One thing that did seem to work was a management trick suggested by John Doerr, not long after the VC made his original $12.5 million investment in Google and joined its board. Doerr was a fan of a complicated system called Objectives and Key Results, usually referred to by the acronym OKR. It was something Andy Grove had devised at Intel (he’d called it Management by Objective), but Doerr believed it was even more useful for start-ups. “It’s really important in rapidly growing companies because it allows you to be superclear about what priorities are,” he says. His efforts to start OKRs at previous start-ups had met with mixed results, so he had no idea what reaction Larry and Sergey would have. But they were enthusiastic enough to have Doerr come and present it to the company.
So one day in 1999, Doerr took Googlers into a conference room and did a PowerPoint presentation on how OKRs worked. The idea was not just to identify what one wants to do but to break down the task into measurable bites (“key results”). In his book High Output Management, Grove imagined the OKR system applied to Christopher Columbus. The explorer fell short of his objective of finding a trade route to India, but he did carry out some subsidiary OKRs: he gathered a crew; he bought supplies; he avoided pirates; and by discovering the New World, he brought riches to Spain.
Doerr had Google at metrics. “Google did more than adopt it,” says Doerr. “They embraced it.”
OKRs became an essential component of Google culture. Every employee had to set, and then get approval for, quarterly OKRs and annual OKRs. There were OKRs at the team level, the department level, and even the company level. (Those last were used sparingly, for important initiatives or to address gaping failures.) Four times a year, everything stopped at Google for divisionwide meetings to assess OKR progress.
An outsider might have wondered if this were a sign of Dilbertization at Google, an annoying program that diverted energy from real work. But Googlers didn’t seem to think so. They saw the OKRs as data, a means of putting a number on the traditionally gooey means of assessing performance. It was essential that OKRs be measurable. An employee didn’t say, “I will make Gmail a success” but, “I will launch Gmail in September and have a million users by November.” “It’s not a key result unless it has a number,” says Marissa Mayer. The OKR embodied ambition. “It sanctions the ability to take risks,” says Doerr. Even worse than failing to make an OKR was exceeding the standard by a large measure; it implied that an employee had sandbagged it, played it safe, thought small. Google had no place for an audacity-challenged person whose grasp exceeded his reach.
The sweet spot was making about .7 or .8 of your OKR. (Geekily enough, the metric was measured by a decimal representation of how close an employee came to the OKR, with the integer 1 being an exact hit.) At the end of every quarter, employees set their OKRs for the next quarter, and six weeks later, they saw their managers and gave a progress report, using a traffic-light system for grading. “Green light, I’m good to go on that one. Red light, I’m having serious issues. Yellow, possible danger,” says McCaffrey. Toward the end of the quarter, all the OKRs were graded, and if an employee was hitting 100 percent, he or she needed something else to do.
What’s more, OKRs were not private benchmarks shared only with managers. They were public knowledge, as much a part of an employee’s Google identity as the job description. The OKRs appeared on every employee’s biographical information on MOMA, Google’s internal website. (The name didn’t stand for anything in particular—according to Marissa Mayer, Larry Page just wanted something fast and short and easy to type.) You could even see Larry and Sergey’s OKRs. “We needed to run our company somehow, and I think having an organizing principle makes sense,” says Brin. “We really like transparency and like the idea that we communicate to everybody on roughly one or two pages of paper every quarter what we want to accomplish.”
That kind of sharing was another hedge against the creeping impersonality endemic to big companies. At a start-up, everybody knows all their colleagues and what they are working on. Even as Google grew to more than 20,000 employees, it tried to maintain the ability to keep up with everyone else. In addition to MOMA, Googlers could access Project Database (PDB, as it was referred to at the Googleplex) to follow all the things the company was up to—engineering allocations, product manager allocations, product definitions, engineering documents, and specs. Also, Googlers looking for a cool new project could access a section called simply “Ideas,” where their colleagues pitched promising concepts that needed manpower.
