The MMA Mobile ProgrammaticCOVER Video Guide Navigating a Growing, Complex, Media Marketplace

MARCH 2019

©2019 Mobile Marketing Association, Inc. All Rights Reserved. TABLE OF CONTENTS

INTRODUCTION: A Growing Marketplace, With Few Adopted Standards ...... 3

SECTION 1: WHY MOBILE IS NOT LIKE DESKTOP...... 7 1. Mobile Has Two Ad Environments: Mobile Web and In-App...... 5 2. In-App Inventory Partners Play a Significant Role...... 5 3. Mobile Programmatic Is Still Evolving...... 6

SECTION 2: VIDEO FORMATS, PLAYBACK AND TECHNICAL STANDARDS...... 7 Understanding How the Experience Affects KPIs...... 8 VAST, VPAID and MRAID: Pros and Cons of the Three Technical Standards ...... 8

SECTION 3: UNDERSTANDING AND LEVERAGING CREATIVE OPTIONS...... 10 Vertical and Square Video...... 10 Interactive Video...... 10 Dynamic/Personalized Video...... 11 Fully baked dynamic video versioning ...... 11 VAST-based personalization...... 11 VPAID overlays...... 11 VAST end-card personalization...... 11 Playables...... 12 Rewarded (or Opt-In) Video...... 12

SECTION 4: BUYING CHALLENGES AND CAVEATS ...... 13 Limited VPAID Inventory In-App...... 13 Interference from Pre-Bid Verification...... 13 Targeting Parameters That Reduce Scale...... 14 Building Creative Capabilities That Can Achieve Scale...... 14

SECTION 5: LEVERAGING DATA...... 15 Deterministic vs. Probabilistic Data...... 15 Location Data...... 15

SECTION 6: GETTING A HANDLE ON MEASUREMENT...... 16 Getting Viewability for In-App Inventory...... 16 Measuring Conversions...... 17 Measuring Reach and Frequency...... 18

SECTION 7: AVOIDING FRAUD...... 18 1. Ads.txt...... 18 2. App-ads.txt...... 19 3. Ads.cert...... 19

CONCLUSION: Making Mobile Video Programmatic Work for Your Brand...... 20

GLOSSARY...... 21

ACKNOWLEDGEMENTS...... 24

ABOUT MMA...... 24

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 2 INTRODUCTION A Growing Marketplace, With Few Adopted Standards

If you’re reading this report, you probably already have an idea what programmatic advertising is. Although it has been defined in many ways, at its core, programmatic is the automation of ad buying and selling, resulting in work efficiencies and ideally, in a better match-up of supply and demand, giving the publisher a better price and the advertiser a more relevant audience.

No matter how you look at programmatic, it dominates the digital ad ecosystem. Just as mobile now accounts for the majority of digital ad spending, programmatic buying accounts for the majority of transactions. And one particular segment of mobile – mobile video – is poised to continue growing at a breakneck pace over the next three years, which is why the MMA is publishing this guide. According to eMarketer, it will total $15.93 billion by the end of this year and $24.81 billion in 2022. The digital ad marketplace is moving toward being dominated by mobile and programmatic and video.

Some current statistics on mobile, programmatic and video: • Mobile will contribute the most to global ad growth by 2020, at $72.6 billion, far ahead of any other platform. TV comes in second, at only $6.9 billion (Zenith). • The global programmatic ad marketplace was expected to hit $34.1 billion by the end of 2018 (Magna Global). This is an upward revision from an earlier estimate of $31 billion. • Ad spending in mobile programmatic video was anticipated to rise to $11.83 billion by the end of 2018, representing 52.5% of all domestic programmatic video spending (eMarketer). (Part of this total is due to a reclassification by eMarketer of social in-feed video into the digital video category.) • U.S. mobile Internet users spent 88 percent of their time in-app in December 2017. • As the chart below demonstrates, in August 2017, eMarketer predicted that digital video advertising would reach $15.4 billion by the end of 2018, but later upped that estimate to $17.9 billion by year-end 2018 (eMarketer).

Programmatic advertising accounts for so much mobile video spend that the marketplace is almost reaching a point where if you’re buying digital video, you’re buying programmatic.

But here’s the rub. For such a dominant marketplace, the speed with which it has grown means there are still many extant issues that need to be identified and addressed so that buyers and sellers can fully trust this channel. If the size of the market has reached scale, the solutions haven’t.

Some prevailing industry standards are insufficient, while others are being modified, but have yet to achieve scale – and much of the market is a black box, particularly when it comes to mobile inventory providers and their advertising software development kits or SDKs. (All terms in blue are defined in the glossary appending this report that can be viewed here.) The goal of this report is to crack open the black box, bringing some level of understanding and transparency to the marketplace.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 3 Mainly focused on the U.S. market, it will unpack, explain, and provide order so that brands, and their partners, can better plan and optimize their mobile ad budgets. Mobile inventory providers are fully supportive of this effort to demystify how they operate, since transparency is key to opening the market to more spending by brands.

The accompanying Mobile Programmatic Video Buyer’s Guide spreadsheet, which is being crowdsourced by members of the ecosystem, aims to help advertisers understand features and functionality of different SDKs, detailing their scale, supported verification tools, video formats and more. We encourage readers of this guide to contribute. Ultimately, this will help the marketplace move towards standardization, allowing more scale for measurement and standardized ad formats.

As a side note, we should mention that eMarketer defines programmatic as anything transacted through automation, which includes the mobile video giants YouTube and ; that definition clearly helps to push the overall programmatic numbers skyward. The majority of this guide will deal with mobile programmatic video transacted through DSPs (demand-side platforms) and trading desks, in other words, inventory that is outside of those two proprietary media silos. This still represents a very significant chunk of inventory, and includes direct buys through premium video providers including the major TV broadcasters.

