The Hotel School. Cornell. SC Johnson College of Business.

CENTER FOR HOSPITALITY RESEARCH

The Billboard Effect: Still Alive and Well

By Chris K. Anderson and Saram Han

EXECUTIVE SUMMARY

As a follow-ups aon follow-up two earlier on studies,two earlier this studies, report thisconfirms report the confirms so-called the so-calledbillboard billboar effect ond effect demand that occurson demand when that online occurs travel when agents online (OTAs) travel agents include (OT a As)particular include hotel a particular in their hotel listings. Evenin though their listings. many guests Even though book directly many guests with the book hotel directly brand, with this the study’s hotel brand, findings this are similar to study’sthose of findings earlier arstudiese similar which to those showed of earlier that beingstudies listed which on showed an OTA that site being listed increasedon an OTA reservations site increased through reservations the hotel through brand’s the hotel site. brand’sThe findings site. The in thefindings report in presented the report presentedA here underscored consumers’ reliance on websites when researching and booking their here underscored consumers’ reliance on websites when researching and booking their rooms,rooms, although non-dirnon-directect channels channels still still have have some some influence influence in lodging in lodging pur purchasechase decisions. In decisions.determining In whichdetermining web-based which marketing web-based efforts marketingpr oduce effortsthe best produce results, hotelthe best operators results, should hotel operatorsmake sure shouldtheir online make pr sureesence their is easy online to find,presence is attractive, is easy and to find,stands is upattractive, to the competition. and stands T upo tobetter the competition.understand changes To better in consumerunderstand online changes behavior in consumer this report online revisits behavior aspects ofthis the report billboar d effect through use of publicly available data sources. Contrary to research suggesting that the revisits aspects of the billboard effect through use of publicly available data sources. billboard effect is dead, this study’s results show that reports of its demise may have been Contrary to research suggesting that the billboard effect is dead, this study’s results show exaggerated. that reports of its demise may have been exaggerated.

Cornell Hospitality Report • April 2017 • www.chr.cornell.edu • Vol. 17, No. 11 1 ABOUT THE AUTHORS

Chris K. Anderson, Ph.D., is an associate professor at the Cornell School of Hotel Administration in the Cornell SC Johnson College of Business. Prior to his appointment in 2006, he was on faculty at the Ivey School of Business in London, Ontario Canada. His main research focus is on revenue management and service pricing. He actively works with industry, across numerous industry types, in the application and development of RM, having worked with a variety of hotels, airlines, rental car and tour companies as well as numerous consumer packaged goods and financial services firms. Anderson’s research has been funded by numerous governmental agencies and industrial partners and he serves on the editorial board of the Journal of Revenue and Pricing Management and is the regional editor for the International Journal of Revenue Management. At the School of Hotel Administration, he teaches courses in revenue management and service operations management. He earned his B.S. and Msc degrees from the University of Guelph, and his MBA and Ph.D. degrees from the University of Western Ontario, Ivey School of Business.

Saram Han is a Ph.D. student in Marketing at the Cornell School of Hotel Administration in the Cornell SC Johnson College of Business.His research Interests include data science, natural language processing for OTAs review, service operation, service marketing, and survey methodology. He earned a Bachelor of Business Administration, Tourism Management, from Kyung-Hee University, Seoul, Korea, and an MS degree from the Michigan Program in Survey Methodology, University of Michigan.

2 The Center for Hospitality Research • Cornell University CORNELL HOSPITALITY REPORT

The Billboard Effect: Still Alive and Well

By Chris K. Anderson and Saram Han

Changes in the online travel market are causing hotels to rethink their relationships with online travel agencies (OTAs) and to take a closer look at the impact on bookings from hanges in the online travel market are causing hotels to rethink their listing their properties with OTAs. One outcome of being listed on an OTA is additional relationships with online travel agencies (OTAs) and to take a closer look at the bookings on the brand’s own website, a phenomenon that co-author Chris Anderson impact on bookings from listing their properties with OTAs. One outcome of labeled the billboard effect. In a 2009 study, Anderson presented an experiment in which a being listed on an OTA is additional bookings on the brand’s own website, a group of hotels was listed and then removed from Expedia. com in alternate weeks. This phenomenon that co-author Chris Anderson labeled the billboard effect.In a 2009 study, Anderson testpresented found anthat, experiment compared in towhich being a grhidden,oup of beinghotels listedwas listed on the and site then increased removed reservationsfrom Expedia. 9 percentC to 26 percent (above transactions that occurred at Expedia).1 That was followed by com in alternate weeks. This test found that, compared to being hidden, being listed on the site aincr 2011eased study reservations examining 9 per consumers’cent to 26 per onlinecent pre-purchase(above transactions research that occurrthat founded at aboutExpedia). 75 1 That percentwas followed of consumers by a 201 1who study made examining reservations consumers’ with a majoronline hotel pre-pur brandchase had resear visitedch that an OTAfound inabout advance 75 per ofcent booking of consumers directly who with made the brand.2 reservations In this with report a major we show hotel brandthat the had ability visited of ana second-partyOTA in advance channel of booking to influence directly anwith eventual the brand. reservation2 In this r eportmay bewe lowershow now,that the but ability the of a billboardsecond-party effect channel still occurs, to influence sincean many eventual consumers reservation visit mayan OTA be lower prior nowto booking., but the billboard effect still occurs, since many consumers visit an OTA prior to booking.

