When Karen took a vacation, we ordered a thousand plastic playground balls and filled her cube with them. They were still being thrown from office to office and rolling around under desks a year later.

Composing a list of new CRM vendors didn't take long. Fewer than half a dozen major players offered stable, well-tested systems. Google's tech evaluation team would ensure we weren't sold a bill of goods (though they hadn't kept us from choosing Miasma), and Larry had a college friend who would advise us on desirable features. The friend, David Jeske, counseled us on what to ask for, then added that, by the way, he and a buddy were building a CRM product called Trakken—if we were interested. It wasn't really finished yet, but Larry's other Stanford pals at Wunderground.com were using it.

Interested? Interested in an untested CRM product still in development with one tiny client? Created by a company of two people? Sure, that's just what I was looking for—another risky technology with no support and no track record behind it. I thanked David for his help and, because he was a friend of Larry's, assured him we'd be happy to send him our request for proposal.

Meanwhile, our real search was well under way. One vendor couldn't provide any support for non-English email. Another had a terrible UI because it was a first-generation product. A third seemed overpriced and their salesman's aggressive stance made us wary of doing business with them. Only one company offered a reasonable solution, and we began negotiating with them in earnest. With our leading contender scheduled to make a presentation to our finance, operations, and sales departments, I felt confident I could convince Larry and Sergey to loosen the purse strings and do it right this time: spend money for a high-quality, stable system from a respected vendor.

I hoped Larry's friend had taken the hint and forgotten about us. It would be a frosty day in Hades before we'd make the mistake of buying a bargain-basement CRM solution again. No such luck. Jeske came back ready to present his proposal. He emailed us his slides and let us know we should print copies for the attendees and that we would need to set up a projector for his demo. I had to laugh at his chutzpah. I didn't really have the time for what I knew would be a dead end, but a friend of Larry's is a friend of Larry's, so I agreed to give him a half hour. What he showed us was surprisingly well thought out, but still not ready for beta testing. Many of the essential features we required were missing, and the interface lacked the polish of the others we'd seen.

[...]

"Larry," I explained slowly and carefully, "we just went through hell with an undeveloped product. I can't burden my team with another flaky piece of software that will just slow us down. We're close with a real CRM company and should have a proposal in a couple of days. I'll let David down gently."

"No. Really," Larry repeated. "You should hire these guys. Look, they're a small company and they'll be very responsive. We can give them space in the office and they'll live here and build their product to our specs. We'll be their most important client, and we'll benefit from their growth based on our product design ideas. Have Biz Dev negotiate the contract and make sure we get some equity."

[...]

That guy was kind of a jerk anyway, so telling him no didn't bother me, but I'd still be cursing Larry's decision today if not for one small thing: Larry was absolutely right. Though we wasted weeks negotiating our investment in Jeske's nascent company NeoTonic (we squeezed just an extra one-tenth of one percent in equity out of them), by the end of October 2001 we had the new Trakken CRM system running in parallel with Miasma. David and Brandon lived in our office and Denise Griffin, our user-support manager, gave them a daily list of desired features and bug fixes. Unlike the big "reliable" company I had wanted to hire, NeoTonic didn't have hundreds of customers using the same product. They didn't release upgrades only twice a year. They fixed things as they came up, in priority order. Within a couple of months we had the CRM system we wanted, built to our specs, fully stable and intuitive to use. We cut our ties with Miasma and never looked back. A year and a half later, we bought the rest of NeoTonic, making its two founders full-time Googlers.

Insofar as we had a clear strategy, a big part of it seemed to be getting other people to do our work for free. Nowadays that's known as "crowd-sourcing." We just called it "cutting costs." Self-service AdWords, porn cookies, affiliate programs, viral marketing—all were based on many hands lightening the load and the unbeatable value of unpaid labor. Google parsed all its tough problems into manageable pieces and parceled them out.

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.

