That still left a need for translation of our ad products, user—support responses, licensing, and operations. The task of building a globalization group within Google to accomplish these things had been tossed around like a beach ball at a rock concert. Various outside translation agencies billed us enormous amounts for work that sounded awkward to native speakers. No one in the company had any real experience with internationalization (known as "i18n" for short because there are eighteen letters between the word's initial i and its final n). So, as with user support, this non—engineering function found a home in marketing's realm.
Fortunately, the responsibility came with a "Sergey." To get control of the accelerating hiring within the company, the executive team had decided to allocate positions by department, with each approved job opening represented by a laminated photo of Sergey. The departments had some flexibility in the way they allocated their Sergeys, but when you showed up at the hiring committee, you needed to bring three "enthusiastic endorsements" for your candidate from current Googlers plus a Sergey to trade for the proposed hire. No Sergey, no hire.
It had taken us a while to post our position, interview candidates, and extend an offer, so it was August 2001 before Stephanie Kerebel, a native of France, joined our group as globalization manager. She had years of experience in dealing with professional translators and immediately implemented cost—saving measures, such as paying for translation by the word and not the job. That alone cut our expenses in half.
We had been directing our professional translators to use the console alongside the volunteers, to save time. We needed professionals to help with our most important languages because waiting for volunteers would delay product launches. There was also a risk that volunteers might intentionally sabotage us with bad translations, a risk we were unwilling to take with popular interfaces that might reach millions of users.
Stephanie saw other limitations to our system as well and recommended we supplement it with Trados, the industry standard translation—management software. Trados had a number of features useful to translators, including version control and a customized glossary that increased in scope over time, ensuring consistency across multiple projects and speeding translation while reducing the cost. Professional translators found it helpful, especially those who did not have an easy way to do all their work online. Stephanie announced she was buying a copy of the software for Google.
And so the i18n war began.
Marissa fired the first shot. The ellipsis at the end of "Forse cercavi ...," our Italian translator's rendering of "I'm feeling lucky," felt awkward to her, so she removed it. She admonished Stephanie to tell our contractors to "pay attention" and not create formatting issues. The head of our Italian office replied that he preferred the original translation: the three dots made the phrase more elegant. I asked Marissa why she had overruled two native speakers and our localization expert. The cork came flying out of the bottle.
Marissa claimed that one of the other engineers thought it looked as if our site had been hacked because the punctuation was unusual. She told me she had delayed pushing out the new Italian interface because of her translation concerns, meaning we had violated the agreement that localization would never delay a launch. And while she was on the subject, Marissa poured out a litany of issues with marketing's i18n approach: the slowness of professional translators, the time required for quality assurance, the resistance of marketing staff and translators to using the translation console. The capper was our decision to buy Trados, which would increase engineering costs because someone would have to insert translations manually. Engineering didn't have time to build Trados features into the translation console as we had requested. Why, Marissa wanted to know, couldn't we just hire people who were not only good translators but also comfortable working in the translation console?
Wayne Rosing, the head of engineering, weighed in with his perspective. He wanted to keep Google's back—end technology as uncomplicated as possible. Google ran faster with fewer systems, and each new technology we introduced slowed our progress like a remora attached to the streamlined body of a great white shark.
I hadn't realized there was contention around our use of Trados or problems with how localization was proceeding. It seemed to me things had improved enormously. Quality was up. Costs were down. I needed more data to support our position, so I talked with the one translator I knew who had the tech chops not to be intimidated by the console or any other online tool.
Dennis Hwang—the same Dennis who drew homepage doodles—had used the translation console to render our interface into Korean. In his opinion, the technology had two major flaws. The most serious was a lack of context. Translators using the console saw one word or phrase at a time, without any awareness of where on a page it might appear. So when a translator saw the word "Bulgarian," he didn't know whether to use the word for a Bulgarian man or the Bulgarian language. And since each piece of text might go to a different translator, there was no continuity in the flow. Once it was reassembled, it often read awkwardly or even nonsensically.
The other problem was that when the console presented a word to be translated, it took the translation and pasted it into exactly the same place the English word had been, and with the exact same formatting. That led to inappropriate bolding, italics, and spacing that could change the sense of a phrase.
I shared Dennis's feedback with Marissa. It was unlikely we'd find translators who were more comfortable with the console than Dennis, and if he felt there were problems, we should probably address them. I was sensitive to burdening engineering, but the primary concern of the marketing group was that the finished translation not embarrass us as we tried to build market share in a new language.
The engineers saw clearly that the problem was the people, not the technology. If we could improve the quality of the translators, we could get by with just the translation console. From my perspective, translators were not engineers and shouldn't have to waste time learning to use software mismatched to the task at hand. Even Marissa admitted having felt frustration while translating our site into Bork, Bork, Bork.
