising market for Internet radio.)

19. The tenacity we demonstrated in surviving years of serious economic hardship, and our ability to seize a new opportunity through innovation and creativity, have been rewarded with a very successful consumer product that has grown almost entirely by word of mouth. It is personally gratifying to see Pandora fill such an enormous unmet market appetite for personalized Internet radio. As of June 2014, some 77 million Americans (nearly one out of every three Americans over the age of 13) tuned into Pandora on its various platforms for an average ofI hours per month. Demand for our service comes not only from our listeners. Pandora receives thousands of monthly submissions directly from artists and record labels looking to be played on Pandora in order gain a wider audience for their music; approximately ~ tracks were submitted to Pandora in 2013 alone. 20. We continue to invest heavily in the music classification "engine" — the MGP— that lies at the heart of our service offering. The database that Pandora started with has expanded to include approximately analyzed tracks; Pandora now streams more than 1.5

billion listener hours each month and plays songs from more than 120,000 artists each month.

The vast majority — some 80% — ofthese artists are independent, working musicians whose recordings receive no airplay at all on terrestrial radio.

21. As witnessed by the direct artist feedback below, as well as objective sales data discussed in the accompanying testimony of Stephan McBride, we are without question promotional 'ofrecord sales. In addition, this year, we will pay more than statutory royalties to SoundExchange for the benefit ofrecord labels and artists, both featured and non-featured.

22. Furthermore, Pandora is now actively developing the release of a set offeatures that will allow musicians to communicate with and activate their fans on Pandora. More than

80% ofthe top spinning musicians on Pandora have more listeners that have created a station using their name than they have Twitter followers. We understand these listeners'usical preferences, we know their geographic location, and we can communicate with them. The potential ofthis platform to provide marketing support for working musicians, at this scale, is unprecedented. Arguably, the value this could bring recording artists in concert revenue, commerce, merchandise and other forms ofpatronage could dwarfthe revenue from royalties.

23. Despite Pandora's dramatic success in the nearly ten years since its launch, it should be borne in mind that the Company has endured and continues to endure significant economic challenges. Indeed, despite our years of innovation and enormous investments in both

'or example, Pandora stations have been seeded with Jason Aldean's name, whereas Jason Aldean has 2,380,000 Twitter followers. Similarly the band Rusted Root has been the seed for Pandora stations, while Rusted Root has only 10,700 Twitter followers. our intellectual property and in developing the growing Internet-radio audio advertising market,

Pandora has yet to make an annual profit. As further described in Mr. Herring's accompanying

testimony, Pandora's payments to the record industry under its existing Section 112 and 114

statutory licenses have, to date, constituted a disproportionate and, in Pandora's view, unreasonable, percentage ofPandora's overall revenues that, ifpermitted to continue at such

levels, will severely impede Pandora's ability to continue to fulfill the market's demand for our service, and to invest in the kind of software capabilities and advertising possibilities that could drive substantial value to musicians.

Pandora's Intellectual Pro er and Pla list Technolo

The Music Genome Pro'ect Toda

24.. Today, the MGP continues to be the heart of Pandora, and is the key feature that differentiates our service from those of our competitors in the personalized radio space. The

MGP that exists today represents an enormous and continuing investment in software, data, infrastructure, and content management. While it retains much ofthe fundamental architecture and algorithms that Will Glaser, Dr. Gasser and I developed back in 2000, we have spent, and continue to spend, significant resources to continue to develop it. Unlike some ofPandora' competitors (whose custom radio products incorporate fully computer-driven song selection models), the MGP coding process relies extensively on input from expert music curators and analysts. Indeed, Pandora has experimented in the past with such exclusively computer-driven tools, but was not satisfied with the results it obtained. In short, Pandora determined that the automated tools that are currently available cannot grasp the same musical subtleties as a trained

'ndeed, as Simon Fleming-Wood explains in more detail in his testimony, Pandora's music recommendation tool is starkly different from interactive music services, which make all songs in their database available for search and on-demand listening. For this reason, we do not consider on-demand services to be our fundamental competition. human ear, and are equally unable to effectively surface all the lesser-known music to which popularity-based algorithms are inherently blind.

