In The Plex

by Steve Levy

  What did Patel learn about being CEO? “Anything that’s wrong sort of bubbles up, so you have to deal with all these problems that aren’t the sort of problems I would want to solve,” he says. “It’s not a job I would want.”

  He had a better job, anyway. He was an engineer at Google.

  The fact was, 2001 was a tough time to be Google’s CEO. Funds were getting so low that Schmidt instituted a tight-pocketbook policy that limited expenditures to one day a week: if an executive wanted to spend money, he or she would have to petition Schmidt for approval in his office at 10 A.M. on Friday. The VCs were screaming bloody murder. Tech’s salad days were over, and it wasn’t certain that Google would avoid becoming another crushed radish.

  Then came a development that was sudden, transforming, decisive, and, for Google’s investors and employees, glorious. Google launched the most successful scheme for making money on the Internet that the world had ever seen. More than a decade after its launch, it is nowhere near being matched by any competitor. It became the lifeblood of Google, funding every new idea and innovation the company conceived of thereafter. It was called AdWords, and soon after its appearance, Google’s money problems were over. Google began making so much money that its biggest problem was hiding how much.

  2

  “When we became profitable, I felt like we had built a real business.”

  “I hate ads,” says Eric Veach, the Google engineer who created the most successful ad system in history.

  Veach hailed from Sarnia, a small city in Ontario, Canada. The son of a chemical engineer and a chemistry teacher, he’d been obsessed with math from an early age. He was on the national team in the Math Olympiad, won a contest for a scholarship at the University of Waterloo, and placed in the top twenty in the prestigious William Lowell Putnam Mathematics Competition. After earning a computer science degree at Stanford, he got a job at Pixar, working on the software that renders computer images into lifelike animation. (If you squint, you can see his name in the credits of A Bug’s Life, Toy Story 2, and Monsters, Inc.) He liked the work but felt his group at Pixar was “screwed up politically”—he’d had two managers in two years—and began looking for a new gig. He was impressed at the technical chops of the people who interviewed him at Google, so he joined the company in 2000. He found himself working on ads. “At the time, it was a backwater of the company,” he says. Seven people worked there.

  Considering Veach’s loathing of advertising, it was an interesting job switch. But contempt for traditional advertising permeated Google from the top down. In their original academic paper about Google, Page and Brin had devoted an appendix to the evils of conventional advertising. The founders weren’t sure what their ads would be but were adamant that they somehow be different.

  When Veach arrived, Google’s search ads were plain blocks of text that were deemed relevant to the search query that a user typed into Google’s search engine. The text blocks had highlighted links that led to a page on the advertiser’s website known as a landing page. This had two advantages over traditional advertising: the ads were more effective because they related to what people were looking for at that very moment, and the clicks that registered interest by users could be tracked by Google in its logs. Nonetheless, the early Google ads worked like traditional ones in one key aspect: the advertiser was billed according to how many people viewed the ad. This CPM (cost per thousand) model was the basis of almost all ad markets.

  Google ads were sold by actual salespeople. The head of the New York sales force was Tim Armstrong, a tall, engaging veteran of the brief dot-com boom who had majored in sociology and business at Connecticut College. He’d been captain of the lacrosse team. Armstrong had been impressed with Sergey Brin during a breakfast job interview when Sergey made a compelling argument that Google wanted its ads to be not fluff that imposed itself on users but important information that its users wanted. While Google expected to make most of its money from licensing, Armstrong was told, advertising might one day account for as much as 10 to 15 percent of its revenue. Not long after he took the job, a media director at an agency he’d worked with lectured him on the huge mistake he was making. “I don’t know much about this place Google,” the director said, “but I can tell you that whatever it is, it’s not advertising—you should get out of there as quickly as possible.” Nonetheless, Armstrong hung on.

  Brin emphasized frugality—Eric Schmidt would often admiringly say, “he’s cheap”—which Armstrong experienced firsthand when he began signing up customers. The standard way to confirm an ad buy in the business was faxing the insertion orders. But when Armstrong ordered a fax machine, he got a call from George Salah, Google’s director of facilities. “Larry and Sergey want to know why you need a fax machine,” Salah said. Armstrong explained about insertion orders. Then he got another call. This time, Larry and Sergey wanted to make sure there would be enough sales in the pipeline to justify the cost of the machine.

  Google’s name for the ads from big accounts that Armstrong visited was “premium sponsored links.” They were positioned on top of the search results, against a background of yellow to distinguish them from the search results. Most of his team was in New York City, the hub of the advertising world. (His apartment on the Upper West Side was unofficially the first Google office in New York.) As salespeople had done for nearly a century, Armstrong’s team took customers to dinner, explained what keywords meant, and told advertisers what it would cost to buy ads, which were priced according to the number of people who saw them.

  But Google wanted something that would work on Internet scale. Since Google searches were often unique, with esoteric keywords, there was a possibility to sell ads for categories that otherwise never would have justified placement. On the Internet it was possible to make serious money by catering to the “long tail” of businesses that could not buy their way into mass media. (The long tail is the term used to refer to smaller, geographically disparate businesses and interests. The Internet—particularly with the help of a search engine like Google—made long-tail enterprises easy to reach.) If you made the system self-service, you could handle thousands of small advertisers, and the overhead would be so low that customers could buy ads very cheaply. So in October 2000 Google launched a product catering to smaller operations that had not previously contemplated an online buy. (Armstrong’s team kept selling premium sponsored links to big advertisers.)