We negotiated a compromise. Work on all but the top eight languages* would be done exclusively by volunteers in the console, while professional translators could use Trados. Those translations would be imported into the console by a yet-to-be-developed automated script. That script remained yet-to-be-developed for years, but the agreement meant we had clear guidelines for moving ahead. The tradeoff was accepting the risk of unprofessional translation in all but a handful of languages. A couple of months later Sergey noticed that a volunteer had used the console to change the Search button on our Russian homepage to say "Click here bitch." Another volunteer changed Google Malta's Search button to a traditional Maltese insult, "penis in a can."
The word, according to engineering, was that a few bugs remained in the system.
Yet, once again, risk reaped rewards. The willingness to suffer a few quickly eradicated indignities opened up enormous gates to international audience growth. The world tolerated awkward translations and the occasional insult in order to access Google's search technology. It was a reminder that perfecting the polish was not as important as giving people access to the product behind it. The results we returned and the speed with which we returned them were ultimately all that mattered. They were the essence of Google's brand.
Can You CPC Me Now?
Unfortunately, the results we were getting for another part of Google were not proving satisfactory. Our AdWords system kept growing, but we were concerned (and this time it wasn't just me) about the rapid expansion of our competitor GoTo. Their "search results" were actually ads, sold by online auction, with the top listing going to the highest bidder. These ads were distributed across the Internet, and each time someone clicked on one the advertiser paid GoTo, who gave a small percentage to the site on which the ad appeared—the pricing model known as "cost per click" or CPC. It was an innovative approach to monetizing se
arch. In the Googleplex, most of us thought it was a load of crap. GoTo required no algorithms to determine relevance, and it presented paid ads as "objective" search results.
GoTo's rationale was simple: the more someone paid to have an ad show up on searches for a particular keyword, the more relevant that ad was likely to be. Where we had a democracy of the web, GoTo had a dictatorship of the dollar. I found GoTo's auction-based results almost unusable. A significant number of their advertisers bid high for popular, but irrelevant, search terms just to lock in the top position on as many searches as possible. Even when the top result was not pure spam, the whole approach seemed misleading.
Ralph Nader agreed. Nader's group Commercial Alert filed a deceptive-advertising complaint with the Federal Trade Commission (FTC) in July 2001 to stop the practice of "inserting advertisements in search engine results without clear and conspicuous disclosure that the ads are ads [which] may mislead search engine users to believe that search results are based on relevancy alone, not marketing ploys."* The complaint called out eight search companies, including AltaVista, AOL, iWon, and Microsoft. Google was not listed among the offenders, but news articles grouped us with companies that had no scruples about crossing the line. That bothered Larry and Sergey a lot.
Larry had thought he was done with ads after AdWords launched in September 2000. Ads were the price we had to pay for building a really cool search engine, but he viewed them as "tainted meat," according to an engineer who was in the ads group at the time. We had made ads better, but GoTo proved we hadn't completely solved the problem. Fortunately for Google, the ads engineers had not been content to leave AdWords alone. They continued to innovate because they saw the danger presented by CPC ads as life-threatening for our company. Fortunately for those engineers, Salar saw it that way too.
Chapter 19
The Sell of a New Machine
SALAR FELT HEMMED in. It was early 2001 and he was not getting his way on consumer products. As Larry's first PM, he had helped launch Google news and been involved in the rebirth of Deja.com as Google groups. But there were always too many cooks stirring those pots.
"Everybody had strong opinions about everything," he remembers, "because the consumer product was what we all lived and breathed." I understood that perspective entirely. I kept bumping into concerned parties who wanted to rewrite my copy. Salar decided it would be nice to find some area that fewer people cared about. He settled on ads and began informally working with the ads engineers.
"Larry and Sergey had the strongest views about things on the consumer side," Salar realized. "We all knew that they were less interested in the details on the ads side." Omid, on the other hand, had a deep and abiding interest in ads. As head of sales and business development, he liked having Salar involved and suggested Salar make the arrangement official. Larry agreed and named Salar product manger for ads.
Salar did his best thinking late at night. He walked the bike-lined halls of the Googleplex after dark, thinking about how ads were sold, how they were displayed, and how they could be improved. In the cubicles around him, Matt Cutts and a handful of other engineers* worked on maintaining the AdWords system.
Another team focused on ads optimization—a new system to predict which ads users were most likely to click. Predicting user behavior was an enormous technical challenge that required machines to learn in real time and then make educated guesses. Veterans Chad Lester, Ed Karrels, and Howard Gobioff were on that team,† along with Noogler Eric Bauer, whose initial project had added a hundred thousand dollars a day to Google's revenue stream by replacing low-performing ads from the original system with the best-performing AdWords ads. Their leader was a redheaded Canadian by the name of Eric Veach, who had come to Google after making fur and smoke look more realistic in movies like Monsters, Inc. Eric enjoyed a challenge.