1 Anderson, CK. “The Billboard Effect: Online ravelT Agent Impact on Non-OTA Reservation Volume,”Cornell Center for Hospitality Research Report, Vol. 9 No. 16. http://scholarship.sha.cornell.edu/chrpubs/2/ 2 Anderson, CK. “Search, OTAs, and Online Booking: An Expanded Analysis of the Billboard Effect,”Cornell Center for Hospitality Research Report, Vo. 11 No. 8. http://scholarship.sha.cornell.edu/chrpubs/4

Cornell Hospitality Report • April 2017 • www.chr.cornell.edu • Vol. 17, No. 11 3 Exhibit 1 Domain visitation (60 days prior to reservation)

Booking Reservations Site Visitation Prior to Reservation Channel OTAs Hotel Sites Web Search TripAdvisor Other Meta OTA 2,776 48% 68% 39% 33% Direct 2,317 65% 66% 34% 21%

Note: Sample OTAs include Expedia.com, Hotels.com, and Booking.com. Sample Hotel sites include Hilton.com, Marriott.com, and IHG.com. Searches include searches at Google, Yahoo, and Bing. Sample Meta sites include Kayak.com, Trivago. Com, and GoSeek.com.

Exhibit 2 Average number of visits per reservation (60 days prior to reservation)

Booking Reservations Site Visitation Prior to Reservation Channel OTAs Hotel Sites Web Searches TripAdvisor Other Meta OTA 2,776 8.4 3.4 4.6 2.9 2.4 Direct 2,317 7.2 6.5 5.1 4.1 2.3