  Google named the self-service system “AdWords.” It was a do-it-yourself marketplace for keywords, purchased by credit card. When someone came to Google and searched using one of those keywords, a few words of text with a link to the advertiser’s home page would appear. The ad would be very similar to a search result, only paid for. Those ads ran to the right of the search results, as suggested by an adviser to Google, the Israeli high-tech investor Yossi Vardi. If you drew a vertical line two-thirds of the way across the page and put text ads to the right, he told Brin one day, it would be clear which were the real algorithm-discovered search results—known as “organic” results—and which were paid links. Google also made sure to label the ads “sponsored links” to further distinguish them from the purity of its organic search results.

  AdWords prices were fixed according to the position on the page an ad would occupy. If it was in the most desirable position, the top ad on the right, the client would pay $15 per thousand exposures. The second position cost $12, the third $10. There was one feature built in to try to ensure that the most useful ads would appear: advertisers couldn’t pay their way to secure the best positions. Instead, the more successful ones—the ones that lured the most people to click on them and go to the advertiser’s landing page—would get priority. The percentage of people exposed to ads who responded to them became known as the click-through rate.

  This was Google’s first stab at what became known as ad quality. It would become a vital component of the company’s strategy, which viewed the ad system as a virtuous triangle with three happy parties
: Google, the advertiser, and especially the user. Unwanted ads made unhappy customers, so Google made it a high priority to calibrate the system to drive out ads that were irrelevant or annoying.

  One day in October 2000 the engineers who coded the system tested AdWords with a little text ad of their own that read, “Have a credit card and 5 minutes? Get your ad on Google today.” It was shown to only a small number of users. Within minutes, someone had clicked on it and began filling out the form. And barely a half an hour after that, someone who typed in the words “Live Lobsters” on Google would see a “sponsored link” on the right side of the search results that read, “Live Mail Order Lobsters,” placed by a small business called Lively Lobsters that had never previously placed an online ad.

  Though the system quickly became popular, it was too easy to game. Advertisers had a huge incentive to click on their own ads to generate a high click-through rate and thus improve the position of the ads in subsequent searches.

  As a consequence of the VC pressure on Google to make some real money, Page and Brin had instructed Salar Kamangar to look into ways to make more money with the ad system. In November 2000, Kamangar visited Veach, and as they spoke Veach realized that Google’s desperate financial situation would give him an opportunity to use his mathematical expertise to improve the concept of advertising. Maybe, he thought, he could even make advertising itself less hateful. Veach believed that a well-placed search ad could be more useful than a search result. They began working together.

  Every week or so, Brin or Page, sometimes both, would come by to toss ideas around and ask why the system wasn’t done yet. Page was adamant that the system be simple and scalable. He thought that the system should be so easy for advertisers that all they would need to do was give their credit card number and point Google to their website. They shouldn’t even get involved with choosing keywords—Google would choose them. That was an idea that made sense, though many advertisers always want a say in choosing keywords.

  Some other suggestions from Page, though, were baffling. “Larry always has far-fetched ideas that may be very difficult to do, that he wants done now,” says Veach. During one session, when discussing the fact that not all countries commonly use credit cards, Page proposed taking payments in barter appropriate to the home country. For instance, Page suggested, for transactions in Uzbekistan, Google could take its payment in goats. “Maybe we can get to that,” Veach responded, “but first let’s make sure we can take VISA and MasterCard.”

  One of the key breakthroughs came when Veach and Kamangar decided to use auctions to sell ads. It made perfect sense. In a dynamic marketplace, auctions allow you to find the sweet spot where buyers and sellers both win. The source of their idea was the business model of one of Google’s competitors. GoTo was the brainchild of one of the most fecund minds of the Internet age, an energetic Caltech grad named Bill Gross. Gross’s IQ and geek factor were both off the charts. He began to make a name for himself in the 1980s as an entrepreneur who came up with ideas that applied clever technological tricks, often ones that exploited tempting market niches.

  During the late 1990s Internet boom, Gross created Idealab, a company that would incubate new companies. He envisioned creating several tech start-ups a year, rolling them out the way a movie studio launches films. During the next few years, several Idealab companies had smashingly successful IPOs—and even more spectacular crashes when the music stopped in 2000. But one Idealab company had emerged as a winner, its search company GoTo.

  In a way, GoTo was a Bizarro-world version of Google. Whereas Google had skyrocketed to fame as a search engine with innovative technology and no discernable way to make money, GoTo got pans for its search strategy, specifically its mixing of paid and organic search results. But its revenue model was brilliant. Gross’s basic model was Yellow Pages ads, in which businesses paid a premium to place their ads in the relevant category. The biggest impact was made by a full-page ad, and the equivalent of that in a search engine was a high place in search results. Gross’s innovation was to have advertisers compete for those places: to get your ad in the search results under a given keyword, you had to outbid other advertisers in an auction. His colleagues didn’t warm to it. “Everybody in the room had a look on their faces like, ‘You’ve gone nuts.’ But I kept pitching it, and they admitted that there might be something to it, but it would be controversial,” he says.