The internal transparency was especially startling because Google had a phobia about leaks that rivaled that of the Nixon White House. The company was an information lobster, hard-shelled on the outside but soft and accessible on the inside. Sometimes employees didn’t get the distinction, as was the case with Mark Jen, a twenty-two-year-old Noogler who started a blog in 2005 called “ninetyninezeros” about his experiences; among the items that apparently displeased his bosses was a comparison of his salary and benefits to those at his previous employer (Microsoft), which paid more. He also noted that Google’s business was booming; even without his mentioning numbers, that was interpreted as data best withheld from competitors. Jen was fired before he completed a month at Google.
Google’s OKR system was only one of many processes, many imposed by Schmidt, intended to bring a sense of order to a company growing to 20,000 employees. “Google’s objective is to be the systematic innovator of scale. Innovator means new stuff. And scale means big, systematic ways of looking at things done in a way that’s reproducible,” Schmidt says. So Google spent a lot of effort on actual bureaucracy—a regular set of launch meetings and reviews, weekly meetings of the Operating Committee of the top leaders, global product strategy meetings, and a companywide peer review system that consumed enormous time.
“We try not to expose all those things,” Schmidt would joke about the organizational scaffolding, “because we want it to look chaotic.”
Perhaps the best illustration of Google’s creative denial of its Brobdingnagian size was a startling move that its founders made in 2007. For a number of years, Brin and Page drew organizational and clerical support from a pool of four sharp young women known as LSA, or Larry and Sergey Assistants. (Googlers referred to LSA as if it were a single organization. You would say, “I’ll check with LSA to see if Sergey can come to this meeting.”) The system seemed to work well, but Brin and Page felt constrained. By having assistants, they noticed, it was easier for people to ask things of them. “Most people aren’t willing to ask me if they want to meet with me,” says Page. “They’re happy to ask an assistant.” When a meeting request came, an LSA would have to see if Page or Brin actually wanted to do it. In truth, the founders almost never wanted to do it. So one day, Brin and Page abruptly dissolved LSA. They would thereafter have no assistants. Whatever they felt was important at the moment would be their work. Sergey sometimes liked to move his workplace right in the middle of a project he found interesting. And sometimes he or Larry would just take off somewhere. Even the communications people would have no idea where they were.
On one hand, the shift offloaded a lot of menial work to the assistants of other executives. Sergey wasn’t about to spend his time mailing packages, so one of Eric’s assistants wound up doing it. Using Google’s calendar application, which allowed people to share their schedules, certain Google executives and their helpers would make appointments for the founders. Larry and Sergey would regularly attend the key weekly meetings—the Operating Committee on Monday, the global product strategy sessions on Monday and Tuesday, product reviews at the end of the week, and TGIF at 4:30 on Friday. You could expect them to be present when the board convened. But they wanted to keep much of the week open. “My favorite meeting,” says Page, “is the absence of meetings.”
It went without saying that Page’s least favorite meetings were one-on-one press interviews. “Larry can be a very, very sensitive and good person,” says one former Google PR hand, “b
ut he has major trust issues and few social graces. Sergey has social graces, but he doesn’t trust people who he thinks don’t approach his level of intelligence.”
Googlers learned to adapt to this system. If someone needed the founders’ approval for a purchase or project, the accepted strategy was stalking. Like the network of amateurs who sit with binoculars at airports and track the peregrinations of private planes, an informal Google pipeline delivered a steady stream of Larry or Sergey sightings. Canny Googlers hoarded knowledge of key interception points. “If I want an opportunity to meet with them, my best option is to go to Building 43 and just plop myself on a couch somewhere,” says George Salah. An APM named Jini Kim once got a key approval from Larry by gathering intelligence on his movements and loitering in his expected trajectory. Googlers also knew that there was an elusive window of access in the few minutes following a TGIF. But sometimes you could carefully plot a collision course with a founder and be frustrated when one of them was engaged in deep conversation with the Mexican mogul Carlos Slim or some other visiting dignitary. Or they would simply be in a hurry. Page in particular was a master of the drive-by greeting, flashing a wide, happy-to-see-you smile while slightly picking up his pace, so that if you attempted to follow up with anything more than a quick hello, you found yourself talking at his receding back. “Larry got rid of his assistants so that he would never meet with anyone who couldn’t figure out how to get a meeting with him,” says product manager Wesley Chan. “If you wanted a meeting with him, you had to find out where he was and harass him.”