A primaryA primary reason reason for this for change this change is consolidation is consolidation and reservations (including air, rental car, and hotel). A total innovationand innovation among among the online the online travel travelfirms. Expediafirms. Expedia has of 5,093 hotel reservations were made by the sample: acquiredhas acquir bothed bothTravelocity Travelocity and Orbitz, and Orbitz, while whilePriceline Priceline 54.5 percent (2,776) of these reservations were made at acquired Kayak and Expedia and also took a major equity OTAs and the remaining 2,317 (45.5 percent) were made acquired Kayak and Expedia and also took a major equity position in Trivago. Much of this merger activity has been directly at hotel websites.4 Using domain level informa- positionallowed in (fr Trivago.om a competition Much of this standpoint) merger activity by the has moves been tion for each website visited prior to the hotel reservation, allowedof Google (from and a T competitionripAdvisor standpoint)to become meta by the OT movesA sites, of as we focus on travel related behavior for 60 days prior to Googleare Kayak and and TripAdvisor Trivago, toand become their continued meta OTA evolution sites, as areto purchase. Because comScore only provides domain level Kayakbecoming and fullTrivago, fledged and theirOTAs continuedthat offer evolutionfacilitated to direct information (e.g., Hilton.com), we have information booking. There has also been an upsurge in hotel-OTA on which domains consumers visited (and how often), becoming full fledged OTAs that offer facilitated direct interactions, with several large hotel brands launching but we don’t necessarily know which pages or content booking. There has also been an upsurge in hotel-OTA direct booking campaigns. All of this activity has encour- consumers focused on.5 We do know whether they visited interactions,aged research with findings several that large imply hotel that brands the launchingbillboard effect web search related sites (Google, Yahoo, or Bing), but directis dead. booking This conclusion campaigns. stems All of fr thisom activitythe rapid has gr owth of we don’t know which keywords they searched. For web encouragedthe two major research OTAs (Expediafindings that and imply Priceline), that the which billboard now search related visits, we do know which site they went effecttake a is lar dead.ger shar Thise ofconclusion online transactions stems from (and the rapid transactions growth to after visiting the search engine. If this next site was a in general). To understand changes in consumer online travel related domain, we can infer that this was a travel of the two major OTAs (Expedia and Priceline), which now behavior we revisit aspects of the billboard effect through related search. Exhibit 1 summarizes the percentages of takeuse ofa largerpublicly share available of online data transactions sources. (and transactions hotel bookers who visited travel related sites. Exhibit in general). To understand changes in consumer online 1 is separated into two rows: the first owr represents behaviorPre-Purchase we revisit Web aspects Search, of the Social, billboard and effect OTA through consumers who book hotels at OTAs and the second row useVisitation of publicly available data sources. In this studyIn this we use study a randomly we selected use a samplerandomly of more thanselected 50,000 consumers sample from of a 4 We focus only on hotels that have at least 30 days pre-purchase panel of over two million online consumers maintained by comScore, which tracks all of more than 50,000 consumers from a panel of over two information (reservations made in February onwards) and those which the sample members’ 2015 online behavior.3 In our analysis we focus on some 13,000 there was a gap of at least 30 days followi=ng any prior travel related million online consumers maintained by comScore, which travel-related reservations (including air, rental car, and hotel). A total of 5,093 hotel reservations. 3 reservationstracks all were of made the bysample the sample: members’ 54.5 percent 2015(2,776) onlineof these reservations behavior were. 5 One methodology to address this issue is eye tracking. See: Bref- made at OTAs and the remaining 2,317 (45.5 percent) were made directly at hotel In our analysis we focus on some 13,000 travel-related fni Noone and Stephani K.A. Robson, “Using Eye Tracking to Obtain websites.4 Using domain level information for each website visited prior to the hotel a Deeper Understading of What Drives Online Hotel Choice,” Cornell reservation, we focus on travel related behavior for 60 days prior to purchase. Because 3 The comScore panel used only includes non-mobile, desktop Hospitality Report, Vol. 14, No. 8 (2014), Cornell Center for Hospitality comScorepanelists. only provides domain level information (e.g., Hilton.com), we have information Research. on which domains consumers visited (and how often), but we don’t necessarily know which pages or content consumers focused on.5 We do know whether they visited web search4 related sites (Google, Yahoo, or Bing), but we don’t know which keywords they The Center for Hospitality Research • Cornell University searched. For web search related visits, we do know which site they went to after visiting the search engine. If this next site was a travel related domain, we can infer that this was a travel related search. Exhibit 1 summarizes the percentages of hotel bookers who visited travel related sites. Exhibit 1 is separated into two rows: the first row represents consumers who book hotels at OTAs and the second row represents consumers booking directly at hotel websites. The exhibit summarizes the percentage of these consumers who visit OTAs, hotel websites and search engines, as well as sites such as TripAdvisor and meta sites (e.g., Kayak, Trivago, GoSeek) within 60 days prior to making a hotel reservation. Exhibit 3 OTA. (The six y-axes, showing relative frequency, are on the same scale, allowing a comparison of direct book- Distribution of visits to hotel websites prior to ers with those using OTAs.) The figures are noteworthy booking via a hotel site vs. an OTA as they indicate that web search activity (Exhibit 5) is happening fairly consistently during the entire 60-day research phase (although it gradually picks up just before the booking), whereas visits to TripAdvisor (Exhibit 5) and OTAs (Exhibit 6) tend to be intensive just prior to the booking. For OTA visitation prior to OTA booking (left panel of Exhibit 4), we exclude the OTA visit during which the transaction occurred. The intensity of TripAd- visor and OTA visitation prior to booking indicates that these travel sites may be greatly influencing the purchase decision. This observational data indicates that consumers remain actively engaged in researching their hotel stay. represents consumers booking directly at hotel websites. Review sites and OTAs are critical components of the The exhibit summarizes the percentage of these consum- purchase decision, although consumers rely less on search ers who visit OTAs, hotel websites and search engines, engines compared to our 2011 report, probably as a result as well as sites such as TripAdvisor and meta sites (e.g., of OTA consolidation and increased familiarly with the Kayak, Trivago, GoSeek) within 60 days prior to making a internet. hotel reservation. WhileWhile Exhibits Exhibits 3, 4, and 3, 4, 5 and illustrate 5 illustrate the role the travel role travelsites play The percentages in Exhibit 1 are reasonably consistent insites online play research, in online in r esearExhibitch, 7 in we Exhibit focus on7 we the focus start onof that 6 with those from our 2011 study. At that time, about 75 travelthe start research. of that Exhibit travel r7esear lists ch.the Exhibit travel related 7 lists thesites travel where percent of consumers who booked directly with a hotel related sites where the consumers’ research phase was the consumers’ research phase was initiated in advance of online visited an OTA prior to purchase (compared to initiated in advance of the hotel reservation. It summa- 65 percent in this study), while 83 percent of consumers therizes hotel the reservation. percentages It of summarizes first visits theoccurring percentages at meta of sites, first performed a web search in the earlier study (compared to visitsTrip Advisoroccurring, hotelat meta sites, sites, OT As,Trip and Advisor, web searhotelches sites, acr oss 66 percent in this study). OTAs,all consumers and web as searches well as separatedacross all consumersinto OTA versus as well dir asect As shown in Exhibit 2, the average number of visits separatedbooking channels. into OTA Itversus indicates direct that booking web sear channels.ch and ItT rip Advisor share similar percentages as the initial site for per reservation is not radically different for those who indicates that web search and Trip Advisor share similar booked on the OTA, compared to those who booked both OTA and direct bookers, while OTA bookers have percentages as the initial site for both OTA and direct with the hotel brand directly. In terms of web visits, the almost twice the frequency of meta and OTA visitation as online research behavior is consistent between OTA bookers,direct bookers. while OTA bookers have almost twice the bookers and hotel direct bookers, but hotel direct bookers frequency of meta and OTA visitation as direct bookers. visited TripAdvisor about 33 percent more often than Implications for the Billboard Effect OTA consumers. On average, hotel direct bookers make ResearchResear conductedch sinceconducted our 2009 report since reflects our new 2009 opinions report regarding reflects the billboard about twice as many visits to hotel websites (6.5) as OTA effect.new Estis opinions Green and r Lomannoegarding contend the that billboar the effectd is effect.considerably Estis less prevalentGreen than indicated earlier.7 They summarize work done by P.K. Kannan at the University of bookers (3.4). However, the distribution of these visits Marylandand Lomanno describing online contend consumer that behavior the using effect comScore is considerably data from 2012 and 2014 7 versus just the average (see Exhibit 3) shows that those (theless same pr evalentdata used here, than but indicatedfrom different years).earlier For. ease They of discussion summarize we show a two groups’ behavior is fairly consistent. That is, those reproductionwork done of results by Pfrom.K. this Kannan study in Exhibit at the 8. The University key insights from of thisMaryland exhibit are OTA bookers who visit hotel websites tend to visit about thedescribing low probabilities online of consumers consumer moving from behavior an OTA (labeled using as comScoran intermediary)e data to a the same number as those who book direct. The average hotelfrom website 2012 (9.3 and percent 2014 for 2012 (the and same 7.0 percent data for used2014) versus here, the but high frprobabilitiesom of consumers moving from OTA to OTA (90.7 percent in 2012 and 93.0 percent in 2014). shown in the exhibit is smaller because only about half of different years). For ease of discussion we show a repro- This indicates that it is unlikely that awareness is created at an OTA with consumers then the OTA bookers visit hotel websites prior to booking at switchingduction sites of and results booking withfrom hotels this directly study (as suggested in Exhibit by the 8.billboard The effect).key the OTA. Oneinsights detail that fr receivesom this less exhibitattention in ar thee Estis the Green low andpr obabilitiesLomanno report ofis that con - Exhibits 4, 5, and 6 show (on the x-axis) the distributions of the number of days before Exhibits 4, 5, and 6 show (on the x-axis) the distribu- thesesumers switching moving probabilities from are for an consecutive OTA (labeled website visits as (from an t-1intermediary) to t in Exhibit 8) booking that consumers perform web searches, visit TripAdvisor, or go to an OTA. (The and not for the consumer’s entire research process. As summarized in Exhibit 2, sixtions y-axes, of showing the number relative frequency, of days are on befor the samee booking scale, allowing that a comparison consum of- to a hotel website (9.3 percent for 2012 and 7.0 percent for consumers who visit OTAs, prior to booking direct with hotels do so 7.2 times on average, directers bookersperform with thoseweb using sear OTAs.)ches, The visit figures T areripAdvisor noteworthy as, theyor go indicate to anthat not once. We can create an approximation of the 9.3 and 7.0 figures from Exhibits 1 and web search activity (Exhibit 5) is happening fairly consistently during the entire 60-day 6. Exhibit7 Estis1 shows Gr thateen, 65 C percent and MV of consumers Lomanno. booking 2016. directly “Demystifying with the hotel thevisited an research phase (although it gradually picks up just before the booking), whereas visits to 6 OTADigital prior Marketplace:to booking direct, Spotlight and Exhibit on 6 (right-hand the Hospitality panel) shows Industry that about,” HSMAI 18 percent TripAdvisorThe (Exhibit sample 5) and in OTAs 2011 (Exhibit included 6) tend hotel to be dir intensiveect bookings just prior forto the July booking. and August of 2008, 2009, and 2010, with data provided by comScore. (ofFoundation. this 65 percent) visit an OTA on the day of the booking (day 0 on x-axis of Exhibit 6), For OTA visitation prior to OTA booking (left panel of Exhibit 4), we exclude the OTA visit the product of these two being 11.7 percent. That figure is higher than the 7.0 or 9.3 during which the transaction occurred. The intensity of TripAdvisor and OTA visitation percent as it ignores other (non-OTA) travel site visits on the same day of the booking prior to booking indicates that these travel sites may be greatly influencing the purchase that might have occurred between the OTA visit and the hotel direct booking. This decision.Cornell Hospitality Report • April 2017 • www.chr.cornell.edu • Vol. 17, No. 11 5 estimate of the OTA impact (like any click-to-click switching probability) ignores all the other OTA visits in Exhibit 6 (those not on the same day as the booking) and provides a conservative estimate of the effect. Exhibit 4 Time before booking of web searches

