In The Plex

by Steve Levy

  Google finally got its license in June 2007. The dispute had been resolved in secret. And to a large degree, the level of service stabilized. Another boost that year was that Google was granted a valuable concession: simply typing “g.cn” would take Chinese users to the Google.cn site. But by then, many Chinese had written off Google as an unwelcome outsider with less reliable service. In the summer of 2007, a group of young associate product managers from the United States spent an afternoon interviewing Chinese consumers. One woman in a T-shirt that read brasil soccer seemed surprised that she’d even be asked what search engine she uses. “Baidu.” Why? “Because it is the product of Chinese people who naturally know more about China than Google,” she said. Though she conceded that some people with advanced education and familiarity with English may want to use Google, “Most Chinese don’t speak English. They will never use Google.” A young man told them, “Google needs to be more close to the Chinese people.” And another young woman said that she likes Google but doesn’t use it much because it often stops working. Did she know why? “A broken cable under the ocean,” she said.

  When Google set up in a country, it typically would have two leaders, one to head the engineering operations and the other to run the business side. In China, Google’s business head was Johnny Chou. Officially, he was the equal of Kai-Fu Lee, but Lee’s celebrity and exalted status as the man who had started Microsoft’s China Research Lab far outweighed Chou’s reputation. Ultimately, Google China was not big enough for both of them. Chou’s replacement was an even-tempered executive named John Liu, who formerly had headed the China operations of SK Telecom (Korea’s dominant mobile network). Liu was content to let the more celebrated man take the spotlight. “I respected Kai-Fu as the in-country leader,” he says. “I think we needed one. I’m not an ego guy.” In any case, Lee told Liu that he wasn’t a businessman and was perfectly happy for Liu to act on his own in generating sales and revenue in the Chinese marketplace.

  Liu found that though Baidu’s nationalistic sales pitch worked well with consumers, advertisers wanted results and Google—with its superior AdWords technology—could offer that. Chinese businesses also liked the idea of a competitor to Google. “Baidu did a great job of building up local Chinese search, but the Chinese will never accept that there’s only one search engine,” he says. “They want Google to be here, they want Google successful.”

  But Liu believed that to compete in China, Google had to try harder to get its message across. “A lot of normal users in third- and fourth-tier cities don’t really know Google,” says Liu. It depressed the Google China people to see Baidu’s name everywhere. When you used an ATM machine, the log-in screen often showed a Baidu ad. When you ate at a KFC restaurant, the paper on your tray would have the Baidu logo on it. Google China’s leaders—including Kai-Fu Lee—wanted the sort of aggressive marketing that Google never had done in the United States. But the people in Mountain View, perhaps because they were still conflicted about the entire China effort, would not give those efforts full blessing.

  Early on, Google’s marketing team spent six months on a big media campaign, including print, radio, and television. They’d hired Ogilvy & Mather to coordinate and shoot television commercials with real Chinese Google users. In one, a teenage boy couldn’t find the Nike sneakers he wanted anywhere—until he used Google. Google made six variations of those success stories. But at the last minute, the bosses at Mountain View pulled the plug. This was disheartening to the China Googlers, who felt that the company should cover the third- and fourth-tier cities to tell people of the existence of Google. But that wasn’t the Google way.

  Google was already at a disadvantage because its ethics prevented the company from practices that were common in China but too unsavory to pass the “evil test” applied in Mountain View. Some of these were as simple as paying expense money (known as “red pockets,” typically fees that exceeded cab fare) to reporters attending press conferences. Google angered the local press by not paying.

  More complicated were fees paid to managers of Internet cafés. A substantial percentage of Chinese users accessed the net in these basement operations, smoky parlors that looked like a cross between a telemarketing boiler room and a video poker casino, with hundreds of terminals active at any hour. The large companies that franchised these establishments preloaded the computers with their chosen software, and Google and Baidu paid for the privilege of being the default search engine. But often the managers of individual cafés would take money under the table to replace one search engine with another. Google generally avoided such arrangements. But the company sometimes used representatives who weren’t as finicky and looked the other way.

  Navigating these precarious situations was especially tricky for Google because it presented itself in China as a righteous force, a trustworthy avatar of the digital age. “In my mind there’s a new kind of China—the WTO and the Olympics, where there’s a rising middle class that expects justice and expects things to work,” said Sunny Oh, who became Google’s director of marketing in Beijing. “We represent this new China, a trustworthy institution that’s not going to fiddle with the results because someone’s paying us to insert search results or suppress results because there’s bad publicity.” (Apparently this was a distinction from suppressing results to accommodate government censors.)

  Google had a chance to trumpet its incorruptible search standard when its competitor Baidu was embarrassed by a commercial arrangement in its own search results. In 2008, the Chinese company Sanlu Group sold baby formula containing melamine, which caused kidney stones to form in hundreds of thousands of babies, killing six of them. Chinese news agencies reported that Sanlu had paid millions of renminbi to Baidu for its program called PR [page rank] Services to remove news articles about the scandal from search results. Baidu denied the story, but on September 12, a reporter for 21st Century Economic Report found that Google delivered 11,400 search results for the incident, while Baidu had 11. Even CCTV, the government-controlled television network, reported the story, thus reprimanding Baidu.