For more on programmatic, here are some additional resources: • The MPV glossary addended to this report • OpenX’s Core Concepts of Programmatic • The Media Crossing video 200 Milliseconds: The Life of a Programmatic RTB Ad Impression

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 4 SECTION 1: Why Mobile Is Not Like Desktop

When advertising moves from desktop to mobile, it does not just mean a shift from one device to another. There are a number of mobile-specific issues which we will discuss here:

1. Mobile Has Two Ad Environments: Mobile Web and In-App 2. In-App Inventory Partners Play a Significant Role 3. Mobile Programmatic Is Still Evolving

1. Mobile Has Two Ad Environments – Mobile Web and In-App. The most familiar mobile ad environment is mobile web, which – on the surface – is directly analogous to the desktop advertising experience because ads appear in a browser. The second, in-app, is completely different; instead of an ad being viewed and served in a browser, the code that loads and manages it is baked into an app. Since it does not scale for every app developer to write their own code to retrieve and display ads, mobile inventory providers entered the picture and developed advertising SDKs. These are libraries of code that developers integrate into their apps to support monetization through advertising.

The in-app environment has its own unique challenges. It slows the pace at which the ecosystem can change; to rollout new features, a mobile inventory provider needs to put the features into its SDK, compel the apps using its SDK to update to the latest version, and wait for the updated app to be downloaded and installed by users so that it can be implemented.

2. In-App Inventory Providers Play a Significant Role. This universe of in-app inventory providers has created an additional layer within that ecosystem that is not well understood or even visible to many advertisers, but which ultimately completely controls the ad experience.

Here’s how it works:

For desktop programmatic, the advertiser works through a DSP, possibly with a trading desk to actually execute buys. That DSP will make inventory available on different websites, each of which may or may not be transparently disclosed to the advertiser. While there is a layer of code to govern the playback experience in video, the browser itself handles a lot of the work, often using standardized player technologies like JW Player or Brightcove. Compared to in-app, the playback environment is relatively transparent.

For in-app, the DSP requests inventory from the SDK of a “black box” inventory provider, and that SDK entirely manages placement, playback, and the role of the SSP (supply-side platform). The SDK may also assume the role of the exchange where a mediation layer is involved. The fact that these SDKs differ in supported functionality, along with practices and standards, and other aspects – unlike the web browser used for mobile web – results in a more complicated ecosystem for buyers.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 5 3. Mobile Programmatic Is Still Evolving. It’s probably abundantly clear that since the majority of mobile is app-based rather than web-based, it leads to issues with leveraging mobile programmatic inventory. In addition to the difficulties in mass adoption of SDKs, the mobile ecosystem has yet to support real-time bidding (RTB) across all inventory. Additionally, connectivity has been more of an issue in mobile, which is why the in-app market has had to bring the role of the exchange into the app itself by way of a mediation layer. The mediation layer references the set of cascading business rules (called a waterfall) to decide between inventory provided by ad networks that are plugged into apps.

Fortunately, the market is starting to embrace real-time decision-making in-app, as connectivity has improved. Advanced bidding solutions from a number of mobile in-app inventory providers now incorporate innovations from the desktop RTB space, including header bidding – which allows publishers to offer their inventory for bidding to ad exchanges at the same time – instead of using waterfalls. There are also new auction pricing methods.

This has increased availability of mobile inventory. Rather than working with a handful of known networks, advertisers can access inventory from all DSPs and networks. The downside is that this also opens the door to more re-selling and arbitrage.

Buyers can more easily buy mobile, but also more easily go astray in the process; they need to be aware of where their ads are running if they choose to bid on the open exchange. PMP (Private Marketplace) deals are one mechanism buyers increasingly use to ensure inventory quality. As is the case throughout the advertising ecosystem, any steps toward inventory transparency (see “Section 7: Avoiding Fraud“ - on page 18) are likely to be beneficial.

Now that we’ve outlined what makes the mobile ad ecosystem fundamentally different than desktop, let’s discuss how this affects the experience for consumers and marketers, starting with its ramifications for creative – and then moving on to how the ecosystem plays out when it comes to media buying, data, and measurement.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 6 SECTION 2: Video Formats, Playback and Technical Standards

In digital video – particularly mobile – all video ads are not alike, because size, sound, and how the video plays can vary. This section details both back-end and front-end variables.

Mobile represents a wider variety of video variations than any other class of device. Here’s a short list:

1. Video size can range from full-screen to quite small. 2. There are a wide variety of playback experiences and formats, ranging from auto-expanding to muted. (For complete details on these, check out the MMA’s Buyer’s Lexicon for Mobile Video.) 3. There are three main technical standards – VAST, VPAID and MRAID – in addition to proprietary offerings from many providers.

Any variation can affect the experience, so it can be a challenge to understand the outcomes of using a given type of inventory, which in turn makes it difficult to buy inventory that supports campaign goals. An accidental screen touch can make an ad play unintentionally; mobile auto-play video is more likely to interfere with device performance than on desktop or certainly OTT (Over-the-Top). Additionally, the mono-tasking nature of mobile devices makes it more likely that video will intrude on activities like calls or listening to music, or be skipped, muted, or otherwise dismissed by the user.

There are many variations in mobile video placement and size.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 7 Understanding How the Experience Affects KPIs It can be difficult to generalize how the variation in playback experiences can impact campaign goals, since a particular placement can have multiple impacts. Here are some specific examples, albeit with speculative results:

EXPERIENCE RESULT

Auto-start May drive completions up, but annoy the consumer

Muted May affect clarity of message

Pop-up or Scroll-by May negatively impact viewability and completion

Rewarded Drives completions up; opt-in may have positive brand association

The catch-phrase “Your results may vary” definitely applies here. While the performance of advertising is always some combination of content, format and context, the variables therein can make coming up with a list of standard caveats impossible. A clear and visually compelling video may be able to meet an advertiser’s KPIs even when muted, while another ad may fail. The success of rewarded video has everything to do with how compelling the reward is, and so forth. All video formats – in-stream and out-stream – have value, and it’s up to the advertiser to figure out the alchemy that results in their videos meeting KPIs.