OTA Bookers Direct Bookers

2014) versus the high probabilities of consumers moving the transitions from hotel (H) website to hotel website from OTA to OTA (90.7 percent in 2012 and 93.0 percent (60 percent chance) as well as hotel to OTA transitions in 2014). This indicates that it is unlikely that awareness (40 percent). Because of these two transitions a consumer is created at an OTA with consumers then switching sites who started at a OTA has a 10.71 percent (0.042+0.0651) and booking with hotels directly (as suggested by the chance of ending up at the hotel website, up from the billboard effect). One detail that receives less attention in original 7 percent. So the probability that a consumer the Estis Green and Lomanno report is that these switching ends up booking directly at a hotel, given she was at an probabilities are for consecutive website visits (from t-1 to OTA earlier, depends upon how many of these website- t in Exhibit 8) and not for the consumer’s entire research to-website transitions are made. This probability con- process. As summarized in Exhibit 2, consumers who visit verges at about 15 percent after about four transitions. OTAs, prior to booking direct with hotels do so 7.2 times on The 15 percent (and the 7 percent) are path independent average, not once. We can create an approximation of the transition probabilities, which means the chance of a 9.3 and 7.0 figures from Exhibits 1 and 6. Exhibit 1 shows consumer moving from Expedia.com to Hilton.com is that 65 percent of consumers booking directly with the the same whether she is starting her travel research or is hotel visited an OTA prior to booking direct, and Exhibit almost finished and knows where she wants to stay. 6 (right-hand panel) shows that about 18 percent (of this We can’t read too much into these transition prob- 65 percent) visit an OTA on the day of the booking (day 0 abilities as they are simply click-to-click behavior and on x-axis of Exhibit 6), the product of these two being 11.7 don’t include the entire search process. In fact, as noted percent. That figure is higher than the 7.0 or 9.3 percent as by Estis Green and Lomanno there is a stronger effect it ignores other (non-OTA) travel site visits on the same of consumers moving to OTAs from hotel direct sites day of the booking that might have occurred between the versus the opposite, with a single click probability of 40 OTA visit and the hotel direct booking. This estimate of the percent of consumers clicking over to OTAs from hotel OTA impact (like any click-to-click switching probability) direct sites. ignores all the other OTA visits in Exhibit 6 (those not on AnotherAnother way to examine way these to switching examine probabilities these is switching to consider them pr inobabili aggregate- the same day as the booking) and provides a conservative acrossties isthe to entire consider research process them versus in aggr from egateclick-to-click acr actions.oss the In our entir samplee of 5,093 hotel reservations, 4,273 of these consumers visited OTAs, with 2,776 booking at estimate of the effect. research process versus from click-to-click actions. In OTAs and the remaining 1,497 booking direct with hotels (see Exhibit 10). This indicates To illustrate the potential impact of the 7.2 (on average) OTA visits consider a consumer at To illustrate the potential impact of the 7.2 (on average) 35our percent sample of hotel ofroom 5,093 purchasers hotel who r eservations,visited OTAs eventually 4,273 booked of these direct. Our an OTA who only makes two transitions (i.e., moves to two websites). Using the 2014 OTA visits consider a consumer at an OTA who only makes sampleconsumers also shows visited 232 customers OTAs, who visitedwith OTAs 2,776 but bookingbooked direct at with OT hotelsAs results from Exhibit 8 there is a 7 percent chance that she moves to a hotel website and 93 two transitions (i.e., moves to two websites). Using the 2014 withoutand thevisiting remaining hotel websites 1,497prior to thebooking purchase. dirThisect 5.5 percentwith ofhotels the sample percent chance she moves to or remains at an OTA. Following that click she could also represents unique shoppers, as they never visited hotel websites until the purchase moveresults to a hotel from or an Exhibit OTA. Exhibit 8 ther9 showse isall athe 7 possible percent outcomes chance of a consumer that she making (see Exhibit 10). This indicates 35 percent of hotel room moment. They are also active travel researchers, making an average of 14.1 visits to twomoves transitions, to addinga hotel the transitionswebsite from and hotel 93 (H) per websitecent to chancehotel website she (60 movespercent purchasers who visited OTAs eventually booked direct. travel related sites in the 60 days prior to booking. The 5.5 percent figure serves as the chance) as well as hotel to OTA transitions (40 percent). Because of these two transitions a to or remains at an OTA. Following that click she could also lowOur end sampleof this switching also behavior shows and 232 the 35customers percent figure whoserves asvisited a high end OT As consumer who started at a OTA has a 10.71 percent (0.042+0.0651) chance of ending up at move to a hotel or an OTA. Exhibit 9 shows all the possible estimate,but booked with the billboard direct effect with falling hotels somewhere without in between. visiting Exhibit hotel11 provides a the hotel website, up from the original 7 percent. So the probability that a consumer ends up outcomes of a consumer making two transitions, adding summarywebsites of billboard prior effect to theestimates, pur chase.comparing This the original 5.5 perestimatecent from of our the 2009 booking directly at a hotel, given she was at an OTA earlier, depends upon how many of report with the current estimate, as well as an estimate based on step-to-step transition these website-to-website transitions are made. This probability converges at about 15 probabilities and steady state transition probabilities from Estis Green and Lomanno. percent after about four transitions. The 15 percent (and the 7 percent) are path independent6 transition probabilities, which means the chance of a consumer moving from The Center for Hospitality Research • Cornell University Expedia.com to Hilton.com is the same whether she is starting her travel research or is almost finished and knows where she wants to stay. Exhibit 5 Time before booking of TripAdvisor visitation OTA Bookers Direct Bookers