  “We saw this problem as a great opportunity to drive our core values,” says Sunny Oh. “It was a chance to say, ‘Why does integrity matter?’”

  But Baidu had its revenge. A couple of months after the Sanlu flap, Robin Li undoubtedly enjoyed Google’s discomfort when CCTV criticized Google for serving search ads for unlicensed medical products for keywords such as “diabetes.” That year, Baidu won the sponsorship of one of the hugest events in the country’s television schedule, the CCTV Chinese New Year gala, watched by more than 400 million people.

  It was as if the government was sending Google a message: you can be in our market, but you must not be the leader. As Google’s market share inched into the twenties and approached 30 percent, some Google executives believed the number was hitting an artificial ceiling: the government would never allow Google to accumulate more than a 35 percent share.

  Meanwhile, Kai-Fu Lee and his directors were organizing their smart young engineers to do great work. It took a while for some of the Chinese to adjust to the Google style. For instance, many had difficulty with the concept of 20 percent time—despite assurances, they did not feel comfortable pursuing a part-time project on their own initiative. At one point a visiting Mountain View executive called an all-hands meeting and asked for all the managers to leave the room. When only the engineers remained, he emphasized that they did not need permission to do a 20 percent project. Even that wasn’t sufficient, so Lee set up brainstorming sessions where people could talk freely about cool ideas and then vote on the best. “It gives you more confidence if your idea was voted number one in a group of ten,” says Lee. Another impetus for engineers to pursue 20 percent projects was to pair them, on the theory that having a partner would build confidence.

  Because Google had a firm policy against storing personal data inside China—to avoid the problems of having the government demand that Google turn over the data—it did no
t offer a number of its key services for local Chinese users. No Gmail. No Blogger. No Picasa. Other services had to be drastically altered. YouTube was blocked entirely.

  In 2007, Kai-Fu Lee assigned Mark Li to head the Google Maps team. At the time Maps was judged the worst product in Google China. In large part this was due to the government restrictions Google labored under. It had no license to gather geographical data and had to buy information from other companies. Li began working closely with the government to get various functions approved. Google also shared its information and technology with other companies. For instance, one of Google’s partners had a service that identified the best restaurants in an area but did not have a license to show the restaurants on a map. Under the arrangement, Google could pinpoint the restaurants on its maps—both helping the partner’s business and making Google Maps more useful. As more information came online, Google began to draw more users.

  The breakthrough for Google Maps came during the 2008 Chinese New Year, when a huge, unexpected snowstorm hit. Millions of people were stuck in their home provinces; more than 100,000 people were stranded at the Guanxou train station alone. A group of seven or eight Googlers who regularly ate dinner together brainstormed on how they could help and returned to the office to start on a project inspired by the fire maps around San Diego during forest fire season. By the end of the next day, the Google team published a detailed interactive snowstorm map that aggregated information from dozens of different sources—things like news, weather reports, airport closures, and road status. It was wildly popular, and Google did a variation on other holidays. When a major earthquake hit China, the Googlers combined the system with Google Earth to bring in satellite images. Google provided the Chinese government information it had not gathered on its own. The government actually presented Google with an award for its efforts. By 2009, Google was the market leader in maps.

  But arguably the most important project at Google was the Pinyin Input Method Editor (IME), a system that sped up and streamlined the often awkward task of producing Chinese-language ideographs on a computer keyboard. (Pinyin is a phonetic system that generates Chinese characters from Latin alphabet input.) Google’s system took smart guesses from minimal keystrokes and suggested which characters the user might want to use. It was able to make these predictions by applying user behavior data it had collected from its search engine. As people began to use its IME system more, Google would get even better data (by noting which suggestions the user accepted and rejected), and the system would presumably get even better. Thus Google’s system had the chance, when it debuted on April 10, 2007, to become a huge asset in the company’s crusade to win market share in China. “We were really proud of this product,” says Yonggang Wang, who headed the project. But the launch was one of Google China’s worst disasters.

  Google had invited the Chinese business press to attend a roundtable where the engineering director and product manager would announce the product. During the presentation, Jin Cui, the Google PR rep, got a call to go to Kai-Fu Lee’s office. Apparently, some of the Chinese bulletin boards were reporting that Google’s IME was based on intellectual property stolen from a local company. Google’s engineers discovered that there was indeed a problem with rogue information inside the IME product but weren’t sure how widespread it was. Jin Cui returned to the meeting, pretending that all was fine, and the first day’s reports on Google’s new product were positive. But over the next few days, the bulletin board reports accusing Google of theft only accelerated.