VAST, VPAID and MRAID: Pros and Cons of the Three Technical Standards

On top of those front-end variations, it’s necessary to layer on even more complexity in the form of digital’s three technical standards for delivering video ads - VAST, VPAID, and MRAID. Each have their own pros and cons, and are not available across all providers. Below are brief descriptions:

VAST (Video Ad Serving Template) is a structured text document that gives video players instructions on how to play a video ad, offering details such as which file to play, how it is supposed to render, its duration, and its skippability. It has come into broad demand because unlike VPAID, it does not require JavaScript.

VPAID (Video Player Ad-Serving Interface Definition) provides the video file wrapped in a layer of code to be executed within video players. This makes it possible to layer on overlays or features that encourage interactivity, as well as to run code that allows advertisers to better measure effectiveness. It is mostly available in mobile web.

MRAID (Mobile Rich Media Ad Interface Definitions) is a similar layer of code to VPAID built specifically for mobile. Its purpose is to support rich media ads rather than video, and it now incorporates VPAID events to make it possible for video ads to run in a more standard fashion in what were previously rich media placements.

Agencies and advertisers typically find themselves between a rock and a hard place regarding these standards. They have to choose between much more prevalent VAST inventory, sacrificing viewability and verification in the process; or more limited VPAID inventory, which supports viewability, but may

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 8 well be a substandard viewer experience because the additional code can increase load times on mobile devices. Even within the context of individual standards, formats can also vary based on the publisher or exchange, and then there are the variables introduced by whether an ad is in-app or mobile web.

Fortunately, things are evolving in a positive direction. Thanks to the OM SDK (Open Measurement SDK), viewability and verification measurement is making its way to VAST inventory on in-app, which represents a lot of the available inventory.

Produced by the IAB Tech Lab and initially led by measurement provider Integral Ad Science, OM SDK is a solution that takes the bulk of the code needed to measure viewability and verification and moves it to operate within the app itself. Its supporters include Comscore, DoubleVerify, , and Oracle’s Moat – and it is being rolled out across the mobile in-app ecosystem. As of December 2018, seventeen companies across the U.S., Europe, and Asia have received OM SDK certification, including Google, InMobi and Pandora. It is also enabled on approximately two billion devices and supported by the Media Rating Council. (Read more about that in “Section 6: Getting a Handle on Measurement” on page 16.)

Still, as referenced earlier, it can be difficult to get enough information to understand the playback experience for the ads you’re running. You can buy from publishers where you know the experience, or rely on the data available via OpenRTB today, which provides some signals like skippability. (As the name implies, OpenRTB provides open standards for communication between buyers and sellers of publisher inventory that is bid in real-time.) A late 2018 update, OpenRTB 3.0 has been released “to meet the market’s demand for security, transparency, authentication, and trust in programmatic advertising.” The playback experience is typically reported separately by each DSP rather than across all inventory.

The latest revision to VAST – VAST 4.1, which was released in November 2018 – proposes the use of standardized macros to be filled in by the publisher player when an ad is requested. This should provide not only much more detail on the playback experience, but also provide it further downstream, making it more likely the information will be accurate; it should also provide a level of standardization.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 9 SECTION 3: Understanding and Leveraging Creative Options

The mobile programmatic video marketplace is highly fragmented, which means not all creative formats are available on all inventory. This section is designed to give marketers some guidance on how to manage the nexus between formats and inventory. Vertical and Square Video While Snapchat pioneered vertical video, the format has become increasingly available, at least within the so-called “walled gardens” of Google and Facebook. Though the specs don’t always adhere precisely to the 9:16 aspect ratio (for instance, sometimes the format is square, at a 1:1 aspect ratio), it is now accepted on Instagram and Facebook, and as of September 2018, YouTube. The adoption aligns with the fact people use their mobile devices vertically, and programmatic standards should more openly embrace these trends with asset flexibility beyond the walled gardens. (For more on vertical video, check out the MMA report “A Different Perspective: Demystifying and Simplifying Vertical Video for Marketers.” Interactive Video There are three options for running interactive video on mobile. 1. You can buy VPAID inventory, which, as was said earlier, is more available in mobile web than in-app. (For the few inventory providers who support VPAID in-app, check the Mobile Programmatic Video Buyer’s Guide spreadsheet.) If you are running in mobile web or any placement that is not close to full-screen, be aware of the placement size and other aspects of the playback experience that can impact the viewer’s ability to engage with the unit. For in-app, make sure you are not affected by latency issues. Targeting Wi-Fi-connected inventory can overcome that.

You may be able to run MRAID units on mobile-first inventory providers. Since they are technically rich media placements, they are becoming less popular for running video. As with VPAID, an MRAID video can end up in different placements and is subject to latency issues, so understand the playback experience of the inventory provider you are using, as well as making sure you get the metrics you might need.

2. VAST can also deliver interactive video. The VAST end card format is most popular because it loads an HTML VAST companion as a full-screen element after the video has completed. There is work being done to standardize this approach of supporting interactivity via different uses of the companion and potentially other VAST elements. These elements, referred to as “VAST interactive templates,” are part of VAST 4.1.

3. The last option – proprietary formats supported by one particular inventory provider – can provide unique functionality, but they are inherently limited in scale.