Exhibit 6 Time before booking of OTA visitation

OTA Bookers Direct Bookers

sample represents unique shoppers, as they never visited (namely, OTA, meta, TripAdvisor, hotel direct, and web hotel websites until the purchase moment. They are also search), we coded major brand sites (e.g., marriott.com, active travel researchers, making an average of 14.1 visits hilton.com) as hotel direct and also coded independent to travel related sites in the 60 days prior to booking. The hotel websites and hotel specific sites as hotel direct. Dis- 5.5 percent figure serves as the low end of this switching tinct from the earlier study we also subdivided interme- behavior and the 35 percent figure serves as a high end diaries into a series of categories (i.e., OTAs, web search, estimate, with the billboard effect falling somewhere in meta, TripAdvisor, airline direct). The result of this coding between. Exhibit 11 provides a summary of billboard shows considerably different site share than that reported effect estimates, comparing the original estimate from in the study summarized by Estis Green and Lomanno. If our 2009 report with the current estimate, as well as an we focus on just hotel direct and OTAs we find 34.5 per- estimate based on step-to-step transition probabilities and cent of these visits are to hotel direct and 65.5 percent to steady state transition probabilities from Estis Green and OTAs, compared to their 2014 numbers of 15.2 percent to Lomanno. hotel direct and 84.8 percent to intermediaries. Similarly, One aspectOne of theaspect comScore of thedata iscomScor the need to ecode data URLs is into the the need appropriate to travel this coding shows 16 percent of consumers visited hotel categories,code URLs which requiresinto the understanding appropriate of the travel travel industry. categories, During the coding which of direct only and did not visit OTAs, versus the 7 percent URLs into our specific categories of interest (namely, OTA, meta, TripAdvisor, hotel requires understanding of the travel industry. During the reported in Estis Green and Lomanno, and 28.5 percent direct, and web search), we coded major brand sites (e.g., marriott.com, hilton.com) as hotelcoding direct andof URLsalso coded into independent our specific hotel websites categories and hotel specific of inter sitesest as hotel visiting OTAs only (versus 64 percent) and 55.5 percent direct. Distinct from the earlier study we also subdivided intermediaries into a series of categories (i.e., OTAs, web search, meta, TripAdvisor, airline direct). The result of this coding shows considerably different site share than that reported in the study summarizedCornell Hospit by Estisalit Greeny R eandport Lomanno. • April If 2017we focus • wwwon just.c hotelhr.cor directne ll.eanddu OTAs • V weol. 17, No. 11 7 find 34.5 per-cent of these visits are to hotel direct and 65.5 percent to OTAs, compared to their 2014 numbers of 15.2 percent to hotel direct and 84.8 percent to intermediaries. Similarly, this coding shows 16 percent of consumers visited hotel direct only and did not visit OTAs, versus the 7 percent reported in Estis Green and Lomanno, and 28.5 percent visiting OTAs only (versus 64 percent) and 55.5 percent visiting both (versus 29 percent). Our 65.5 percent (for OTAs) and 34.5 percent (for hotel direct sites) share visitation figures are not far off from numbers recently reported by Phocuswright,8 which indicated that 72 percent of U.S. consumers use OTAs for hotel shopping, with 44 percent using hotel websites. Exhibit 7 Exhibit 9 Domain visitation—first visit distribution Two-click transitions

Visitation Site Booking Channel All OTA Direct Meta 6.2% 8.2% 3.8% Trip Advisor 8.3% 8.6% 8.0% Hotel Sites 25.5% 14.6% 38.7% OTA 42.3% 51.9% 30.7% Web Search 17.7% 16.8% 18.7%

Exhibit 8 Switching probability between types of sites

Exhibit 10 Channel switching

Booking Channel Visited OTA OTA 2,776

Notes: The probability that a user will visit an intermediary and then go to a hotel Hotel Direct 1,497 site is very small—7% in 2014. Most likely, if a user visits an intermediary site he Total 4,273 will stay there or continue going there. See: Estis Green, C and MV Lomanno. 2016. “Demystifying the Digital Marketplace: Spotlight on the Hospitality Industry,” OTA 65 % p. 45, HSMAI Foundation. Hotel Direct 35%