  As Idealab prototyped the idea, Gross had another one. Every month he would gather the CEOs from his fifteen or so companies and have them compare how much they paid to get traffic to their websites through the banner ads that were then the only form of Internet advertising. The most useful metric was arrived at when the cost of the ad was divided by how many times someone clicked on a banner and actually went to a site. Even though ads were paid for according to how many people saw them, it was the clicks that made them worthwhile. “So the thing hit me,” says Gross. “Why don’t we make a search engine where you just pay by the click?” That way, advertisers could know the values of ads from the start.

  Gross announced GoTo at the TED conference, a high-profile industry conclave, in February 1998. His presentation introduced the hugely innovative pay per click and auction, but what stuck in people’s minds was that GoTo’s paid search results showed up in the sacred territory of organic results. Techno-pundits viewed the ethics of search engines like the ad/editorial separation in newspapers and magazines. There seemed something fishy, even venal, in selling results that would be intermingled with the best guesses of algorithms. (For its nonpaid results, GoTo licensed search engine technology from Inktomi.) The audience at TED, where even fairly tepid presentations often get standing ovations, actually hissed during Gross’s demo. (Page and Brin considered GoTo’s mixing of paid and organic links an abomination.) “It was very distasteful to people,” says Gross. “But I didn’t consider that the paid links were part of the organic results.”

  GoTo’s search capabilities weren’t strong enough to lure users to its site. Instead, Gross paid other Internet companies to use GoTo in the search engine they offered visitors, figuring he’d come out ahead when people clicked on the ads. His biggest and most successful arrangement was struck in late 2000: GoTo paid AOL $50 million to become its search engine. When AOL’s users did a search, they would see Inktomi web results mixed with GoTo’s ads. In 2000, GoTo reaped $100 million in revenue and as was customary in the dot-com world, it went public while still in the red. The IPO brought in a billion dollars.

  In all the excitement, GoTo made a big mistake of omission. “We were ready to go public and were on fire, revenues going through the roof and all that, and were getting our IP [intellectual property] portfolio together for the bankers, and everybody was like, ‘What patents do we have?’ And we didn’t have too many,” says Gross. Worse, since patents had to be filed within one year of public exposure, GoTo had missed the window to patent ad sales with real-time auctions and pay per click. All GoTo could do, Gross says, “was patent everything else we could think of, a bunch of obscure things like the way we accepted bids. These were silly patents, but the real patents would have been worth billions.”

  In 2001, GoTo changed its name to Overture. The new moniker reflected the direction the company had taken. Very few people thought to “go to” Gross’s company. Instead, like a musical introduction, Overture, embedded in various portals such as AOL, was a prelude to an ultimate destination. Gross himself felt that the approach was misguided. Originally, he had thought of GoTo as a consumer brand. That was gone. “We thought we could win more deals by only being a service provider and not having our own site. It was the beginning of the end for us, but Overture was still worth a fortune.”

  Google knew all about Overture, of course. At the TED conference in 2001, Gross had actually suggested to Page and Brin that the companies merge. The Googlers would have nothing to do with any system that mixed organic search results with ads. Still, they wondered whether taking over Overture’
s contracts would solve some of their revenue problems at the time, and there was talk of a partnership. Bill Gross even ginned up a demo called GOTOOGLE, with two columns of results, one of them the Google organic results and the other GoTo’s paid result. But Salar Kamangar successfully argued against any kind of deal, saying that Google could do it alone. He was sure that he could build a better system, beating Overture at its own cost-per-click, auction-ized game.

  Eric Veach particularly disliked one aspect of the Overture auction system: the fact that advertisers were bound to pay the amount they had bid, even if the next lowest bidder had offered significantly less. “That means that advertisers always have an incentive to lower their bids [in subsequent rounds],” he says. (This was known in the auction world as “bid shading.”) As an example, he would cite the case where an advertiser bid 50 cents and the next highest bidder offered only 40 cents. Clearly the high bidder would be unhappy, because the optimal bid was 41 cents, and the winner was stuck with paying nine cents too much. A cottage industry of software vendors had provided programs to automate bid shading on Overture, so winners would keep submitting slightly lower bids, and losers would edge up. “I wanted to avoid that cat-and-mouse game,” says Veach.

  So Veach devised a different model: the winner of the auction wouldn’t be charged for the amount of his victorious bid but instead would pay a penny more than the runner-up bid. (Example: If Joe bids 10 cents a click, Alice bids 6, and Sue bids 2, Joe wins the top slot and pays 7. Alice is in the next slot, paying 3.) It was incredibly liberating because it eliminated the fear of “winner’s remorse,” where the high bidder in an auction feels suckered by paying too much. In the Google model, no one would feel like an idiot for paying a dollar a click when the competitor below them bought a slot on the same page, positioned just a few pixels lower than their ad, for only 10 cents a click. In that case, a winner would get the prime position for 11 cents.