25. For songs to be included in the MGP, Pandora's music curation team first selects the new music to be incorporated. The team spends significant time and resources searching for new music through exhaustive research, chart tracking, as well as comprehensive coverage of a wide range of music publications, blogs, and other forms of commentary and criticism. As noted, Pandora also receives submissions directly from artists and record labels looking to be played on Pandora. When music is submitted for consideration, Pandora's curators make every effort to listen to every song that is submitted. Generally, ofthe hundreds of submissions Pandora receives each week, roughly ~ ofthe artists have some tracks that make it into Pandora's system. Some of the artists that have found success on Pandora through this self- submission process include Bronze Radio Return, The Gospel Whiskey Runners, Caravan of

Thieves, and Chances End.

26. Once the curators have deliberately chosen the songs to be included in the MGP, they pass the tracks on to Pandora's team ofthirty music analysts, all ofwhom are musicians with deep academic grounding in music theory. The music analysts then study the song and analyze it according to its component characteristics, or "genetic" parts, to create a musical thumbprint for each song. Depending on the genre, each song will be analyzed along 148 to 450 musicological traits. Each trait represents an identifying element ofthe song that must be aurally detected and understood by the analyst. Through multiple listens, analysts evaluate each track

These four acts have enjoyed over performances on Pandora, respectively, since the beginning of 2012. 'andora has rigorous hiring and training requirements to ensure maximal integrity ofthe MGP. Each prospective analyst must pass an exam even to be considered for a music analyst position. Once hired, music analysts in training must review the same song and compare results, and repeat that process until the results are consistent.

10 according to both sonoric traits (such as tempo, vocal range, and instrumentation) and more

stylistic traits (such as blues influence and lyrical quality).

The above chart displays just a portion of the hundreds of elements used to analyze a song. Note that most traits are assigned a value of 1 to 5 based on the prevalence ofthe trait in a given song.

27. All of Pandora's music analysts have the training and musicological expertise to be able to identify and describe these musical characteristics and to assess them according to a uniform system, so that songs can be compared to one another on an objective basis. This consistency, which is the result of a rigorous training method developed by Pandora, ensures the kind of accurate data that can be input into the mathematical algorithm that underlies the MGP.

This analysis takes roughly 15 minutes for a simple pop song, and far more for more complex music, such as classical symphonies. Our team of music analysts is currently analyzing

11 approximatel~ songs per month. Over the last fourteen years since the MGP was invented, our music analysts have devoted as much as ~ hours, or the equivalent of about I years, to listening to and cataloging the musicological traits of each song in the Genome. 28. A short video demonstration ofthe technical features and capabilities ofthe MGP

is attached to this testimony as Pandora Exhibit 1.

Creatin Pla lists

29. The MGP is the cornerstone ofPandora's playlist system. It forms the

musicological basis for.connecting songs. However, the intellectual property of this system has

expanded substantially. Pandora now uses three different components to create playlists from the songs included in its database: (1) Pandora's proprietary content-based recommender, the

MGP algorithm; (2) collective intelligence; and (3) collaborative filtering. Together, each of these approaches allows us to create the best possible playlist for our listeners in a radio-like

"lean back" listening experience.

30. The Music Genome Project Algorithm. The MGP algorithm uses the musical content of songs to determine matches based on objective data. In short, the MGP algorithm utilizes data compiled as part ofthe MGP to find songs with similar musical thumbprints or

"DNA." These thumbprints of analyzed tracks are compared using patented technology developed by Pandora to identify songs with the greatest similarity across the traits. In brief, if a listener selects an artist or genre to "seed" a station (a process explained more fully in the testimony of Simon Fleming-Wood), the MGP's patented song-matching technology identifies songs that share similar characteristics with the source song, and will populate a channel of music for the listener based on those attributes. To identify which songs are similar to the seed, the MGP. uses a nearest-neighbor calculation to determine how geometrically "close" one song is

12 to another. This calculation begins with an artist, song, or genre and adjusts in real-time with

each new input — be it a song or artist. This requires a very high-performance algorithm that can perform highly complex calculations across a vast and. constantly growing database, in a &action

of a second. Pandora has spent approximately hours developing and improving this technology at a cost ofmore than

31. The MGP's song-matching technology is entirely blind to the popularity of a given song. In fact, a listener may be presented with tracks sharing similar musical DNA that are

Rom disparate time periods, relatively unknown artists, or even different genres or cultures. The objective nature ofthe matching process makes the MGP a uniquely effective tool in helping listeners to discover new music. For example, if a listener seeds a station with the 80s band

Journey, they may be introduced to the otherwise little-known band Orion the Hunter, which musicologically is a close match to Journey in the MGP.

