63. Jason Fairchild – developing GoTo, OpenX and tvScientific

67. Ramsey McGrory – DoubleClick to Mediaocean via Right Media

Ramsey McGrory started his career in the U.S. Army and later found himself at the mighty DoubleClick [https://paleoadtech.com/2023/12/10/11-kevin-oconnor-co-founding-doubleclick-and-more/] in 1999 at the peak of the dot-com boom as an ad sales exec “banging the phones” selling direct-response ads to finance and insurance companies. He joined the seminal Right Media [https://paleoadtech.com/2023/12/10/40-brian-okelley-part-2-the-right-media-experience/] in 2004 as VP of platform and ad sales, staying on after Yahoo acquired the company [https://www.reuters.com/article/technology/yahoo-to-buy-rest-of-right-media-for-680-mln-idUSN30370375/] in 2007. Later, Ramsey was CEO at AddThis [https://techcrunch.com/2016/01/05/oracle-addthis/?guccounter=1&guce_referrer=aHR0cHM6Ly93d3cuZ29vZ2xlLmNvbS8&guce_referrer_sig=AQAAAIwtl9NQYFNxXJiUMaYobpIdoMvT_iWkBu8Wy-Pz2gU3M4T_GdUEJ5kkMVZ9omkcorUj4g6PyIrjxMz5QjsZ1h9Y1KCkQCtaeQAC1TyoiW4NX0OhjkLAhZfsuf7u8HhcvenzKY9scekCBEN7pBUJLWDAtdgGF9wbYl0GmShuxEkW], acquired by Oracle, and president at Scout Media before joining his old friend and former PaleoAdTech guest Bill Wise [https://paleoadtech.com/2023/12/10/28-bill-wise-accounting-for-the-rise-of-right-media-and-more/] at Mediaocean, where he is CDO. As Ramsey tells Marty in this fascinating episode, he fell into advertising by accident in the 1990s after starting his career in New York at Citibank, working on then-trendy derivatives; a hiring manager moved over to a little-known startup called DoubleClick and suggested he join up. Ten interviews later (none with founder Kevin O’Connor), he was hired as a display ads sales rep, a career change for this former Army logistics specialist. He thrived on the chaos and excitement of the boom. “It was a very diverse community,” he recalls. “You’d get the super-engineering quant types, you’d get the creative types and everything in between.” In the process, Ramsey learned a lot about how to make advertising work on the internet as part of a direct-response group (distinct from the more premium DoubleClick brand-sales group) led by Bill Wise and Mike Walrath. The process was manual: he got a daily spreadsheet describing publishers, impressions, cost, conversions; and he would make tweaks to targeting and frequency capping. Daily frustration led to innovation, years later: “The concept of Right Media came out of DoubleClick’s unwillingness to invest in an ad server and a media business that was focused on meeting direct-response goals,” Ramsey says. DoubleClick went through 7-8 rounds of layoffs after the dot-com crash as the brand-ad market cratered and colleague after colleague was called into the HR office on the 13th floor of the luxe “Click City” on 33rd and 10th. Ramsey’s niche in direct-response stayed relatively steady: “How many times have these direct response advertisers said, if you’re meeting my goals, I’ve got unlimited budget?” Ramsey followed ex-DoubleClicker Mike Walrath to Right Media, which Walrath founded in 2003 in the offices of Joe Zawadzki’ [https://paleoadtech.com/2023/12/10/3-joe-zawadzki-the-original-mediamath-man/]s Poindexter (later called [x+1]). Of course, Right Media gained an inestimable asset when the technically-brilliant Brian O’Kelley [https://paleoadtech.com/2023/12/10/39-brian-okelley-part-1-from-la2nite-to-right-media/] joined and a series of fundamental ad-serving and network innovations unrolled, setting up the era of RTB. Ramsey describes a market with rampant impressions on sites like MySpace, requiring scale (O’Kelley’s specialty as an engineer and the topic of his Princeton thesis paper) but also inspiring new ideas. Optimization was primarily by “suppression.” Meaning: the advertiser picked a price point (e.g., $2.50 CPM) and inventory would be optimized by changing geos, websites, frequency — optimization happening by removing placements, which lowered scale and decreased spend. The team experimented, creating campaign segments at different price points in the ad server. This tactic raised the frequency cap for ads: “You’re bombing the same ad unit and the ad unit is effectively sort of competing with itself,” Ramsey explains. At this point, O’Kelley had one of his insights: “Brian looked at it and said, ‘We just built this wrong.'” Meaning: the ad server (called Manage) optimized inventory against a set price. But what if that model could be changed — what if campaigns could flip the model from optimizing inventory against price to optimizing price against inventory? Thus: dynamic pricing. It was a fundamental conceptual and technical change. Now theoretically every advertiser