Note: N = 4,273 visiting both (versus 29 percent). Our 65.5 percent (for OTAs) and 34.5 percent (for hotel direct sites) share visita- tion figures are not far off from numbers recently reported a hotel room (stayed with family and friends) or they by Phocuswright,8 which indicated that 72 percent of U.S. booked hotel rooms in a manner not tracked by comScore consumers use OTAs for hotel shopping, with 44 percent (e.g., phone, travel agent, or mobile). If we assume some fraction of these airline bookers and hotel specific using hotel websites. If we assume some fraction of these airline bookers OTAand visitors hotel required specific a hotel OTA room visitors but made requir thated reservation a hotel r offlineoom butor Estimating Impacts for Other Hotel throughmade a that traditional reservation travel agent, offline we can or derive through an estimatea traditional for the offline component of the billboard effect. For example, if you needed a Transactions travel agent, we can derive an estimate for the offline hotelcomponent room only 10of percentthe billboar of the dtime effect. you took For a flight,example, then 10if percentyou Our sampleOur samplecontains 13,867contains reservations 13,867 r eservationsby 5,970 consumers, by 5,970 ofneeded the 1,598, a orhotel 159.8, room represents only 10a billboard percent effect of the of 5.8 time percent you (159.8took includingconsumers, airlines including (6,364), rental airlines cars (2,410),(6,364), and rental hotels cars (5,093). (2,410), Of the / 2,776a flight, OTA thenhotel reservations).10 percent of If theyou 1,598,needed orhotel 159.8, accommodations represents 50 totaland number hotels of (5,093). consumers, Of the2,948 total visited number an OTA of but consumers, did not make a percenta billboar of thed time, effect the effectof 5.8 is peralmostcent 29(159.8 percent. / 2,776 We assume OTA hotelthat hotel2,948 reservation, visited an with OT 2,414A but of thedid 2,948 not makingmake aan hotel airline r eservareservation.- thesereservations). 160 consumers If you (rounded) needed who hotel visited accommodations a hotel specific OTA 50and Iftion, we focus with only 2,414 on OTAs of the that 2,948 don’t making sell airline an products airline (e.g., reservation. madeper centan airline of the reservation time, the made effect an offline is almost hotel reservation29 percent. that W wase Hotels.com,If we focus Booking.com, only on OT Hotwire.com)As that don’t we have sell aairline sample pr ofoducts 1,598 influencedassume by that their these visit to160 the consumers hotel specific (rOTA.ounded) We assert who that visited the visit a consumers(e.g., Hotels.com, who visited Booking.com, a hotel specific OTA Hotwir and madee.com) an weairline have tohotel an OTA specific influenced OT theA andhotel made purchase an asairline a result reservation of our earlier made reservationa sample butof 1,598did not consumersbook a hotel roomwho online. visited These a hotel consumers specific analysisan offline (65 percent hotel of reservation hotel direct reservationsthat was influenced were preceded by by their an eitherOTA did and not made need a an hotel airline room (stayedreservation with family but anddid friends)not book or they a OTAvisit visit, to andthe thesehotel visits specific are concentrated OTA. We assertor close that to the the time visit of to bookedhotel rhoteloom rooms online. in a These manner consumers not tracked by either comScore did not(e.g., need phone, purchase).an OTA influenced the hotel purchase as a result of our travel agent, or mobile). earlier analysis (65 percent of hotel direct reservations 8 Phocuswright’s Search, Shop, Buy: The New Digital Funnel.

8 The Center for Hospitality Research • Cornell University Exhibit 11 Comparison of billboard estimates

Study Data/Approach Estimate of Billboard Effect Anderson (2009) Experiment 7.5% to 26% Estis Green and Lomanno (2016) Archival comScore panel 2012 and 2014 7% (single move) to 15%* (multiple moves) Anderson and Han (2017) Archival comScore panel 2015 5% to 35%

Note: *15 percent is steady state approximation derived in Anderson and Han (2017) using Estis Green and Lomanno (2016) single step transitions.

Exhibit 12 Domain visitation 30 days after reservation

Booking Channel Reservations Visiting OTAs (%) Visiting Hotels (%) OTA Visits Hotel Website Visits OTA 1,016 81 54 9.2 2.2 Direct 620 74 73 8.5 6.4 were preceded by an OTA visit, and these visits are con- Exhibit 13 centrated or close to the time of purchase). Number of unique hotel sites visited post purchase Post-Purchase Behavior Many hotels engage in some form of revenue management using price to manage supply and demand imbalances. As a result, hotel prices may fluctuate, caus- ing consumers to check prices and time their purchases to get the best price. Also, most hotel reservations have flexible cancellation policies allowing consumers to cancel without penalty if the cancellation is made at least 24 hours prior to check-in. This combination of flexible cancellation policies and fluctuating prices may result in consumers second-guessing their purchase decisions. A potential outcome of this buyers’ remorse is that consum- ers continue to check prices or compare hotels post-pur- 9 Exhibit 12 shows that, in fact, consumers are more chase. We investigate this aspect of consumer behavior active post-purchase than they are pre-purchase, with by looking at travel site visitation after the hotel reserva- the percentage of hotel direct consumers visiting OTAs tion. To isolate pre-purchase research from post-purchase rising to 74 percent from 65 percent, along with hotel site activity we look only at transactions where there is a time visitation increasing to 54 percent from 48 percent for OTA gap of at least three months between purchases. A subset bookers. The increased level of OTA visitation by hotel of the 5,093 hotel transactions where there was at least direct bookers is consistent with consumers checking 90 days between the hotel transaction and the next travel prices to determine whether they paid too much. Exhibit purchase reduces our sample to 1,636 consumers (1,016 13 shows the number of different hotel sites that consum- OTA consumers and 620 hotel direct reservations). For ers visit post-purchase, indicating that while the average these 1,636 consumers we track OTA and hotel website number of times consumers visit hotel sites is over six, visitation for 30 days after the reservations. they are only visiting one or two sites as they seek addi- tional details about their hotel, or to check prices. 9 This is a long standing practice. See, for example: Gary M. Thompson and Alexandra Failmezger, “Why Customers Shop Around: A Comparison of Hotel Room Rates and Availability across Booking Channels,” Cornell Hospitality Report, Vol. 5, No. 2 (2005); Cornell Cen- ter for Hospitality Research.