Units running on mobile can drive higher levels of engagement and single-person devices – phones or tablets – are ideal for personalization. Buyers need to challenge inventory providers to agree upon basic, scalable mobile-interactive video formats like the VAST end card in addition to innovative proprietary features.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 10 Dynamic/Personalized Video The objective of creating and delivering dynamic or personalized video is no different than delivering it across any channel. These videos are more relevant, and personalized content yields better performance. However, as was hinted at above, getting to a fully scalable solution is slightly longer and more complex.

There are a couple of options to execute dynamic video today. However, to get to a scalable and efficient machine that will crank out dynamic videos on-the-go, a number of tools needs to be brought together, including video authoring and asset production, dynamic video advertising strategy and tactics, and translation of these assets into actual creative. (See “Building Creative Capabilities That Can Achieve Scale” in “Section 4: Buying Challenges and Caveats” on page 14.)

Knowing your starting point and availability, as well as the depth of dynamic video layers or changes is a good place to start:

• Fully-baked dynamic video versioning: In this case, the dynamic elements are rendered into the video itself, and it can be run and targeted pretty much anywhere. While scale is not a problem, the main barrier comes from the availability of assets and production time. The big positive for this approach is its ability to run on any screen. Negatives include the lack of support for responsive layout (the video is fixed once rendered, so you cannot support the same asset across different aspect ratios), and difficulty in supporting real-time data. There are two ways to execute this type of dynamic video: 1. Shooting video across all dynamic versions. A big, but not impossible, leap for some brands, it requires careful pre-shooting planning. 2. Augmenting video with motion graphics dynamic elements. This is easier to achieve and can be executed via post-production tools such as Adobe After Effects with extensions coming from various providers.

• VAST-based personalization: If you have a limited number of video versions, or if you want to better target, you can replace a whole video within the same VAST tag. This is easy to execute and can be done at scale via feeds. All creative versions must be ready prior to launching the campaign in order for the most relevant creative to show, based on the dynamic triggers.

• VPAID overlays: Even without a full personalized production strategy, you can take advantage of personalization by utilizing VPAID rich overlays, which are capable of generating data-driven, interactive, creative experiences based on a set of contextual data triggers. VPAID inventory is more limited, but it does support real-time dynamic capabilities as well as interactivity.

• VAST end-card personalization: Another way to take advantage of dynamic video is to personalize the message following the video, by utilizing the companion standard as an end card. Positives here are that it can pair a fully- responsive, dynamic or interactive message with a standard video, essentially following the video asset with a rich media ad. On the negative side, in-app inventory is limited. Distribution is heavily dependent on each SSP’s ability to support it in its SDK, and only a few (albeit major) SSPs support it today.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 11 Playables • These are highly interactive ads that, for instance, allow a user to try out a game within the advertising experience and then usually give the user the opportunity to install it. Playables are not video, per se, but mentioned here because – as a carryover from app-install advertising – they rival video and some rich media in providing sight, sound, and motion. Additionally, at least among app download advertisers, they are catching up with video as the most effective marketing channel. According to AdColony, they rank fifth as an effective channel or format among app developers (see the chart at right).

So far brand advertiser interest has trailed well behind that from app marketers, but brands may find them useful if they are demoing an app or if they have a strong interactive concept they want to execute without video.

Rewarded (or Opt-In) Video Rewarded video is more a type of inventory than a type of creative, but because brand advertisers are beginning to see its value and it is featured by some inventory providers, it is worth discussing.

Like playables, rewarded video is a carryover from app-install advertising, specifically in games. Users opt-in to watch a video, and then the marketer gives him or her a reward typically by providing some kind of currency (an extra life in a game, for example). Although in-app advertisers have historically been wary of gaming inventory, the opt-in, non-interruptive experience is what is attracting interest from brands.

A 2018 study by OpenX found that 70% of US mobile users had watched a rewarded or opt-in video recently. Of that group, 80% said they preferred the video ad-experience to other types of traditional video ads. This preference from consumers, paired with above industry-standard performance metrics in areas such as viewability and completions, are what makes rewarded video a compelling solution.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 12 SECTION 4: Buying Challenges and Caveats

For marketers, most of the discussion of technical standards, playback experiences and in-app vs. mobile web really boils down to the essentials: What should buyers know before they even begin the process of looking to purchase mobile programmatic video?

Even after making decisions on creative options, data, measurement, KPIs, and price, it can still be challenging to execute the buy. This is true especially if you’re looking for specific features, such as vertical video or a VAST end-card allowing for interactivity – or you require measurement in VAST in- app inventory which uses the OM SDK.

In some cases, specific buys can be enabled on the open exchange using relevant OpenRTB flags, or you may need to rely on Deal IDs. Some inventory providers will have evergreen package IDs set up to access specific inventory, while for others you may need to set up a PMP deal to get everything you are looking for.

It’s probably obvious that the complexities of the mobile programmatic video marketplace present challenges in reaching your audience at scale while meeting all of your other objectives. Let’s outline the major ones:

Limited VPAID Inventory In-App: The first challenge is the limited availability of VPAID inventory in-app, which – as was said earlier – is usually required by advertisers to provide viewability and verification. In mobile web, it is more likely to be associated with a variety of different placements, not all with the best user experience. While out-stream definitely has value, its very definition – i.e. everything that is not in-stream – requires an informed buyer to be sure to understand exactly what playback experience they are buying.

There is an ever-changing relationship between VAST and VPAID inventory. While VAST has almost always dominated, its percentage of inventory has seen some dramatic shifts.

The dominance of VAST inventory dovetails with consumer time spent in-app and the dominance of mobile in-app video ad spending, which has grown from 63.2% domestically in 2017 to an expected 72.9% in 2019. (NOTE: In-app’s share may look disproportionately large in the chart above since it includes social inventory, but there is little argument that in-app spending exceeds mobile web regardless).