could look at every impression, which improves yields for publishers. Advertisers were told that the eCPM (average CPM over the campaign) could be lower overall than the previously-set price but the scale would go up 5X (say) because more inventory is available. This technology, rebuilt by O’Kelley, became Yield Manager, and it was licensed to a number of ad networks including Right Media’s own. At this point the 3-4 networks in the group noticed they had all trafficked demand and supply tags into one another. This observation in turn led the ad server to morph into the first “exchange,” where users (at first, ad networks) would have a “seat” and connect to other seat-holders without having to use a lot of cross-tagging. This was the pioneering Right Media Exchange (RMX) of song and legend. The only tags used were the creative tags on the page. There was an auction for each impression. However, a limitation was that all the intelligence required had to be synched into the exchange. (This data-sharing requirement made some large buyers like P&G uncomfortable.) The next evolution was the emergence of DSPs and SSPs which pulled decision intelligence from the exchange itself into buy- and sell-side platforms. How did buyers and sellers react to the idea of an ad exchange? At first, Ramsey says, it was not difficult to describe an auction — after all, everyone knew how the stock market and paid search ads worked, — but “dynamic pricing” made buyers uneasy. The Right Media sales team came up with the idea of “PANT People,” styrofoam figures brought into meetings to demonstrate for visual-tactile learners how the platform worked. PANT stood for the parties in the exchange: Publisher-Advertiser-Network-Technology. [https://i0.wp.com/paleoadtech.com/wp-content/uploads/2024/09/image-5.png?resize=782%2C401&ssl=1] [https://i0.wp.com/paleoadtech.com/wp-content/uploads/2024/09/image001.png?resize=782%2C695&ssl=1] The challenge with dynamic pricing was that it made ad serving fees (then materially higher than today, say, $0.05-0.07) difficult to forecast. Objections were met by focusing the market on eCPM and setting a maximum bid that was higher the eCPM so more inventory could be accessed. After Yahoo’s acquisition of Right Media, Ramsey stayed at Yahoo for four years, which he says he enjoyed. He started at Mediaocean eight years ago as CRO, building a sales function beyond the original holding-company core and now shepherds mergers and acquisitions, of which Mediaocean made 10-12 in recent years. The largest of these was Flashtalking, run by PaleoAdTech guest John Nardone [https://paleoadtech.com/2023/12/10/62-john-nardone-flashtalking-about-modem-media-and-x1/], which powers the fast-growing digital component of Mediaocean’s business.

66. Daniel Jaye – starting it all with Engage

Daniel Jaye was co-founder in 1995 of Engage, a pioneer in bringing database marketing to the internet. A competitor of DoubleClick, Engage built arguably the largest database [https://web.archive.org/web/20010331053635/http:/engage.com/promotions/admanager_promo.cfm] of pseudonymous profiles at the time, and Daniel and his team created innovative technologies for ETL, large-scale analytics and behavioral targeting. Daniel was also the man behind much of the technical resourcefulness of the ad network Tacoda. In 2010, he co-founded another technically ingenious startup called Korrelate, an ad attribution solution. Today he is CEO Aqfer [https://aqfer.com/], a marketing data plataform-as-a-service company based in Florida. An astrophysicist inspired by Tom Swift [https://en.wikipedia.org/wiki/Tom_Swift] stories in his youth, Daniel began his adult journey in management consulting and then joined Epsilon, where he worked on an early pen-based PDA before enlisting at Fidelity. At the time — as Daniel tells Marty in this erudite episode — there were two companies pushing the boundaries of data management: Walmart (supporting scale) and Fidelity (supporting complexity). At Fidelity, Daniel’s job was to run its Teradata-based massively-parallel processing (MPP) environment. In this era before Hadoop, Daniel developed a global reputation as a cross-platform high-scale data management expert. This reputation reached the ear of David Wetherell [https://www.crunchbase.com/person/david-wetherell], the energetic CEO of CMGi, and Wetherell approached Daniel with a compelling (if rather vague) idea: “Bring database marketing to the internet.” It was 1995. Netscape was one year old. At the time, CMGi — the name stood for College Marketing Group Inc. — had a database of consumers linked to their book- and magazine-buying behaviors. So it could be used to tie people to their areas of interest and serve