Cornell Hospitality Report • April 2017 • www.chr.cornell.edu • Vol. 17, No. 11 9 Summary average, shoppers conduct a lot of research online in the It has become increasingly difficult for hoteliers to -de 60 days before purchasing a hotel reservation, making 25 termine which sales and marketing efforts lead to demand, visits to travel related sites. While research suggests that and how these efforts (e.g., sponsored search, banner the ability of a non-direct channel to influence an eventual ads, OTAs, and offers) interact. Without this attribution reservation at a hotel may be low (between 5.5 and 35 it is impossible to determine ROI of marketing efforts or percent), there is still a billboard effect on customers as true channel specific acquisition costs. So, while it’s true they visit one of these non-direct sites prior to booking. that the demand funnel is more complex, to state that Hoteliers who ensure that their online presence is easy the billboard effect is dead as a function of this complex- to find, attractive, and competitive will capture more of ity assumes that a hotel’s listing at OTAs only influences these customers. those consumers booking at the OTA, and that consumers It is importantIt is important to note to that note our that data our sample data sample is observational, is booking direct with hotels are not influenced by listings andobservational, any inferences and weany draw infer doences not westate draw causation. do not stateThe only at OTAs, even though as indicated in Exhibit 10 over 30 waycausation. to truly Theknow only the way impacts to tr ofuly marketing know the actions impacts upon of hotel marketing actions upon hotel transactions is to perform percent of these direct bookers started their research pro- transactions is to perform experiments as was done in our experiments as was done in our 2009 study. We do not cess at an OTA. This study indicates that OTAs now get an 2009 study. We do not indicate that the 35 percent of increasingly larger share of the transaction landscape, but indicate that the 35 percent of consumers visiting an OTA consumers visiting an OTA who book direct would not have OTAs are visited by almost two-thirds of all online hotel who book direct would not have booked at the specified direct consumers, down about 10 percent from our 2011 bookedhotel if atthat the hotel specified had not hotel been if that listed hotel at thehad OT notA. been After listed all, study results, showing that the magnitude of the billboard atther thee OTA.are many After other all, there methods are many for cr othereating methods product for awar e- effect is decreasing, even though it has not disappeared creatingness. Individual product awareness. hotels that wantIndividual to solve hotels the that attribution want to puzzle must conduct a series of these pseudo-experiments entirely. solve the attribution puzzle must conduct a series of these Booking aBooking hotel online remainsa hotel a complex online activity remains for all but a the complex most loyal ofactivity hotel for in which they stop certain actions for short periods of pseudo-experiments in which they stop certain actions for shoppers,all but and the while most almost loyal39 percent of ofhotel direct bookersshoppers, start their and travel while research almost at a time (e.g., preferred placement at OTAs, sponsored search hotel site (Exhibit 7), 31 percent of consumers who start their search at a hotel site end up shortat Google, periods ads of withintime (e.g., hotel preferred finder atplacement Google) andat OTAs, compar e booking39 per at ancent OTA. ofOn diraverage,ect shoppersbookers conduct start a lottheir of research travel online resear in the ch60 daysat sponsoredtransaction search volumes at Google, across the ads tr withineatment hotel periods. finder nat beforea hotel purchasing site a (Exhibithotel reservation, 7), 31 making per 25cent visits of to travelconsumers related sites. who While start researchtheir suggests search that at the a abilityhotel of asite non-direct end channelup booking to influence at an an eventual OTA. On Google) and compare transaction volumes across the reservation at a hotel may be low (between 5.5 and 35 percent), there is still a billboard treatment periods. effect on customers as they visit one of these non-direct sites prior to booking. Hoteliers who ensure that their online presence is easy to find, attractive, and competitive will capture more of these customers.

10 The Center for Hospitality Research • Cornell University Center for Hospitality Research Publication Index chr.cornell.edu

2017 Reports CREF Cornell Hotel Indices Vol. 16 No. 21 FRESH: A Food-service Sustainability Rating for Hospitality Vol 17. No. 10 Ethics from the Bottom Vol. 6 No. 1 Cornell Hotel Indices: Sector Events, by Sanaa I. Pirani, Ph.D., Up, by Judi Brownell, Ph.D. Fourth Quarter 2016: Hotels Are Getting Hassan A. Arafat, Ph.D., and Gary M. Costlier to Finance, by Crocker H. Liu, Thompson, Ph.D. Vol. 17. No. 9 Entrepreneurship Is Ph.D., Adam D. Nowak, Ph.D., and Global: Highlights from the 2016 Global Robert M. White, Jr. Vol. 16 No. 20 Instructions for the Entrepreneurship Roundtable, by Mona Early Bird & Night Owl Evaluation Anita K. Olsen, Ph.D. 2016 Reports Tool (EBNOET) v2015, by Gary M. Vol. 16 No. 28 The Role of REIT Thompson, Ph.D. Vol. 17. No. 8 Total Hotel Revenue Preferred and Common Stock in Management: A Strategic Profit Diversified Portfolios, by Walter I. Vol. 16 No. 19 Experimental Evidence Perspective, by Breffni M. Noone, Boudry, Ph.D., Jan A. deRoos, Ph.D., and that Retaliation Claims Are Unlike Other Ph.D., Cathy A. Enz, Ph.D., and Jessie Andrey D. Ukhov, Ph.D. Employment Discrimination Claims, by Glassmire David Sherwyn, J.D., and Zev J. Eigen, Vol. 16 No. 27 Do You Look Like Me? J.D. Vol. 17 No. 7 2017 CHR Compendium How Bias Affects Affirmative Action in Hiring, Ozias Moore, Ph.D., Alex M. Vol. 16 No. 18 CIHLER Roundtable: Vol. 17 No. 6 Do Property Susskind, Ph.D., and Beth Livingston, Dealing with Shifting Labor Characteristics or Cash Flow Drive Hotel Ph.D. Employment Sands, by David Sherwyn, Real Estate Value? The Answer Is Yes, J.D. by Crocker Liu, Ph.D., and Jack Corgel, Vol. 16 No. 26 The Effect of Rise in Ph.D. Interest Rates on Hotel Capitalization Vol. 16 No. 17 Highlights from the 2016 Rates, by John B. Corgel, Ph.D. Sustainable and Social Entrepreneurship Vol. 17 No. 5 Strategic Management Enterprises Roundtable, by Jeanne Practices Help Hospitals Get the Most Vol. 16 No. 25 High-Tech, High Varney from Volunteers, by Sean Rogers, Ph.D. Touch: Highlights from the 2016 Entrepreneurship Roundtable, by Mona Vol. 16 No. 16 Hotel Sustainability Vol. 17 No. 4 What Matters Most to Anita K. Olsen, Ph.D. Benchmarking Index 2016: Energy, Your Guests: An Exploratory Study of Water, and Carbon, by Eric Ricaurte Online Reviews, by Jie J. Zhang, Ph.D., Vol. 16 No. 24 Differential Evolution: A and Rohit Verma, Ph.D. Tool for Global Optimization, by Andrey Vol. 16 No. 15 Hotel Profit Implications D. Ukhov, Ph.D. from Rising Wages and Inflation in the Vol. 17 No. 3 Hotel Brand Standards: U.S., by Jack Corgel, Ph.D. How to Pick the Right Amenities for Vol. 16 No. 23 Short-term Trading Your Property, by Chekitan S. Dev, in Long-term Funds: Implications for Vol. 16 No. 14 The Business Case for Rebecca Hamilton, and Roland Rust Financial Managers, by Pamela Moulton, (and Against) Restaurant Tipping, by Ph.D. Michael Lynn, Ph.D. Vol. 17 No. 2 When Rules Are Made to Be Broken: The Case of Sexual Vol. 16 No. 22 The Influence of Vol. 16 No. 13 The Changing Harassment Law, by David Sherwyn, Table Top Technology in Full-service Relationship between Supervisors and J.D., Nicholas F. Menillo, J.D., and Zev J. Restaurants, by Alex M. Susskind, Ph.D., Subordinates: How Managing This Eigen, J.D. and Benjamin Curry, ,Ph.D. Relationship Evolves over Time, by Michael Sturman, Ph.D. and Sanghee Vol. 17 No. 1 The Future of Hotel Park, Ph.D. Revenue Management, by Sheryl E. Kimes, Ph.D.