Interference from Pre-Bid Verification: A second challenge is that pre-bid verification can sometimes interfere with the delivery of in-app inventory because the verification tool is missing signals it would typically get from web inventory. It also interprets the absence of these signals as a trigger to block the impression. These tools will likely get better, but for now, check with the in-app mobile inventory provider, and turn off pre-bid verification if necessary.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 13 Targeting Parameters That Reduce Scale: Of course, any limiting factor on your inventory will reduce scale, and having multiple limiting factors pares back inventory quickly. For example, if you build an audience using the following criteria: gaming inventory, targeted only to specific device users or Wi-Fi-connected users at a specific location – and additionally can only buy inventory measurable by a given verification partner – your sweet spot of inventory is now the (small) Venn diagram sum of only partly overlapping inventory pools.

Unavailability of data also limits scale if you are targeting inventory that relies on it. The most pervasive data problem in mobile is the inability to obtain a unique identifier, since the Device ID – which allow marketers to identify individual consumers anonymously – may not always be received from the publisher, and cookies are frequently disabled or expire quickly. According to a Q4 2017 study from ad serving company Flashtalking, some 75% of mobile web cookies are blocked or deleted, and in-app typically relies on the inventory provider to pass an ID. If you don’t have that identifying data, any strategy specific to individual users will require you to decide whether to sacrifice scale over certainty, or push for scale even when the ability to identify the viewer is a “maybe.” “Section 5: Leveraging Data” on Page 5 provides some insight into situations where data availability will be compromised.

Building Creative Capabilities That Can Achieve Scale: Another challenge is achieving scale if your campaign focuses on sophisticated, personalized creative. In fact, adding media partners and channels across various types of inventory typically requires so much creative adaptation that brands fall back to static creative or stick with standard inventory instead. While we discussed particular formats and how well they can be scaled in “Section 3: Understanding and Leveraging Creative Options,” the first requirement is an organizational approach that allows marketers to have creative capabilities that can leverage scale. Marketers need to:

• Bring the media and creative teams closer together. Collaboration and planning, ideally supported with data, is key to meaningful scale. • Make the creative process more efficient, so the focus can be on sophisticated solutions. • Onboard a creative management platform to scale and organize teams. It should include: - Built-in requirements and capabilities across digital, native, video and social. - Easy re-use of assets and creative concepts across inventory. - Centralized reporting and analytics for creative optimization. - Universal tagging for fast and efficient trafficking. - Automated versioning and creative adaptation.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 14 SECTION 5: Leveraging Data

Deterministic vs. Probabilistic Data Any conversation about leveraging data should probably start with the two types of data that are typically used to power data-related advertising.

Deterministic data is a type of first-party data that definitively identifies a user. It includes log-in data, as collected by Facebook, Google and other entities that require a log-in; or data such as sub- scriber information from cable or other companies that can confirm a user’s identity. The scope of first-party data can range from a few data points to many – including age, gender, location, buying habits and so on.

Probabilistic data, as the name implies, uses data to make statistically-informed guesses. Using any number of data points, including a Device ID, IP address, and location, companies can employ sta- tistical models to guess which impressions are connected to the same user. For example, if a certain Device ID is regularly associated with the same home Wi-Fi IP address in the evening, and a set of IP addresses from another location during the day, a device graph company can speculate this is the same user, and then model the likelihood that other Device IDs associated with the home IP address also belong to the same user.

Deterministic data is the more valuable of the two. Facebook and Google’s access to login data for large swaths of the global populace has played a key role in propelling them to ad revenue leader- ship. Amazon, their most serious challenger, possesses the same kind of data, combined with access to conversion data from its own site, which accounts for approximately 50% of all U.S. online sales.

Outside of the “walled gardens” it is hard to get deterministic data at scale, which is why probabilistic data remains a major resource for most marketers. Location Data Location data is usually deterministic – since consumers opt-in to share the location data on their phone, or within individual apps, and it is most effectively gathered either by companies that provide location-based services, which then resell it to data services companies, or by companies like Face- book and Google, which might have users that are logged in to a running app for much of their day.

While theoretically location data is among the most useful for targeting, it still suffers from inaccu- racies. According to an analysis by Ericcson Emodo’s Institute, as much as 45 percent of location data is inaccurate. Meanwhile, sellers of programmatic inventory know that impressions that include location data can fetch a higher price. So although programmatic impressions for sale via OpenRTB frequently contain location data, it can be hard to understand how accurate it is. A transmitted lat- itude and longitude may end up mapping to the middle geographic point of a country, state or city because that was the extent to which the location could be determined. On the other hand, location data that comes from beacons within stores is on a much more granular level. It is smaller in scale – and relies on Bluetooth – but can be useful, particularly when cross-indexed with things like credit card data.

Marketers need to understand how the location is determined, discover what the data provider is do- ing to ward against inaccuracies, and, if possible, match the data to an accurate truth set.

It is being used in increasingly powerful and innovative ways for both location and predictive analyt- ics. For more on marketer uses of location data, check out the MMA report “Checking In: Use Cases for Location Data Beyond Geo-Fencing”.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 15 SECTION 6: Getting a Handle on Measurement

Accurate measurement is an issue throughout the advertising ecosystem, but mobile programmatic video comes with its own set of concerns. Getting Viewability for In-App Inventory As mentioned above, the first and likely biggest challenge agencies and advertisers face is the ability to get consistent viewability and verification measurement, especially in VAST inventory, which dominates in-app and mobile inventory. The media agency Hearts & Science even coined a term - “the unreachables” - for a new generation of viewers who have moved away from TV to mobile in- app and have dropped off the measurement grid.

Fortunately, as you’ll see below, the industry has done important work to improve the viewability and verification measurement for in-app inventory. But first let’s look at how these difficulties evolved in the first place.