marketers. Seizing the opportunity of the nascent internet, Wetherell used some smart acquisitions to amass a warchest to fund a successful venture arm and an operating division to build its own ideas. At its peak during the dot-com boom, CMGI was a massive holding company [https://venturefizz.com/stories/boston/could-cmgi-have-been-google-boston] with a roster of storied internet brands: Lycos (search engine), GeoCities (web hosting), Planet Direct (portals), FlyCast [https://paleoadtech.com/2023/12/10/47-larry-braitman-fishing-for-dot-com-dollars-with-flycast/] (ad network), AdKnowledge [https://paleoadtech.com/2023/12/10/30-dave-zinman-on-the-first-ad-server-bluelithium-and-beyond/] (ad server), etc. It eventually had 5,000 employees and $1.5 billion revenue and was #7-9 in terms of web traffic to its properties. The New England Patriots’ stadium was called CMGi Field. [https://i0.wp.com/paleoadtech.com/wp-content/uploads/2024/08/cmgi-screenshot.png?resize=756%2C400&ssl=1] Daniel became CMGi’s acting CTO with the proto-database project as his day job. The company was based at Brickstone Square in Andover, Mass. After a privacy audit during his time at Fidelity, Daniel realized that there was actually no need to have personally-identifiable information (PII) to do the analytical work for database marketing: segmentation and list management. PII could be appended later, but it was not needed for analysis. “That’s what actually inspired online profiling,” he recalls. “We could raise the bar versus traditional direct mail because we did not need to know who you are in the real world in order to make … advertising and content relevant to you based on your interests.” So Engage was built on the Fidelity-inspired direct mail idea to build anonymous profiles, plus cookies. Daniel’s first technical challenge was scale, handling clickstream data and data from portfolio partners like Lycos. “I knew that the tools didn’t exist,” he admits. So he had to build them. He ended up crafting three products: * The first parallel-ETL tool for data processing * An analytics tool for discovery and iteration * Interest-based anonymous profiles (which they called “cybertargeting”) At the end of 18 months, the tools were real. The first (called Engage Fusion) and the second (Engage Discover) were soon sold to the data warehousing company Redbrick for about $10 million. CMGi kept cybertargeting, which became Engage. Engage’s solution for cybertargeting could process log data and online event data — and then create cookie-based profiles to target and measure marketing campaigns and analyze web traffic. By 1998, DoubleClick [https://paleoadtech.com/2023/12/10/11-kevin-oconnor-co-founding-doubleclick-and-more/] had gone public and CMGI wanted to get into the ad serving game. It acquired Accipiter [https://www.cnet.com/tech/tech-industry/cmg-buys-ad-server-accipiter/] for $35 million, a rival of the on-premise ad server NetGravity (later acquired by DoubleClick). In the spirit of building things that don’t exist, Daniel also filed what may well have been the first patent [https://patents.google.com/patent/US8219613B2/en] for cookie-synching. So while PaleoAdTech’s friend Lou Montulli [https://paleoadtech.com/2023/12/10/14-lou-montulli-building-browsers-cookies-and-more/] may have invented the cookie at Netscape, Daniel says, “I am the inventor of all the evil uses of cookies [wink].” There was also an algorithm that determined peoples’ interests from their browsing behavior and, eventually, got at various levels of interest. While Engage now offered an integrated data and ad server platform, it had challenges. It was difficult (aka impossible) to build an internet data-selling business until the real-time bidding era a decade later, when Exelate and BlueKai realized Engage’s original data-network dream. There was the persistent, rarely-discussed problem of decaying profiles and stale data. So while at one point Engage may have had 130 million unique profiles in its database, only a portion of them were actually useful. CMGi went public in 1999, launched its own ad network called AudienceNet [https://adage.com/article/news/engage-technologies-rolls-audiencenet/7888] (combining Accipter and Engage profile data), and was hammered of course by the dot-com meltdown that began in 2000, laying off staff and selling assets. For various dot-com financing reasons, the company didn’t have the cash to endure and eventually dissolved. After an exciting interlude at Tacoda — which we will (Marty hopes) discuss in a future episode [in the meantime, enjoy this previous episode [https://paleoadtech.com/2023/12/10/13-dave-morgan-keeping-it-realmedia-tacoda-and-simulmedia/] with Tacoda’s founder Dave