Cornell Hospitality Report • April 2017 • www.chr.cornell.edu • Vol. 17, No. 11 11 Advisory Board Cornell Hospitality Research Note Vol. 17, No. 11 (April 2017) © 2017 Cornell University. This report may not be Syed Mansoor Ahmad, Vice President, Global Business reproduced or distributed without the express permission Head for Energy Management Services, Wipro EcoEnergy of the publisher. Marco Benvenuti MMH ’05, Cofounder, Chief Analytics and Cornell Hospitality Report is produced for the benefit Product Officer, Duetto of the hospitality industry by Scott Berman ’84, Principal, Real Estate Business Advisory The Center for Hospitality Research Services, Industry Leader, Hospitality & Leisure, PwC at Cornell University. Erik Browning ’96, Vice President of Business Consulting, The Rainmaker Group Christopher K. Anderson, Director Program Manager Bhanu Chopra, Founder and Chief Executive Officer, Carol Zhe, RateGain Jay Wrolstad, Editor Glenn Withiam, Executive Editor Susan Devine ’85, Senior Vice President–Strategic Kate Walsh, Acting Dean, School of Hotel Development, Preferred Hotels & Resorts Administration Ed Evans ’74, MBA ’75, Executive Vice President & Chief Human Resources Officer, Four Seasons Hotels and Center for Hospitality Research Resorts Cornell University School of Hotel Administration Kevin Fliess, Vice President of Product Marketing, CVENT, SC Johnson College of Business Inc. 389 Statler Hall Chuck Floyd, P ’15, P ’18 Global President of Operations, Ithaca, NY 14853 Hyatt R.J. Friedlander, Founder and CEO, ReviewPro 607-254-4504 www. chr.cornell.edu Gregg Gilman ILR ’85, Partner, Co-Chair, Labor & Employment Practices, Davis & Gilbert LLP Dario Gonzalez, Vice President—Enterprise Architecture, Lead, Accenture DerbySoft Carolyn D. Richmond ILR ’91, Partner, Hospitality Practice, Linda Hatfield, Vice President, Knowledge Management, Fox Rothschild LLP IDeaS—SAS David Roberts ENG ’87, MS ENG ’88, Senior Vice President, Bob Highland, Head of Partnership Development, Consumer Insight and Revenue Strategy, Marriott Barclaycard US International, Inc. Steve Hood, Senior Vice President of Research, STR Rakesh Sarna, Managing Director and CEO, Indian Hotels Sanjeev Khanna, Vice President and Head of Business Unit, Company Ltd. Tata Consultancy Services Berry van Weelden, MMH ’08, Director, Reporting and Kenny Lee, Vice President of Marketing, Revinate Analysis, priceline.com’s hotel group Josh Lesnick ’87, Executive Vice President and Chief Adam Weissenberg ’85, Global Sector Leader Travel, Marketing Officer, Wyndham Hotel Group Hospitality, and Leisure, Deloitte Faith Marshall, Director, Business Development, NTT DATA Rick Werber ’83, Senior Vice President, Engineering and Sustainability, Development, Design, and Construction, Host David Mei ’94, Vice President, Owner and Franchise Hotels & Resorts, Inc. Services, InterContinental Hotels Group Dexter Wood, Jr. ’87, Senior Vice President, Global Head— David Meltzer MMH ’96, Chief Commercial Officer, Sabre Business and Investment Analysis, Hilton Worldwide Hospitality Solutions Jon S. Wright, President and Chief Executive Officer, Access Nabil Ramadhan, Group Chief Real Estate & Asset Point Financial Management Officer, Jumeirah Group Umar Riaz, Managing Director—Hospitality, North American

12 The Center for Hospitality Research • Cornell University