Eric and Salar bounced ideas back and forth from parallel tracks, though their common destination became clear soon enough. Google would need to build an entirely new ad system to replace AdWords. A new system based on cost per click (CPC) instead of cost per thousand impressions (CPM). Remember that with CPM pricing, an advertiser paid a set amount for each thousand times an ad was shown, regardless of how many people actually clicked on it. AdWords offered three CPMs, at ten, twelve, and fifteen dollars. The more you paid, the higher up on the page your ad appeared, which made your ad more likely to be seen and clicked by users. With CPC, an advertiser paid only when someone actually clicked on the ad, regardless of how many times it was displayed.
Advertisers loved CPC, but it scared the bejeezus out of Google executives. Netscape had offered CPC deals, guaranteeing the number of clicks their clients would get over a certain period of time. The numbers had been very large. When the clicks didn't materialize, Netscape had no choice but to keep running more and more ads. And the more ads they ran, the lower the clickthrough rate went, until every page was saturated with banners that were ignored by users. CPC could spawn a whirling, sucking spiral of death.
Salar, though, had seen a great future. He summed it up in one word: "syndication." Salar believed that to truly grow Google's revenue we needed to distribute ads on other websites. Advertisers would be unwilling to pay CPM rates to have their ads displayed across a network of sites over which they had little control. They would be much more comfortable if they paid only when people actually clicked on their ads. But if we were going to do CPC, we had to do it differently than GoTo. We had to do it better. The question was how.
The engineering director overseeing the ads team in 2001 was Ross Koningstein. Ross had a PhD in aerospace robotics and notions of his own about how the ads system should develop. He thought we could sell ads CPC, but rank them by CPM. By taking data from the logs files, we could calculate how much revenue each ad actually generated for Google, and display them accordingly. Ross told me that he presented his idea to Larry at a meeting in April 2001, but that Larry was not in the mood to hear it.
Larry only wanted incremental changes to AdWords, according to Ross, not a whole new system, even though the other engineers agreed a new system was needed and were already working to make it so. Besides, Ross said, Larry only wanted to communicate with engineers directly involved in writing the code—and with his handpicked PMs. At the meeting, Larry informed Ross that he was making Salar the lead for ads. Salar immersed himself in discussions and brainstorming with the engineering team. They shot down most of his ideas, such as selling keywords in bundles, until one night Salar experienced an epiphany. Google could assign a quality score to each ad. "Quality" would be our prediction of how likely a user was to click on an ad. If that score was factored with the amount the advertiser was willing to pay, we could rank ads by their potential for earning money for Google—their effective CPM. Everyone would benefit. The user would see more relevant ads. The advertiser would get more clicks. Google would make more revenue. It was a brilliant idea.
Ross believed he had already suggested something very similar, only to be ignored. According to Eric Veach, a number of people were arriving at the same place independently. "It's kind of an obvious idea," he claimed, though some of the things evident to Eric made my brain ache.* AdWords already prioritized the ads by CPM. If you paid fifteen dollars, your ad appeared at the top. If you paid less, your ad was lower on the page.
"I know I proposed that at one point," Eric told me about effective CPM ranking. "I said, 'We can prove that it results in the highest expected income if you rank people by their effective CPM,' which essentially means the CPC you charge times the predictive clickthrough rate."
That one idea was worth billions and billions of dollars. So who came up with it? "I think you're going to find very little agreement on that," Eric went on. "It's true that each of us who was involved thinks we contributed. The sum of those numbers adds up to much more than one hundred percent."
But brilliant as the idea was, it wasn't perfect. Two things were still needed. The first was a method for determining the "quali
ty" rating for each ad.
"When Salar was originally talking about multiplying by clickthrough rate," Eric explained, "what he meant was the historical clickthrough rate. If we had shown their ad a thousand times and somebody had clicked on it ten times, they would have had a one percent clickthrough rate." Eric thought that was the wrong approach. "The problem is, that kind of stuff is just, well, garbage really. You have to show a bad ad thousands and thousands and thousands of times to get any good information about how well it's doing."
"The clickthrough rate needed to be a predictive thing," Eric insisted. "It needed to be what we thought the chances were of somebody clicking on the ad given all the information we had about the query right then." That required enormous computing sophistication. And secrecy. Eric didn't want advertisers to know how Google was determining the quality of their ads, so that Google could keep refining and improving the algorithms without advertisers' trying to game the system. He insisted that Google retain complete control—that the ranking mechanism remain a black box. He knew that would frustrate and anger advertisers, but it would benefit users, who would see more relevant ads on every page.
While our competitors made trivial adjustments to their ad programs, Eric led an effort to build one of the biggest machine-learning systems in the world—just to improve ad targeting.
I'm Feeling Lucky: The Confessions of Google Employee Number 59 Page 31