This divide-and-conquer approach even informed the basic algorithms running Google search. Rather than basing search results solely on a single source—the content of individual web pages—Google looked at links created by millions of people to determine a site's importance. Sergey called it "the democracy of the web," because each link was a vote cast in favor of a site's credibility. That approach made Google scale better than the competition, because the more the web expanded, the more links Google harvested for its ranking algorithm.

"Here are Overture's policies," AOL said, dropping a phonebook-sized document on the table with a thud. "These are their editorial policies for what is allowed and what is not. Let's see yours."

"Fine. We'll show you ours," Alan replied. "Let's set up an appointment and do it right." We had no policy manual. But we would have one by the meeting the following Monday. Five minutes after that exchange, Omid was on the phone to Sheryl. "Overture has a binder," he told her. "We have to have a binder! We have to have a binder!"

Sheryl and AdWords staffer Emily White spent the weekend pulling one together. While they were at it, they pulled together an editorial team. AOL wanted to know how many people Google had dedicated to reviewing ads and approving them. By enlisting everyone in advertising operations who had ever looked at an ad, Sheryl stretched four to fifteen. "We didn't lie to them," she asserted, "but we included everyone we possibly could."

Eric Schmidt's choice of Susan's mid-range revenue estimate proved unduly conservative. By the end of the contract's first year, we were far above the highest projections. Part of that success may have been attributable to a small shift made by an enterprising engineer. The day after the deal went live, John Bauer added code that boldfaced the keyword a user had searched for when it appeared in an ad, making it obvious that the ad was relevant. That single improvement increased clickthrough rates by four hundred percent. One engineer. One change. Four hundred percent.

What if we let users opt out of accepting our cookies altogether? I liked that idea, but Marissa raised an interesting point. We would clearly want to set the default as "accept Google's cookies." If we fully explained what that meant to most users, however, they would probably prefer not to accept our cookie. So our default setting would go against users' wishes. Some people might call that evil, and evil made Marissa uncomfortable. She was disturbed that our current cookie-setting practices made the argument a reasonable one. She agreed that at the very least we should have a page telling users how they could delete their cookies, whether set by Google or by some other website.

In mid-2003, Susan put some product plans and strategic documents on MOMA that required a password to access. She was concerned that the sales team might accidentally spill too much to clients. As head of product management, Jonathan told her to make the documents accessible because Google so strongly valued the free flow of information among staff members. Only performance appraisals and compensation were off limits. "This is extremely unusual for a company to do," Eric Schmidt often reminded us at our weekly TGIF meetings, "but we will continue trusting everyone with sensitive information unless it becomes a problem."

In September 2003, it became a problem. Information about our revenue numbers and Larry and Sergey's stock holdings started showing up in news reports. Eric immediately clamped down, telling Omid and me to stop including revenue numbers in TGIF presentations. Passwords on MOMA were no longer forbidden. It was a shame, Eric observed, that reality had finally come to Google.

The source for the stories turned out to be a low-level administrator feeding information to an outsider. She was asked to leave. In January 2004, though, long after that first small leak had been plugged, a much bigger crack appeared in our wall of secrecy. The same month we hired our first corporate security manager, John Markoff from the New York Times wrote a series of articles in which he reported details of products in development and the results of an internal audit conducted in preparation for a possible IPO. The information had been extremely confidential and closely held. The leak was ultimately traced to a senior manager who had known Markoff for years. He left the company as well, though the true reason for his departure was not made public, leading to much speculation.

From that point on, I had to ask for access to the project information I needed to do my job. It felt odd, as if with each ironclad, password-protected gateway the company installed, it locked out a little more of its original corporate culture.

Shortly before going public, Google clamped down completely. According to SEC rules, every employee who had access to intimate knowledge about the state of the business would be restricted from freely buying and selling the company's stock. I, and most others, gladly traded ignorance about our bottom line for the bliss of being able to cash out whenever we were ready to do so. The days of innocence in the garden of data had officially come to an end.