by Steve Levy
Google used this approach to expose the system to some of its biggest potential customers, huge digital publishers such as The New York Times.
All this advertising involved a considerable outlay of cash. Eric Schmidt thought it excessive. That would ultimately lead to one of Schmidt’s “and-it-turned-out-they-were-right” anecdotes: “Sergey walked in and said, ‘I’m going to invent this business,’ and I said, ‘Fine.’ Then he said, ‘I need money because I’m going to preguarantee these deals.’ And I said, ‘That’s bad business,’ and we argued for a good half an hour, because Sergey wouldn’t give up. Eventually, I said, ‘Okay, take a million dollars.’ Which was a lot of money for us then. Two months later, he comes back and they’ve spent a million and a half.
“Sergey, you couldn’t just stick to the million, could you?” Schmidt asked. Brin just gave Schmidt a broad smile.
“We could do that because we had the money,” says Liebman. “Google’s mantra was ‘Get this puppy launched.’” (In fact, the AdSense codename was “Puppy.”) After a few months, Google announced that it was no longer going to buy the ad space outright. By then it had enough data to show that if publishers signed up for the official AdSense program and kept running the ads, the commissions they earned from advertiser-paid clicks would justify the commitment.
It was an easy sell, because publishers could use AdSense for only the spots on their pages that otherwise would go unsold. Advertisers loved it because it gave them a chance of getting their ads on prestigious sites, such as The New York Times or Forbes. (A later twist in the program let advertisers specify where the ads would run.) “Those were brand names that our people could sell,” says Liebman.
The only hitch in the program was the risk that the ads Google placed on a website would be inappropriate or even offensive. When human beings created an ad for a publication, they took care to avoid situations where the combination of a certain ad with a certain type of article would produce a tasteless match that would appall readers and win no business for the advertisers. Google’s algorithms weren’t so sensitive. “The editors would get freaked out,” says Liebman. Some of the unintentionally offensive matches became classics. Liebman would cite an ad that ran alongside a gory murder story in the New York Post: someone had chopped up a body and stuffed it in a garbage bag. Alongside this gruesome text was a Google ad for plastic bags.
“We didn’t foresee that there were times when you don’t want to target ads to the content,” says Georges Harik. “We would analyze a page about a plane crash and happily place an ad for airline tickets. I think we rapidly discovered that this was a bad idea.” Google engineers started working on ways to mitigate this problem, but it would never be totally eliminated. It was just too hard for an algorithm trained to discover matches between articles and ads to exercise human good taste. In 2008, a story about the Mumbai attacks headlined “Terrorists kill the man who gave them water” was accompanied by an ad that read “Terrorism: Pursue a certificate in terrorism 100% online. Enroll today. Ads by Google.” An account of massive food poisoning at an Olive Garden restaurant in Los Angeles was accompanied by a coupon offering a “FREE Dinner for Two at Olive Garden.”
As the program continued, Google opened up AdSense on June 21 to long-tail businesses such as blogs and small-business websites, using a self-service model. Sergey Brin joined the team to monitor its progress. They sat in rapt attention as one publisher after another signed up, pasted the line of JavaScript code in their HTML code that would enter them in the program, and began hosting ads on their websites and blogs. The team hung out at the offices until 3 A.M., transfixed by the response. A few months later, again at Brin’s urging, they “localized” the product, making it available in ten languages. That doubled the business.
Even beyond the revenues, AdSense was important to Google because it showed that the company could make money outside of search. “You can think of the search engine as the crown jewel of Google,” says Gokul Rajaram, the AdSense product manager. “With a program like AdSense, Google was able to make money from its partners—it was kind of a moat that protects the king’s castle.”
One aspect of AdSense was reminiscent of the black box that determined the ad quality rating of AdWords. When someone clicked on an AdSense ad, the money paid by the advertiser was split between Google and the publisher whose site hosted the ad. According to Rajaram, the original thought was to split the money down the middle—Google would take half and the AdSense publisher would take the other half. But Brin thought that such a split gave too much to Google. The idea was to build the program for the long run, and if Google made it clear that it was taking half the money, a competitor might undercut the program by giving 80 or even 90 percent of the fee to the publisher. So Google decided to give the majority of the money to the publisher. Then Susan Wojcicki came up with an idea that some might find strange: What if we don’t reveal the revenue share percentage with the publisher? That way Google wouldn’t have to worry about a competitor boasting a better split.
Gokul Rajaram was startled by the idea. “What?” he said. “How can we not tell the publishers what percentage of the revenue they get?”
“The publishers shouldn’t care about the revenue share,” Wojcicki told him. “What they should care about is the bottom line.”
On the face of it, the scheme flew in the face of Google’s stated goal of transparency. But when Google decided that it was in its interest to keep something secret—as the “hiding strategy” showed—it reconsidered. The dichotomy of closed and open at Google might be traced to Larry Page’s personal library. On one hand there was Don Norman’s book, a polemic urging total fealty to the user, making everything clear to customers and visitors to a site. On the other hand, there was Nikola Tesla, exploited and dying alone in a New York City hotel. With regard to the revenue split in AdSense, Google wasn’t going to be Tesla. Besides, why would publishers complain if Google made them more money than the competitors?