VAST, of course, has dominated in terms of scale, because it does not require JavaScript. Different mobile environments have different requirements regarding the use of JavaScript, meaning VPAID will not reliably run in the same way on all devices. In addition, its extra code layer has historically enabled viewability, but increased the load time on devices, especially intermittently connected devices like phones.

Some entities within programmatic manipulate VPAID code to run what is called client-side arbitration. Basically, in real-time, an entity buys the impression and then tries to re-auction the impression for more money via an empty VPAID tag, and so on, in a daisy chain. In addition to gaming the whole programmatic ecosystem, this introduces a ton of latency for the publisher and viewer.

For all the reasons above and a few more, the mobile in-app ecosystem has been resistant to adopting VPAID, with the notable exception of a few inventory partners who have made it function, even in mobile in-app. Agencies and advertisers have either shied away from in-app or accepted a hit to viewability thresholds to get the additional reach that mobile in-app inventory provides.

Fortunately, OM SDK is replacing a patchwork solution in which several proprietary SDKs were being developed in parallel to work with mobile video inventory providers.

We hope the MMA Mobile Programmatic Video Buyer’s Guide spreadsheet will help track the adoption of viewability across in-app video inventory providers; use it is a starting point before you speak to inventory sources directly (and, as was said earlier, if you can provide your own insights, please contribute to it by filling out the form at the bottom of the guide).

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 16 Measuring Conversions Since cookies aren’t available in-app and usually expire quickly in mobile web – and connectivity can be somewhat spotty in mobile – it can be difficult to measure conversions, as well as other metrics that rely on associating an impression with a given user. Unless you have access to deterministic data (see “Section 5: Leveraging Data”) connecting an in-app click with a conversion on a website requires having a device graph that can connect the Device ID from the in-app click with a cookie ID from the website. This also presumes that both the Device ID and cookie are available for the publisher in question.

Device IDs can be retrieved through installed apps, where they can track actions including pre- install engagement and transactions that occur in-app, from purchases to game activities. As such, they are key to tracking the user journey and offer details on other things as well, such as audience segmentation. Through ID matching, the behavior of users can be aggregated and analyzed.

While Device IDs aren’t perfect – users can block or limit tracking, for instance – they are far more persistent than cookies, giving marketers a much longer timeline into customer activity. Collecting Device IDs is more complex than most people are aware of; they can potentially be hashed (encoded) to make it harder to connect them, and the mechanism for passing Device IDs around the mobile ecosystem has yet to be fully standardized.

For more information on tracking conversions, check out the MMA’s MTA (Multi-Touch Attribution) Tactical Success Guide.

For marketers looking to measure app downloads in the Android or Apple app stores, it’s possible to work with partners like Tune or Kochava, which specialize in app-download attribution. In those cases, the conversion provider’s SDK must be present in the downloaded app, so it can send the Device ID once the app is installed. Measuring Reach and Frequency Measuring reach and frequency remains a challenge outside of the walled gardens, especially across devices. While we’ve already mentioned that cookies expire particularly quickly in mobile web, things have become even more challenging with Apple’s introduction of new levels of browser privacy enforcement (specifically ITP - Intelligent Tracking Protection - 2.0) alongside the rollout of other privacy-centric initiatives like EU’s GDPR (General Data Protection Regulation). As was said earlier, Device IDs are optimum, but have challenges.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 17 SECTION 7: Avoiding Fraud

Fraud, of course, is of paramount concern no matter what part of the marketing ecosystem a compa- ny resides in.

There are several different flavors of fraud that can plague mobile ad campaigns, including in media, when the ad isn’t shown where reported; in the device, when the device isn’t what’s reported, and in location.

Most participants in the ecosystem work with one or more measurement partners to mitigate fraud concerns. These different vendors have released – and are continually expanding on – their ability to identify various kinds of fraud using their own proprietary techniques. Recently, three technical stan- dards have also been introduced with the support of the OpenRTB Working Group.:

1. Ads.txt: This is a text file that publishers can put on their servers that helps protect against unauthorized sale of inventory. 2. App-ads.txt: A version of ads.txt for in-app, it was released in beta in November 2018 and was open for public comment through February 2019. 3. Ads.cert: Closely related to Ads.txt, ads.cert – currently in beta – uses cryptography to “sign” bid requests, verifying the premium inventory – like video – available through sellers listed in ads.txt.

A discussion of their promises and challenges lies below:

1. Ads.txt: Launched in May 2017, ads.txt is essentially a whitelist of authorized sellers and resellers, to ensure that ads don’t end up running on shady inventory that may appear to be legitimate at a glance (e.g. espn.sports- news.com). Ads.txt only whitelists domains (i.e. website addresses), which is why, its companion, app-ads.txt, was developed.

Other flaws exist – for example, ads.txt does not always identify inventory type (display or video), allowing sellers to game the system by selling display inventory as video for higher CPMs. Fraudsters can also make their way onto the ads.txt whitelist by convincing someone in the organization that they are a legitimate seller. From a buyer’s standpoint, the existence of ads.txt is nevertheless a big win, and adoption continues to grow. The ads.txt aggregator currently tracks more than two million domains, and, according to Pixalate, adoption had surpassed 77% as of September 2018 among the top 5,000 most popular websites. Its rapid adoption means that buyers should insist on ads.txt for web traffic.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 18 2. App-ads.txt: Since this was just being released as this paper was being published, it’s too soon to predict how universally app-ads.text will be adopted. While giving marketers some assurance in-app should seem a no-brainer, in-app is different from mobile web here too. For one, advertising is the lesser revenue stream for many popular game apps which may lessen the sense of urgency for adoption. Also, some argue there is far less need for an ads.txt whitelist of legitimate apps because new apps go through an approval process already to be accepted into one of the app stores. So far, the Apple and Google app stores haven’t implemented the metadata tags that would make implementation easier, though there is a workaround.