Morgan], — Daniel resurfaced in 2010 with Korrelate, an ad analytics company. It emerged from an problem related to AOL’s acquisition of Tacoda: how to get AOL’s other properties, like Ads.com, to work with it. Daniel’s solution included a forerunner of header bidding he called federated RTB (sold to Operative Media) and Korrelate, which tied digital clickstream data to offline events like sales. It performed online-to-offline attribution. Korrelate could tie ad impressions to offline sales — for example, partnering with Polk for DMV registration data, — as well as give marketers insights into the impact of their digital channels on sales. In this way, General Motors learned that their online form wasn’t nearly as powerful as their configurator for predicting sales. Ultimately, Korrelate couldn’t compete with Datalogix [https://paleoadtech.com/2023/12/10/42-joseph-zito-the-logic-of-datalogix-and-oracle/], which sold media targeting data and provided its measurement service like Korrelate’s for free. “We were competing with free,” he says. “And I learned my lesson then about selling analytics standalone.” Korrelate ran into trouble [https://www.adexchanger.com/data-exchanges/online-to-offline-data-firm-korrelate-suspends-operations/] in 2014 and was sold to JD Power. Today at Aqfer, Jaye says, he’s still in a position to make use of his prolific technical imagination. “What I do,” he says, “is I sell the technology that I’ve had to build in every company … as a platform.”

65. Adam Singolda – the tabula rasa of Taboola

Adam Singolda is the Founder and CEO of Taboola [https://www.taboola.com/], a performance-focused advertising company he started in 2007 after spending seven years as a cryptological engineer in the Israeli Defense Forces. Today, Taboola is a public company with 2023 revenues of $1.4 billion (and growing nicely [https://investors.taboola.com/static-files/5dae2d16-4862-416a-852b-0ebb9be91ba3]), around 2,000 employees in 22 countries, 18,000 advertiser customers reaching 600M daily users. It acquired Connexity [https://techcrunch.com/2021/07/23/after-going-public-via-a-spac-taboola-acquires-e-commerce-marketing-network-connexity-for-800m/] (formerly Shopzilla) in 2021 for $800M, expanding its offering into commerce- and retail-focused recommendations and ads. It’s based in NYC and still has a large presence near Tel Aviv, where it was founded. Taboola took a circuitous route to its present incarnation. As Adam tells Marty in this lively episode, his vision after leaving the IDF was to start a company that was a video recommender system — to the solve the problem, he says, that he never knew what to watch on TV. Initial investors were friends and family and, on one memorable occasion, an acquaintance of his mother’s Adam accosted at a bat mitvah to which he had not been invited. Adam was always self-directed. Living in Israel as a kid, he tried to start a babysitting and tutoring syndicate and then a short-lived text-based news-and-entertainment alert company he called 24Go. And he had an entrepreneurial family: his father Avi Singolda [https://www.imdb.com/name/nm1294878/] is a well-known studio guitarist and band leader in Israeli who once played “Here Comes the Sun” with Sir George Martin (“the only time I’ve ever seen him nervous,” Adam says of his father). The name Taboola came from the Latin tabula rasa (“blank slate”), combined with the “oo” motif from successful internet companies Facebook, Yahoo and Google; and the domain taboola.com was available for $10 in 2007. The company had a hard time finding a business model, and Adam admits that in the beginning he didn’t know anything about the ad business (a trait shared with many PaleoAdTech founders). At first, he pitched publishers on the idea of paying a fee for a video recommendation engine; then he tried a revenue share from money made from recommended videos. Neither worked. “I almost shut down Taboola three times,” Adam says. “It’s a very hard moment because you’re very small, when you’re a company of 10 people or 15 people and you know, and nobody else does, that you have money for a month or two. It’s really stressful.” It was only in November, 2011 when for the first time Taboola allowed someone to pay them to be discovered on someone else’s site that “the revenue started to go up and to the right.” Adam doesn’t name his first client but admits it’s a large publisher we all know (like Forbes, perhaps). The idea of a content-and-article recommendation engine with sponsored links mixed in with organic links was born, and Taboola did not look back. The company went from almost no revenue in 2011 to $200M in 2014 and is on track to $2 billion by 2024 or so. It charges a CPC (which varies