Larry’s and Sergey’s peripatetic ways could drive Googlers crazy. Even Eric Schmidt sometimes viewed them acerbically: “Larry will call and say, ‘I’m going to go visit Android,’” he says, referring to Google’s mobile phone project. “He’s not going over there to inspect—he’s going over there to have fun.” But Maria Montessori might approve. “To be … helpful,” she wrote, “it is necessary rigorously to avoid the arrest of spontaneous movements and the imposition of arbitrary tasks.”
PART FOUR
GOOGLE’S CLOUD
Building data Centers that Hold everything ever Written
1
“Hi, here I am, and I’ve got to explain this thing that seems creepy and weird, and convince you it’s not so bad.”
Paul Buchheit looked like a fourteen-year-old when he joined Google in 1999, his cherubic face crowned with wisps of blond hair. He had grown up just outside Rochester, New York, a typical hacker kid driven by silicon and curiosity, and by the time he entered Case Western Reserve University in Cleveland, he was full of ideas and projects, one of them being a web-based email program. After graduation, he took a job with Intel. But the chipmaker was big and bureaucratic. “I wasn’t really loving Intel,” he later said, meaning he hated it. He started looking around for an interesting start-up. He’d read about Google on Slashdot, an online discussion site that was like Entertainment Tonight for geeks, and started using its search engine. He was impressed by Google’s imaginative stab at an interesting technical problem. He sent his résumé to
[email protected]. The email bounced—the server was down—but he persisted and eventually was granted a phone screener, followed by a face-to-face interview.
Unlike at the other companies he was talking to, the Google people asked smart, technical questions. The one he remembered was “If you have a server that’s performing too slowly, how do you diagnose the problem?” On one hand, it was straightforward, but when you thought about it, the question was almost spiritual in its depth. “Surprisingly, it’s something that people don’t really get,” says Buchheit, years later still engaged by the profundities. “Your site is slow, what do you do? What resource is being restrained? Because it’s always a bottleneck. Is it CPU-bound? Is it disk-bound? You have to understand the fundamentals of what makes things fast or slow. That’s a pretty good question.”
Google made him an offer. He didn’t have much confidence in the company’s future—“I thought they would probably get crushed by AltaVista or something,” he says—but he viewed it as a learning experience. Even if the company didn’t last long, he’d learn about start-ups. He was employee number 23.
In mid-2001, Buchheit found himself liberated from his assignment by the management putsch Page and Brin launched that year, doing away with all the product managers. So he unilaterally decided to revisit his college project and create a web-based email product for Google. He would have the benefit of something that was only just invented when he was at Case Western: the dynamic JavaScript computing language, which could, if you pushed it, help you create a web-based application that behaved like a desktop application running directly on the computer. (That would make it more responsive and flexible than the current web-based mail systems run by Microsoft and Yahoo) “He was writing with the technology that would later be called Ajax, but that term didn’t even exist then,” says Keith Coleman, who would later head the email product at Google. “Internally, there were a lot of questions about whether this would work at all for a Google project.” Within months, Ajax would become a core technology in thousands of web-based computer applications, and Buchheit would look like a prophet.