3. Ads.cert: By using cryptographically signed and authenticated bid requests, ads.cert aims to combat ad fraud from a different angle. This automated solution validates both the seller and impression at each stage of the supply chain, ensuring they cannot be manipulated. One caveat is that ad.cert’s authentication features only work with the OpenRTB 3.0 spec, which officially launched in November 2018. Its success depends on infrastructure upgrades among ad tech firms, and at this writing, only months after its release, there is no data on adoption. It makes sense for buyers to support OpenRTB 3.0, among other transparency initiatives, but the amount of work involved to roll it out, means it is likely not a near or even mid-term solution.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 19 CONCLUSION Making Mobile Video Programmatic Work for Your Brand

As this report makes clear, the mobile video programmatic marketplace has scale, even if it may still be challenging to successfully meet all your campaign goals. While the industry works on solutions that will standardize and streamline the various components of the marketplace – media buying, creative, data, and measurement – marketers need to master the market as it is now, instead of waiting for it to be easier. To review, here are some topline recommendations:

1 Know the differences between mobile web and in-app ad placements.

Be aware of the variations in the playback experience and how they may affect 2 your KPIs.

Educate yourself on the differences in what the three video advertising technical standards – VAST, VPAID and MPAID – can deliver in terms of measurement, and 3 what they can’t.

Familiarize yourself with how to account for measurement gaps that arise from the 4 intricacies of this marketplace.

Learn about the various industry standard initiatives and how they can help you 5 combat fraud.

Acquaint yourself with the various creative formats that can make video stand out on 6 mobile.

The mobile video programmatic marketplace has many benefits as an automated method of buying advertising that can be targeted toward your customers. By keeping this report and the Mobile Programmatic Video Buyers Guide close at hand, marketers should be able to master it.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 20 MOBILE PROGRAMMATIC VIDEO GUIDE GLOSSARY

Ads.cert: Like ads.txt (see below), the objective of ads.cert is to ensure that impressions marketers get via ad exchanges are actually from the website where they intend to buy them. Currently in beta, ads.cert uses cryptography to “sign” bid requests, giving marketers a better idea of what they are buying, and verifying the premium inventory – like video – available through sellers listed in ads.txt. Whereas ads.txt uses a simple whitelist approach to eliminate undesired sellers, ads.cert intends to correctly represent a particular seller’s inventory.

Ads.txt: Launched in May 2017, ads.txt is essentially a whitelist of authorized sellers and resellers, to ensure that ads don’t end up running on shady inventory that may appear legitimate at a glance (e.g., espn.sports-news.com) but isn’t. Ads.txt only whitelists domains (i.e., website addresses), which is why, its companion, app-ads.txt (below), is in development.

Apps-ads.txt: Released in beta in late 2018, apps-ads.txt is the in-app version of ads.txt. It was open for public comment through February 2019. At this writing, It is simply too early to predict how universally app-ads.text will be adopted, though one school of thinking is that verification is less necessary in the more controlled in-app universe.

Deal ID: A Deal ID is unique to a particular deal between a buyer and seller, allowing the two parties, even in an automated marketplace, to set parameters on their dealings with another. These parameters can take any number of forms from minimum prices to campaign objectives, particular audiences, or viewability thresholds.

Device ID for advertising: This an anonymous identifier, used specifically for digital advertising, that is unique to every mobile device. As such, this type of device ID is useful to marketers because it allows them to track user behavior. Each mobile OS has its own device ID for advertising. In iOS, this is called an IDFA (identifier for advertisers), and for Android it’s AAID (Android Advertising ID). Advertising device IDs, which allow for user control, were developed to deal with privacy issues that arose from marketers using the immutable baseline device ID that is permanently associated with a particular device.

DSP: DSPs, or demand-side platforms, are utilized by agencies and advertisers to help automate the buying process. Customarily, they aggregate impressions across sites, repackaging them according to user data. DSPs buy the ad impressions through ad exchanges, and most of the time inventory is bid on in real-time. They are an evolution of the ad network model in that they usually bundle buying, serving, and tracking.

End card: End cards display additional content at the end of digital video ads that give users the option of further interacting with the brand or content.

GDPR: Short for the General Data Protection Regulation, this is new European Union privacy legislation, enacted in May 2018, which gives more data protection and rights to individuals. Even though it is not law outside of Europe, the global nature of the Internet has meant that it has affected businesses throughout the world. The California Consumer Privacy Act promises to bring some of the same changes directly to the U.S. by 2020.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 21 Header bidding: Header bidding developed as an alternative bidding technique to the waterfall method, in which inventory is put out for bid sequentially until it sells. Instead, header bidding allows publishers to offer inventory to multiple ad exchanges at the same time. The term comes from the fact that the bidding code lives in the website page header. Demand partners are called simultaneously from the visitor’s browser and the winning bidder gets the impression.

In-Stream: In-stream is a mobile video ad format that appears in front of (pre-roll), during (mid-roll), or after (post-roll), video content. In-stream units mimic the look and feel of TV commercials within a video player.

Mediation Layer: A mediation layer is essentially an ad exchange that “lives” inside an SDK, allowing it to connect with multiple ad networks. Its role is to help in managing the waterfall method of bidding; the mediation layer operates off the waterfall’s set of cascading rules that prioritize networks based on ad rates.

MRAID: MRAID stands for Mobile Rich Media Ad Interface Definitions, and is the API used by rich media ads to communicate with the advertising SDKs embedded in apps. Essentially, MRAID lets apps show rich media ad formats within mobile devices. Later versions of MRAID have incorporated VPAID events to support video ads as well, although over time most mobile video delivery has moved to VAST or in some cases VPAID.