from pennies to dollars per click) or sometimes a CPM and shares some of the revenue with publishers. Its customers are “performance ninjas” who are looking for traffic and action, not upper-funnel branding. The algorithm uses pixels on the advertisers’ pages to see if ads result in actions, which improves targeting. So Taboola functions like a performance ad network. There were always competitors, including RevContent [https://www.revcontent.com/] and Outbrain [https://www.outbrain.com/]. The latter and Taboola were reportedly involved in merger talks for years that fell apart [https://techcrunch.com/2020/09/08/taboola-and-outbrain-call-off-their-850m-merger/] near the beginning of the Covid pandemic in 2020. In 2021, Taboola went public [https://www.nytimes.com/2021/01/25/business/media/taboola-public-spac.html] via a SPAC on the Nasdaq exchange at a valuation of about $2.6 billion. [https://i0.wp.com/paleoadtech.com/wp-content/uploads/2024/06/DSC09171-1.jpg?resize=782%2C522&ssl=1] Adam’s goals remain lofty and he admits he’s an optimistic person by nature. He sees Taboola as a content recommendation engine for the open internet (outside the gardens), and the acquisition of Connexity as a way to become the web’s product recommendation engine. Recent features include Maximize Conversion, which is a kind of autopilot for outcomes similar to Google’s PerformanceMax, and an AI-enabled ad-design tool.

64. Greg Smith – Boomerang-ing through DoubleClick, EchoTarget and Aniview

Greg joined DoubleClick in 1998 as a product manager after a stop in management consulting, and he became passionate about the early retargeting solution for advertisers called Boomerang. In 2005, he founded dynamic retargeting company EchoTarget, which was acquired by Acxiom [https://chiefmarketer.com/acxiom-acquires-echotarget/] in 2007. Today he is GM North America at Aniview [https://aniview.com/], an Israel-based video monetization platform for web, mobile and CTV. Joining the diabolically-hot DoubleClick in NYC in the ’90s via a headhunter, Greg was hired by product leader David Rosenblatt, later to become the company’s turnaround CEO. At the time, Rosenblatt ran a small group called Closed Loop Marketing, one of DoubleClick’s three lines of business. (The other two were Wenda Millard’s media network and the original DART ad server.) Rosenblatt’s Closed Loop Marketing unit also had three power plays, as Greg tells Marty in this far-ranging chat. One was what would become DART for Advertisers; another was Boomerang cookie-based retargeting; and the other was DataBank, an ad database that helped tracking and conversion optimization. Sensing its potential, Greg adopted Boomerang and convinced the bosses that it should be part of the media group, under Millard; and thus it was moved. It was an incendiary time, with the legendary parties (just mention “Oompah Loompah” to any staffer of that era and watch their reaction). The staff doubled in 2000, its last boom moment for a long while. Retargeting “was successful right away,” Greg reports. With an early customer “we doubled click rate” — but more important, customers like Victoria’s Secret were able to show much better click-to-conversion rates using retargeting. The music stopped — for everyone — in 2001, and Greg in time found himself at Cendant [https://en.wikipedia.org/wiki/Cendant], a megacorporation with some holdings in travel. While there, Greg was exposed to a technology which he believed could be configured into a dynamic retargeting solution. Rather than a static tag, like Boomerang’s, it could provide data about origin and destination of a trip, which might then be populated into an ad. Greg founded EchoTarget in 2005, using the dynamic tech and the Right Media and DoubleClick exchanges. Early customers like Spirit Airlines adopted what EchoTarget called “dynamic retargeting,” and the company claimed click-rate improvements [https://www.mediapost.com/publications/article/56035/bts-next-stage-custom-creative.html] from 2x-8x. It was an advantaged data solution, although competitors like Dotomi appeared. [https://i0.wp.com/paleoadtech.com/wp-content/uploads/2023/06/Greg-Smith-at-Adtech-0406.jpg?resize=480%2C360&ssl=1]Greg at the Ad:Tech event Eventually signing up non-travel customers such as Nestle Waters, EchoTarget’s portfolio remained about 70% travel-focused at the time of the company’s acquisition by Acxiom in 2007, for an undisclosed amount. Today, Greg is GM North America at Aniview, which he describes as an “end-to-end video platform company,” predominantly serving publishers. It offers a video player, video ad server, CMS and server-side-ad-insertion for CTV. The biggest component of Aniview’s business is a marketplace.