But what would really distinguish Buchheit’s email product from its competitors was storage. He wanted a lot of it. One of the frustrations of people who used email was the constant need to clear out the cramped digital inboxes and archives, products of an era when email use was sparse and storage costs were sky-high, two factors that no longer applied. Of the existing web-based systems, Microsoft’s Hotmail offered 2 megabytes of storage and Yahoo only 4. For people with even moderate needs, those mailboxes would fill up in a few days. Almost as soon as you received an email, you had to consider whether to delete it. Weirdly, this problem also plagued people who worked for corporations. Even though storage was increasingly inexpensive, the information technology (IT) people in charge of the corporate systems policed disk space as if it were made of platinum. They would commonly impose impossibly low ceilings on the amount of disk storage allotted to a given account, and if you exceeded the limit, you were unable to access your mail until you weeded nonessential messages out of your inbox.
Buchheit wanted to eliminate that problem. “It was fundamental to the way the product actually worked that I need to be able to save my mail. Otherwise, it’s a different product. And then, the other thing is that I thought it would be cool.” His mailboxes would have a gigabyte of storage—more than a hundred times what competitors offered. For free. This was enough for more than 10,000 emails. And that would be only the start. As storage got even cheaper (an inevitability in the age of Moore’s Law), Google would offer more and more space.
An email product was a departure for Google, which to date had concentrated on search. Analysts and business writers continually praised the company for “sticking to its knitting,” a practice obviously regarded as a virtue. Anyway, by what logic could the company’s mission include offering email service? Even Buchheit’s colleagues thought that making a web-based email product was too problematic. They thought that implementing the scheme would be expensive and complicated and were especially doubtful that the JavaScript approach would work well. “Almost everyone thought it was a bad idea,” says Buchheit. “Except Larry and Sergey.”
Indeed, Page and Brin were thrilled at the prospect. They saw email as a search problem. Your email, after all, was a crucial information corpus. How ridiculous was it that you could find the most obscure item among billions of web pages but it was a tortuous—or impossible—process to dig up an interesting comment you had made a few weeks ago or relocate a book recommendation someone sent for you last year? The most popular desktop email application—Microsoft’s Outlook—had a search function so slow and cumbersome that no one used it. So what company better than Google to allow you to easily access your information? The founders wanted to use the system Buchheit proposed for themselves, and volunteered to test it. Soon Buchheit and his team (including his offi
ce mate Sanjeev Singh, who became a colead on the project) had a prototype, dubbed Caribou. The name was an homage to “Project Caribou,” a doomed initiative in a Dilbert cartoon. “Larry and Sergey literally became our first users,” says Buchheit. “It was not only key to the product surviving but also becoming a good product.”
The first suggestion from the founders was rather obvious. “All the first version did was store and search email—Larry and Sergey said it would be nice to be able to reply to emails,” says Buchheit. “I said, ‘Okay, I guess I can add that.’”
From the start Buchheit wanted revenues. The product could have ads, the same kind you saw on Google search results pages. Instead of relying on the relevance of keywords, ads in Caribou could relate to something you talked about in your email. “People always asked how things could make money, and putting in ads was the obvious thing to do, because that’s how Google made money,” says Buchheit. “And if we were going to do it, we should do it from the start, so that people wouldn’t feel tricked.”
Remember, it was Buchheit who originally came up with the slogan “Don’t be evil.”
A lot of people at Google hated the idea of ads, saying that users would be repelled at seeing ads alongside their emails, especially ads that related to the content of the messages. Opponents also claimed that it was too hard to match ads to the content of emails—it wasn’t like search, where people started with keywords. Marissa Mayer, who was the product manager at that point, was particularly opposed. Mayer told Buchheit, who shared an office with her at the time, that targeting ads to email “is just going to be creepy and weird,” warning him that people would mistakenly assume that Google had a bunch of drones in some room who were reading your email and matching ads to the private expressions contained therein. Buchheit was on the fence, but his colleague Sanjeev made some back-of-the-envelope calculations and concluded that even a small percentage of people clicking on the ads would produce more than enough revenue to pay for Gmail.