OM SDK: Short for Open Measurement , the OM SDK, which launched its first version in April 2018, “is designed to facilitate third-party viewability and verification measurement for ads served to mobile app environments without requiring multiple ad verification service providers’ (Measurement Providers) Software Development Kit (SDK),” according to the IAB Tech Lab. An industry-wide effort, its development is meant to be used by Integration Partners (app publishers and ad SDK developers) and Measurement Providers. For more on Open SDK, go here.

OpenRTB and OpenRTB 3.0: OpenRTB was a standard that launched in 2011, as an open protocol making real-time bidding possible at scale. Though it has evolved since 2011, OpenRTB 3.0, which officially launched in November 2018, is the most significant revamp of OpenRTB, addressing widespread industry concerns about trust in the programmatic marketplace. That said, at this writing, there is no hard data on its rate of adoption.

OTT: An abbreviation for over-the-top, OTT is TV and movie content, like Netflix, that is delivered over the Internet rather than through satellite or cable providers.

Out-Stream: Out-Stream is a sub-group of video ad types that are defined as NOT in-stream (i.e., pre-roll, mid-roll, etc.) and play as standalone videos without any video content before or afterward. Types of out-stream videos include in-article, in-banner, and in-feed.

PMP: An abbreviation for Private Marketplace, a PMP is an invitation-only auction for media buyers offered by publishers.

Pre-bid verification: A method for ensuring brand safety, pre-bid verification essentially vets for-sale ad inventory for a combination of viewability, brand safety and/or fraud to make sure it is a suitable place for that advertiser’s ad placement – depending on its parameters – before bids are made.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 22 Programmatic media: Programmatic media is the term for the automation of ad buying and selling, resulting in work efficiencies and ideally in a better match-up of supply and demand, giving the publisher a better price and the advertiser a more relevant audience.

RTB: RTB, or real-time bidding, is the term for the real-time auction that takes place between ad buyers and sellers.

SDK: Short for Software Development Kit, an advertising SDK is a library of code that developers integrate into their apps to support monetization through advertising. Since it does not scale for every app developer to write its own code to retrieve and display ads, SDKs are the province of mobile inventory providers, which can deliver scale.

SSP: An abbreviation for supply-side platform, an SSP is an automated ad sales platform, and effectively the flip-side to a DSP. SSPs are utilized by publishers to connect to other parts of the digital media ecosystem – from ad exchanges to networks to DSPs – simultaneously, opening up inventory to many buyers.

VAST: (Video Ad Serving Template) is a structured text document that gives video players instructions on how to play a video ad, offering details such as which file to play, how it is supposed to render, its duration, and its skippability.

VAST 4.1: A substantial update to the VAST technical standard, VAST 4.1 was officially released in November 2018. Its most prominent feature is that it supports server-side ad insertion and verification without VPAID, and also supports standardized macros, integration of audio ads, and a number of other features.

VPAID: (Video Player Ad-Serving Interface Definition) provides the video file wrapped in a layer of code to be executed within video players. This makes it possible to layer on overlays or features that encourage interactivity, as well as to run code that allows advertisers to better measure effectiveness. For mobile, it is mostly available for web inventory, and not in-app.

Walled gardens: Platforms like Google and Facebook, which tightly hold their users’ data, are referred to by the digital marketing ecosystem as “walled gardens,” because of the amount of data they don’t share outside of their own ecosystem. Despite that limitation, the upside of the walled gardens for marketers is their unrivaled scale and their ability to target very specific audiences thanks to the deterministic data they can gather by linking user behavior to individual logins.

Waterfall: Like header bidding, the waterfall model is a yield optimization technique for publishers. The waterfall method gets its name because it is based on a set of cascading rules rather than real- time auctioning of inventory. (Sell to x at price y, if not sell to z at price a, etc.)

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 23 Acknowledgements Special thanks to the MMA Mobile Programmatic and Mobile Video Committees, particularly Michael Tuminello, formerly of Innovid, who was crucial to the production of this paper and the accompanying Mobile Programmatic Video Buyers Guide spreadsheet. Additionally, this guide was substantially benefited by input from these people at the following member companies: Maggie Mesa and Max Nelson of OpenX, Paul Fields and James Llvermore of AdColony, Vanja Brzin and Gefen Lamdan of Celtra, Erin Yasgar of Prohaska Consulting and David Sebag of Yieldmo. Cathy Taylor served as lead writer.

Thanks also to these MMA member companies who serve on either or both the Mobile Programmatic and Mobile Video Committees: 1-800-Flowers.com, AdTheorent, Inc., Akbank, Alphonso Inc., AppsFlyer, CNN, Cuebiq, ESPN, Facebook, Ford Motor Co., Google, IBM Watson Advertising, NBCUniversal, Neustar, Inc., Nutrisystem, Procter & Gamble, Salesforce, Samsung, Snap Inc., The Coca-Cola Company, The Washington Post, Turner Broadcasting System, Inc., , Ubimo, Vistar Media, and Walmart Inc. About the Mobile Marketing Association (MMA) The MMA is the world’s leading global non-profit trade association comprised of more than 800 member companies, from nearly fifty countries around the world. MMA Member companies hail from every faction of the mobile media ecosystem. Our consortium includes brand marketers, agencies, enabling technologies, media companies and others. The MMA’s mission is to accelerate the transformation and innovation of marketing through mobile, driving business growth with closer and stronger consumer engagement. Anchoring the MMA’s mission are three core pillars; to enlighten you with a path to the future; empower you with unique knowledge and connections; and enable you with science-backed truths and tools. Additionally, MMA industry-wide committees work collaboratively to develop and advocate global best practices and lead standards development. For more information about the MMA, please visit: www.mmaglobal.com.

©2019 Mobile Marketing Association, Inc. All Rights Reserved. 24