“It was one of the single biggest painful things for me,” says Rajaram. “On every panel I went to for the first year, I would get questions about why isn’t Google sharing the revenue split and why isn’t Google being transparent. People said we were doing it because we weren’t generous. But quite to the contrary, we were being generous. We just didn’t want our competitors to tell publishers that they were offering a better revenue share.”
(In May 2010, Google finally revealed the split. “In the spirit of greater transparency,” Google reported that of the money received from advertisers on AdSense for content, 68 percent went to the publishers whose pages hosted the ads. Google kept the other 32 percent. That was close to the proportions that participants and analysts had long assumed. Google’s belated announcement only raised more questions as to why it had been a secret in the first place.)
Google took a viral approach to luring small businesses and bloggers to run ads. For bloggers who didn’t have many visitors to their sites, the money they would get from AdSense commissions was a piddling sum, arguably not enough to sully the pure relationship between an impassioned citizen-writer and engaged readers. On the other hand, it was free money. To help build the program, Sheryl Sandberg hired an old friend.
Kim Malone had, in Sandberg’s view, “the perfect Google résumé,” at least for a nonengineer. A Princeton grad (like Eric Schmidt), she had studied Russian literature, venturing into the mother country after graduation to write reports on converting military technology to private-sector uses. Then she started a diamond-cutting factory in Moscow. She returned to the United States to get a Harvard Business School MBA and emerged as a founder of high-tech companies. After Malone went through three start-ups—and two unpublished novels, because she believed that her true calling was fiction—Sandberg urged her to apply to Google. Twenty-five interviews later, she arrived in Mountain View to help sell AdSense to more small publishers.
Malone dubbed herself “the high priestess of the long tail” and came to v
iew her job—the dubiously virtuous task of sticking ads on pages that otherwise would have been ad-free—as a mission to empower small publishers in the age of search. She considered the AdSense arrangement the best sales pitch in advertising history. “Here’s the deal,” she’d say. “You take ten seconds to put a little snippet of code on your website, and from that moment on, Google sends you a check every month.” It became evangelical for her. “I’d tell people that AdSense funds creativity a nickel at a time,” she says. “So if you have an idea, you can start making money on it with AdSense immediately without having to get it published or to raise money from venture capitalists.”
One of the first things she did was become an AdSense customer herself. She formatted one of her unpublished novels as a blog and applied for the program. To her astonishment, Google’s algorithms rejected her, citing excessive profanity. “It made me think that we should reevaluate our policies,” she said, and she began to introduce different means to segment AdSense customers. The idea was to make the program as inclusive as possible. If your site was profane (but not pornographic—there were still boundaries), Google would place ads only from customers who were okay with that.
Even though Malone’s domain was limited to the small players on the Internet—if the page views of a site exceed a certain limit, it would be turned over to direct sales—she understood that her potential customer base was vast. By making it simple to sign up and collect money, Malone saw the number of AdSense publishers grow dramatically. In early 2004, she gave a progress report at a Google Product Strategy meeting. Eric Schmidt asked her how many publishers AdSense had signed up. He figured the number would be in the thousands. But she reported a number well up in the hundreds of thousands. “They almost fell off their chairs,” she said. But not Brin. “That’s pretty good,” he said. Later people would explain to Malone that Brin’s “pretty good” is the equivalent of a Nobel Prize.
As AdSense kept delivering more and more of Google’s revenues, Malone thought of a song from the musical Evita:
When the money keeps rolling in, you don’t ask how
Think of all the people guaranteed a good time now
Later in the year, AdSense achieved a milestone in its run rate—$1 million a day. Kim Malone wanted to have a party for her team, but Google had gone into an advanced mode of its “hiding strategy.” By then its IPO had been announced, and though some of Google’s numbers were public, its stealth had morphed into something more subtle. It was increasingly clear that beyond the minimum disclosures required by law, Google was going to keep the most important figures to itself—such as how much its individual products were making. (In contrast, Microsoft would break down the figures for each of its divisions.) So Malone had to negotiate the terms of her celebration with Google’s lawyers. Only after some hardball dickering did Malone win a concession: she could bring in a cake to mark the milestone. But the lawyers were adamant that the actual milestone—the million-dollar number—could not appear on the cake.
It didn’t really matter. The $1 million a day would soon be $2 million. And more.
And the $2 billion that Brin had brashly predicted to the Applied Semantics group would eventually produce $10 billion for Google—every year.
4
“The barometer of the world”
While AdSense was a great success, the bulk of Google’s revenues came from AdWords. Eric Veach and Salar Kamangar’s auction-based AdWords Select product had first been thought of as a supplement to the more traditional, impression-based ads in the premium program, which was now called AdWords Premium. But it was working so well that Google would sometimes allow its auction-based ads to break out of their side-of-the-page ghetto and leapfrog to the premium zone sitting on top of the search results. If Google felt that the outcome would raise more revenue, a select ad would “trump” a premium ad and knock it out of that coveted position. As more and more auction-based ads trumped the hand-sold premium ads, Kamangar argued that Google should entirely end the practice of selling premium ads by a sales force that set prices and charged by impression. He set up a project, code-named D4, to implement the idea. Most Googlers called the plan Premium Sunset.