  After a couple of days, Google found the problem. One of its student interns had been working on the IME product, and during the testing process, he had taken a shortcut. Instead of using original data, he had gone to a competing search engine called Sogou (which translated to Search Dog) and drawn from Sogou’s search results, which in turn had drawn from an internal dictionary that had originally come from its parent company, the Internet portal Sohu.

  “He was an intern, and he just wanted to find a work-around to speed up the process,” says Wang. “So he borrowed the data to provide a work-around way. It was a bad thing.” Wang explained that such acts are not uncommon in China, where there’s a “more fluid” view of plagiarism. Google’s routine new-employee training did address such issues, but in this case it had apparently failed to make an impression on the intern. (“He went to Microsoft,” says Wang.)

  People at Google were convinced that the early discovery hadn’t happened by accident but had been planted on the bulletin boards by competitors, who had come across the problem by testing Google’s IME as soon as they got their hands on it. Nonetheless, Google had screwed up, and before it formed its response, Sohu was attacking it in the first of what would be a series of almost daily press conferences over the next two weeks. “It was a smart decision on Sohu’s part,” says Cui, “because they had a competing product and they could say, ‘Our product is so good that the mighty Google is stealing from us.’”

  All Google could do was admit its error and apologize. “The Chinese journalists really believe Google values, so they cannot believe Google would do this kind of thing,” Cui says. “All they wanted to know was, is it true?” And it was. The company that had hoped to instruct China on moral business practices had stolen code.

  The number one concern of Google’s engineers was not unfair competition or the pressures of censorship. It was an issue that spoke to the place Chinese employees—and the Chinese business itself—occupied in Mountain View’s estimation. The situation infuriated and humiliated the Chinese engineers every day of their working life. It was their access, or lack of it, to Google’s production code.

  Google was a collaborative company that wanted its engineers around the world to innovate on its existing products and create exciting new ones. It empowered them to do so by giving them access to its production code base. Without such access, engineers were limited in what they could do. But unlike Google’s employees in other locations—Zurich, Tokyo, Tel Aviv, Bangalore, or even Moscow—the China workers did not have such access. Convoluted procedures were required to work on search, ads, and other key projects. The restrictions limited what the engineers could do—and sent a message that they were second-class employees, that Google did not trust them. “China was the only country that had this,” says Boon-Lock Yeo, who headed engineering in the office Google had opened in Shanghai. “It was this fear that something bad was going to happen—somebody breaking into the data center or someone will take out certain information that would be considered very sensitive.”

  “It’s not that I don’t trust Chinese engineers,” Alan Eustace, the Google executive assigned to monitor the China territory, later explained. “It’s the same engineers as here, who went to the same schools, but when you go to a place like China, there’s lots of examples of companies where intellectual property has gone out that door.”

  “We were concerned that employees in China who were Chinese nationals might be asked by government officials to disclose personal information, and all our access policies derived from that,” says Bill Coughran, Google’s engineering director, who enforced the policy. Despite those reasonable concerns, suspicion lingered in quarters of Google China that the engineering executives behind the policy—some of whom had deep concerns about the company’s China policy—had intentionally engineered rigid restrictions as a form of corporate civil disobedience against their employer’s cooperation with censors.

  “Productivity was impacted,” admits Yeo, who tried for years to overcome the problem. “It took longer than people anticipated to solve it.” In the meantime, Google relied on sea-turtle engineers who had come to China to check the code of local engineers. “That was a real pain, because then they’d spend their time not thinking, but just checking people,” says Yeo. Local engineers were urged to pick projects that didn’t involve the global code base. Or they would just be told to make searches and look for unsuccessful queries—basically, performing the tasks that less qualifi
ed testers do for Google in other countries. “Go search every day, search for lots of things, tell us what’s broken,” says Kai-Fu Lee. “That doesn’t require access to code.” But it also put those engineers in a position where any creative genius they had was wasted. And that infuriated them.

  “There’s a lot of pain here,” Ben Luk says. “People feel they are being treated like second-class citizens. You can feel the pain gnawing at you. At one time, after we had a hundred engineers in the office, I had the feeling that if we didn’t give them access, there would be a riot.”

  Beginning in 2008, Boon-Lock Yeo spent more than half his time on the problem, working with a security engineer in Mountain View. Eventually Google was able to implement a system where all but the most sensitive portions of the code base were available to the Google China engineers. But the anger remained. In September 2009, Luk told a visiting journalist that although things were better than they had been, “there’s no clear access policy.” He cited an example where a Chinese engineer “opens a ticket” to access a protected database and the request gets stuck in the queue for months.

  Still, when that conversation occurred in the fall of 2009, Google China was feeling pretty good about security.

  4

  “The worst moment in our company.”

  Google’s success in China depended in part on having a government relations (GR) point person who could navigate the tricky shoals of preserving Google’s values without offending Chinese officials. Google’s first GR head was a former vice president of Sina, who was experienced in the ways of Chinese bureaucracies. But perhaps because she did not speak English, she failed to appreciate issues from the Google perspective. She complained to at least one colleague that Google wasn’t flexible enough with the government and did not work hard enough to please it.