32. The MGP algorithm is also the best available methodology to effectively address the "cold-start" problem — the issue that arises when new songs are used to start stations with no information about their performance history — or similarly, when a new listener begins using the service before anything is known about his or her personal preferences. Most systems will fail dramatically in either ofthese scenarios, but Pandora can use the MGP algorithm to effectively identify appropriate songs without any data. A music service without the MGP would have to guess whether a new song will "fit" within a listener's playlist; oftentimes, this guessing creates a complete mismatch that can do substantial damage to the listener experience and to the reputation ofthe service. Or, a new song may only get limited exposure to an artist's existing fan base without exposing the song to a wider audience. Pandora's ability to address the "cold start" problem is a large part ofwhat allowed the service to grow so quickly from the very

13 beginning. Playlists were good from the start, without requiring time-consuming data input on the part of the listener.

33. Collective Intelligence. Pandora's "collective intelligence" strategy uses the

feedback provided by its listeners to further refine their playlists and to identify musical trends. Over time, Pandora has collected more tha~ combined thumbs-up, thumbs-down, and track skips that listeners have provided during their listening experience. Using this data,

Pandora can correct instances where the MGP matches two songs with similar traits that, for some reason, do not appeal to the same audience. For example, if Song A is the seed song for a station, and Song 8 is the closest "relative" identified by the MGP, Pandora will monitor listener responses to Song 8. If listener responses are negative, then Pandora may stop playing Song 8 on stations where Song A is the seed song, even though the MGP might have otherwise determined that Song A and Song 8 should appeal to the same listener.

34. Collaborative Filtering. The third algorithm Pandora employs involves looking at the feedback an individual listener has provided on each of his or her stations to create or improve playlists. After a listener indicates a thumbs-up or down for a song, Pandora relates the listener's individual preferences in favor or disfavor for each of the thumbed songs and uses those preferences to influence the playlists of that listener, as well as other listeners who have similar preferences as expressed through thumbing behaviors. For example, assume Listener A has thumbed-up Song A, and Listener 8 has thumbed-up both Song A and Song B. It is a statistically reasonable premise that Listener A is likely to enjoy hearing Song 8 simply because ofthe relatedness of both Listener A's and Listener 8's thumb-up of Song A. This strategy thus develops "cohorts" with shared listening patterns that can improve their collective experience.

14 As may be obvious, Pandora only uses this algorithm when a listener is thumbing-up or -down

songs. For Pandora listeners who do not thumb-up or -down songs, this algorithm is not utilized.

35. Experimentation. In addition to these playlist algorithms, Pandora is constantly

experimenting with ways to improve the mix of songs presented to listeners. In the ordinary

course of business, when there is a new idea for improving playlist quality, that idea will be

tested on a small but statistically significant group of listeners. The results are evaluated across

140 different metrics to test listener satisfaction, including whether the listener changed the

amount of time he or she spent listening to Pandora, or whether the listener changed the rates at

which he or she returned to Pandora to listen. If listeners respond positively, then the

improvement that was tested will be rolled out to all Pandora listeners. At any given time,

dozens of such experiments are being run simultaneously.

Pandora's Promotion of Artists

36. We are committed to helping artists harness the power of Pandora's Internet radio platform. Artists and listeners'like have provided feedback confirming that we are fulfilling

those objectives. As just a few examples of Pandora's positive impact on artists:

In September 2014, classical artist Chad Lawson wrote to Pandora's curation team after his songs were featured in a Pandora Premieres event to thank Pandora for its role in promoting his music. Lawson said: "Pandora has allowed indie artists such as myselfto reach an audience never imaginable. And this would 'hat have never... happened had it not been for your vision, diligence, and most importantly, love for all things music...." Lawson. also noted the impressive fact his "Chopin Variations hit 41 on iTunes Classical last night and is still sitting pretty. The album isn 't even releasedyet."