63. Jason Fairchild – developing GoTo, OpenX and tvScientific

Jason was the co-founder with his fellow Yahoo veteran Tim Cadogan of OpenX, a pioneering ad server and then programmatic exchange that launched in 2008. Today he is co-founder and CEO of tvScientific [https://www.tvscientific.com/], a CTV-oriented programmatic platform. Jason began his journey in business development in Southern California at one of the original ISPs, Earthlink [https://www.earthlink.net/], which persists to this day. Then living in Pasadena, he was an avid follower of the Idealab [https://www.idealab.com/] incubator, led by the monumental Bill Gross [https://en.wikipedia.org/wiki/Bill_T._Gross], who became a mentor. One of Idealab’s startups that Jason noticed was GoTo.com – and he joined as one of the company’s first two dozen employees, powering business development and deal-making. As Jason tells Marty in this penetrating episode, it “almost didn’t matter” what GoTo actually did, so eager was he to join the Idealab orbit. [https://i0.wp.com/paleoadtech.com/wp-content/uploads/2023/06/GoTo.png?resize=351%2C144&ssl=1] [https://i0.wp.com/paleoadtech.com/wp-content/uploads/2023/06/GoTo2.png?resize=225%2C225&ssl=1] As legend has it, GoTo [https://thehistoryoftheweb.com/goto-forgotten-search-engine/] started as an algorithmic search engine, until Bill Gross came back after a weekend’s thought and decided it should try something that had never been done before: rank results based on how much advertisers bid on the keyword(s). Paid search was born. It was a tough sell, as Jason admits. Advertisers had to become comfortable with keywords, bidding, measurement, a whole new vocabulary and method of paying for attention and intent. Ironically, it took the dot-com crash to catalyze GoTo’s business, as cash-strapped marketers realized search was a (seemingly) highly-accountable channel. GoTo (then renamed Overture) was acquired by Yahoo in 2003 for a reported $1.8 billion [https://www.latimes.com/archives/la-xpm-2003-oct-08-fi-overture8-story.html#:~:text=Yahoo%20Inc.%2C%20owner%20of%20the,advertising%20into%20Web%2Dsearch%20results.]. Jason and Tim Cadogan founded OpenX in 2008 with the idea of building a “more real-time” version of Right Media. Originally offering an open source ad server, the founding pair assumed their thousands of ad-serving customers would happily join a proto-exchange and provide a willing source of inventory. They were wrong. What followed was a “brick by brick” assembly of an exchange, in the midst of the financial crisis. It was slow going. The build required recruiting a publisher, putting them on the exchange, then finding a buyer; and so on. What tipped the scales was a deal with Fox Audience Network, which was flush with user-level data from its MySpace acquisition and looking for a way to use that extend its inventory elsewhere. OpenX joined with the MediaMath DSP to do an experiment with Fox that provided an “aha moment” — and OpenX shifted from slow-growth to hyperspace. [https://i0.wp.com/paleoadtech.com/wp-content/uploads/2023/06/IMG_1251-2.jpg?resize=782%2C587&ssl=1] [https://i0.wp.com/paleoadtech.com/wp-content/uploads/2023/06/IMG_6556.jpg?resize=767%2C1024&ssl=1] [https://i0.wp.com/paleoadtech.com/wp-content/uploads/2023/06/openx-celebration.jpeg?resize=360%2C640&ssl=1] During his wind-down at OpenX, as Chief Revenue Officer and global expander, Jason started to become interested in Connected TV as a programmatic possibility, and by 2020 he has left OpenX to co-found tvScientific. Bill Gross was an early champion and investor, as was #PaleoAdTech friends’ Joe Zawadazki and Eric Franchi’s Aperiam Ventures [https://www.aperiam.vc/]. tvScientific’s mission is to make CTV as easy to buy as search and social, activating a heretofore dormant long tail of smaller advertisers outside the top 500 who dominate 85% of TV ad spend.