Even as he argued for it, Kamangar had his concerns about the shift. Customers used to certain prerogatives might balk at a system determined totally by auction and algorithms. For instance, it was common for a big advertiser to insist that its ad be the first one to appear above the search results, so that its impact wasn’t mitigated by a competitor’s ad above it. Also, moving to auctions would introduce uncertainty. Clients and agencies were used to guarantees that if they budgeted a specific sum of money they’d get a specific number of ads in predictable positions. Finally, some advertisers didn’t want to budge from impression-based ads. They would insist that their ads were intended to build up their brands and having a percentage of people clicking on their ads wasn’t as important as having lots of people see the ads.
Eric Veach believed that the data showed that the auction-based, pay-per-click model was actually better for everybody. The key was the ad quality, which made sure that ads would appear before sympathetic eyeballs. He did a close analysis and concluded that ads bought through AdWords Select performed better. He also uncovered hard proof that some premium advertisers were paying way too little for some valuable keywords. Armed with a PowerPoint presentation full of this information, he went to an executive meeting and argued that there was no reason to cater to the statistically unsound assumptions of big advertisers. “We should just make our advertisers live with it,” he said.
How did that go over? “Like a lead brick,” he recalls. But the disagreement came from the business people at Google. The engineers, he noticed, were behind it a hundred percent. And considering that this was Google, such support made the adoption of Veach’s solution inevitable. He had the data on his side.
Indeed, after months of wrangling, the ruling troika, nicknamed “LSE”—Larry, Sergey, Eric—signed off on the plan. Sales head Tim Armstrong thought that 99 out of 100 companies he was familiar with would have hemmed and hawed and decided to test some more and revisit the idea in six months. But Google was going for it.
Google had already used scale, power, and clever algorithms to change the way people accessed information. By turning over its sales process entirely to an auction-based system, it would similarly upend the entire world of advertising, removing guesswork-ridden human intervention. It would also provide a leveling function among its customers. “We would have everyone competing for the same ad position,” says Schmidt. “So regardless of whether you were a large company or a small one, you had to bid, and you had to bid at market value.”
Nonetheless, the move would be painful. It meant giving up campaigns that were selling for hundreds of thousands of dollars, all for the unproven possibility that the auction process would generate even bigger sums. “We were doing $300 million in CPM ads and now were going to turn this other model on and cannibalize that revenue,” says Tim Armstrong.
The role of Google’s advertising sales force had always been awkward. Its members had long suspected, not without foundation, that Larry Page wanted to do away with them entirely. At one point, Sheryl Sandberg made a major presentation to Larry, Sergey, and Eric, arguing that with the ad model’s success her team needed reinforcements for things such as ad approval, organization, and management. She thought the presentation was going great. Then Page chimed in. “I have a question,” he said. “Why do we need this team?”
Brin wasn’t terribly engaged with salespeople, either. In December 2001, Google had run its first sales conference, at the Hilton Garden Inn in Mountain View, a mid-price hotel a few miles from the Googleplex. Jeff Levick, who had just been hired by the company, recalls Brin dropping into the tiny conference room where the team of maybe twenty people was huddled. Brin ignored the conversation and instead remained in the back of the room, where the controls of the audiovisual system were located. “Everyone is talki
ng about what’s going on with sales, and Sergey was paying no attention, just pushing buttons on the AV system and trying to unscrew a panel to understand it,” says Levick. “And I remember thinking, this man does not give a rat’s ass about this part of the business. He doesn’t get what we do. He never will. That set the tone for me very early in terms of the two Googles—the engineering Google and this other Google, the sales and business side.”
No matter how much you exceeded your sales quota, a salesperson wouldn’t be coddled as much as a guy with a computer science degree who spent all day creating code. And some tried-and-true sales methods were verboten. For instance, golf outings. “Larry and Sergey hate golf,” says Levick. “Google has never sponsored a golf event and never will.” There would be days when Google salespeople would call agencies and discover that everybody was off on a golf retreat with Yahoo. But Tim Armstrong would tell his troops, “They have to take people on golf outings because they have nothing else.”
The salespeople at Google did have something special, and they were terrified that a change would kill the golden goose. They had worked hard to overcome the reluctance of advertisers. “We spent a tremendous amount of time trying to figure out how to get people to believe in relevancy,” says Armstrong. But AdWords Premium was working. Salespeople were assigned sectors of the economy and they would ring up contacts, if they had them, or just cold-call, and explain the concept of targeted keywords. One difficulty was that agencies were used to discrete ad campaigns where they ran something for a few months, shut down, and then ran something else. Google’s idea was that you could have something running all the time, measure the results, and reinvest as long as the payoff is positive. It had data to prove it.