Cello rock band Break of Reality credits Pandora for helping them build a healthy career. In a November 2013 article in Forbes magazine, the band's members were quoted as saying: "Before Pandora, we were a regional group. As soon as

15 Pandora hit, we were being listened to around the country. That exposure has led to a record and single sales. It's incredible." Iri fact, although Break ofReality's music is not played at all on terrestrial radio, when their songs were added to Pandora, their sales tripled year-over-year. They have noted that their top streaming tracks on Pandora always correlate to the most successful sales of singles. As one ofthe band members has remarked, "Coincidence? No way!" o Bronze Radio Return, an indie band out of Connecticut, has seen spectacular success based largely on the MGP. Their songs are known for having a sound similar to the popular band Mumford & Sons, but before Mumford & Sons released their first album, Bronze Radio Return had very few spins on Pandora and no radio airplay. In approximately mid-2011, as Mumford & Sons became more popular and more Pandora listeners started stations seeded with that band, songs from Bronze Radio Return were picked up on those stations. This year, Bronze Radio Return is scheduled to play major music festivals like Bonaroo, Lollapalooza and Firefly. Below is a chart that illustrates Bronze Radio Return's dramatic improvement in Pandora spins over time;

BRONZE RADIO RETURN MUMFORD & SONS SPINS/YEAR ON PANDORA + Y/Y GROWTH ACTIVITIES & HONORS

2009

2010 Feb.: Released first album, Sigh No More, in U.S.

2011 Feb.: Performs at Grammys, named Best New Artist Sept.: Released second album, Babel 2013

The band All American Rejects, which had several Top 40 hits in the early 2000s and achieved gold and platinum albums, has used Pandora's unique ability to reach listeners in various markets to its advantage. During one visit to Pandora, All American Rejects was presented with data that showed, much to their surprise, that Salt Lake City, Utah, was among the top ten cities in which the band's music was performed on Pandora. Armed with this information, the band booked a show in Salt Lake City on their next tour, and played to a sold-out audience.

16 The band The Lacs, which performs "hick-hop" or a form of country-rap music and is signed to the Average Joe's label, recently used Pandora data concerning the geographic distribution oftheir plays to leverage greater distribution oftheir albums in Walmart stores. Specifically, The Lacs noted that in addition to the southeast, their music was performing well in Indiana, Ohio, and the Pacific Northwest. Their label used this data in a pitch to Walmart, and succeeded in having their albums carried in nearly half of all Walmart Supercenters nationwide; they have also begun to use similar Pandora data to plan their tours to areas in which they might not otherwise have known they had a fan base.

Artist Jonny Manak has voiced his support for Pandora, noting that Pandora has been helpful with both his band The Depressives and his solo career, allowing his "recordings [to] reach an audience that would otherwise be unavailable to an unsigned artist. Because ofthis, I'm selling my music worldwide without label support!"

Artist Aaron Nudelman has noted: "I'm blown away that my music is being offered and displayed adjacent to musicians I have long admired and worshipped. Now when friends and family listen to Mensclub or my solo recordings on Pandora, they are exposed to many artists they might really dig. My music is part of [a] brand new growing community of listeners and musicians."

Artist Charlie Karr explained Pandora's effect on not only new releases, but on catalog works: "My previous band officially stopped performing a few years ago, except on rare occasions. Thanks to Pandora, there is renewed interest in our back catalog and we have new fans present every time we do get out and play."

37. And this impact grows in direct proportion to Pandora's scale. Spins on Pandora have been shown to lead to a variety ofbenefits not only to relatively unknown artists, but to artists of every level of success — from "mega artists" (in our top 10) to "mainstream artists" (top

1,000), and "mid-level artists" (top 5,000). So even artists who have already seen success

17 through other avenues are exposed to new audiences, and a new and increased revenue stream, as

the result oftheir airplay on Pandora.

38. This impact will only be amplified as we begin to fully roll out the

. When it comes to marketing,

Internet radio has some great advantages over broadcast radio — namely that it is a personalized

and connected medium. So this means each listener can be messaged to in a far more relevant

and efficient way, and by a far broader range of artists. An emerging rock band can send

targeted alerts to their fans in Kansas City announcing an upcoming club date. A singer

songwriter can solicit donations from listeners who have thumbed up her music to fund a new

album. The possibilities are endless, and we are investing heavily in building that platform.

Investing in promotional programs for artists provides benefit both to Pandora and to artists.

Programs such as Pandora Premieres, Pandora Presents, and (described in the separate

testimony of Simon Fleming-Wood), among others, create great value for musicians, while

increasing listener satisfaction and retention, thereby generating greater advertising and

subscription revenue for Pandora — revenue that is ultimately shared with the artists. As these

programs continue to develop, there is an enormous opportunity to nurture this symbiotic

relationship and create economic opportunity for artists and labels, while furlher enhancing our

listeners'ove for music. I am very excited and optimistic about what lies ahead.

In addition to the exposure artists obtain through Pandora's platform, we are also developing new tools and functionality to help artists market and promote their music, and continue to work with the music t industry to collaborate on new ways to advance the interests of artists. 18 Sefore the UNITED STATES COPYRIGHT ROYALTY JUDGES THK LISRARY OF CONGRESS Washmgton, D.C.

)

) DETERMINATION OF ROYALTY ) Docket No. 14-CME-0001-WR(?016-?lM) RATES AND TERMS FOR ) KPHEWiKRAL RECORDING AND ) DIGITAL PERFORMANCE OF .) SOUND RECORDINGS (SX8 1F) ) )

CLARATION OF TIMOTHY STE G

I, Timothy%'estergren, dec1are under pena1ty ofperjnIythat the statements contained in my Written Direct Testimony in the above-captioned proceeding are true and correct to the best ofmykaowlcdge, informationand b,elief. Executed this g dtr'r of October 2014 in Oakland,

CaHfornia.

nk & Westergren Proof of Delivery

I hereby certify that on Thursday, September 26, 2019, I provided a true and correct copy of the Sirius XM and Pandora Designated Written Direct Statement, Vol. III (PUBLIC VERSION) to the following:

Corporation for Public Broadcasting, represented by Kenneth L Steinthal, served via Electronic Service at [email protected]

National Religious Broadcasters Noncommercial Music License Committee, represented by Karyn K Ablin, served via Electronic Service at [email protected]

National Association of Broadcasters, represented by Andrew Gass, served via Electronic Service at [email protected]

SAG-AFTRA, represented by David A. Handzo, served via Electronic Service at [email protected]

Radio Paradise Inc., represented by David Oxenford, served via Electronic Service at [email protected]

National Public Radio, Inc., represented by Gregory A Lewis, served via Electronic Service at glewis@npr.org

SoundExchange, Inc., represented by David A. Handzo, served via Electronic Service at [email protected]

Jagjaguwar Inc., represented by David A. Handzo, served via Electronic Service at [email protected]

iHeartMedia, Inc., represented by John Thorne, served via Electronic Service at [email protected]

Sony Music Entertainment, represented by David A. Handzo, served via Electronic Service at [email protected]

American Association of Independent Music ("A2IM"), The, represented by David A. Handzo, served via Electronic Service at [email protected] Google Inc., represented by Kenneth L Steinthal, served via Electronic Service at [email protected]

UMG Recordings, Inc., represented by David A. Handzo, served via Electronic Service at [email protected]

American Federation of Musicians of the United States and Canada, The, represented by David A. Handzo, served via Electronic Service at [email protected]

Warner Music Group Corp., represented by David A. Handzo, served via Electronic Service at [email protected]

College Broadcasters, Inc., represented by David D Golden, served via Electronic Service at [email protected]

Educational Media Foundation, represented by David Oxenford, served via Electronic Service at [email protected]

circle god network inc d/b/a david powell, represented by david powell, served via Electronic Service at [email protected]

Signed: /s/ Todd Larson