"Turn it off. It thinks I'm German." The off-key refrain caught my ear.
"I found MentalPlex mildly amusing, but the different languages on the results page make it harder to use. The joke gets old very quickly."
Discordant voices sang about confusion and annoyance. They swelled from a hoarse whisper to a shrill harangue to a roaring cacophony of insistent outrage that could not be ignored. By ten p.m., it was painfully clear that for many in Google's global audience, jokes about Germany and mind control were just not funny. I emailed Susan, who was still at the office monitoring feedback, and suggested we get rid of the German message. She had already identified the problem and called Sergey. He okayed removing the German error message and the linked results page. Okay. A minor hitch resolved.
Except, it was Friday night. All the engineering staff were with Sergey, washing away the week's woes at Zibbibos, a trendy restaurant twenty minutes away in downtown Palo Alto. No one in the office was authorized to make the change. I sat at home watching more and more messages complaining about Germans, German thoughts, and German jokes drop into my mailbox and explode. Little ice balls formed on the back of my neck, then rolled down my spine.
A half hour passed. Then another. We were taking a pounding on email. Finally, the offensive German ground to a halt. Thanks to some anonymous engineer, no more unwanted German results confused our users. Now our unwanted results were in Portuguese. The engineers thought the joke was just too funny to eliminate entirely, so they simply shifted the interface to another language. Users didn't like German? Fine, we'll give them something else.
"No! No! No!" I told them. The problem wasn't the language; it was how hard the foreign interface made it to use our results. We were breaking our brand, and instead of repairing the cracks they found it more amusing to watch our equity pour onto the floor. I had lived through PR crises before, and this one was becoming a sabakova kashmar.*
Complaints kept coming. Though the tone was less virulent than when the German text was up, users were still unhappy they couldn't navigate the site easily. We were breaking a cardinal rule by making it difficult for them to get to information they sought. One user lamented that our little joke was costing her money—a professional researcher, she could no longer use Google to do her job.
I had worked at big companies long enough that I hesitated to escalate the problem to our top executive. But I had worked at Google long enough not to be intimidated by an org chart. I called Sergey. It was hard to hear over the background noise of rowdy engineers in a crowded restaurant, but I could tell he was surprised when I insisted we drop all the foreign-language results.
I was pooping the party, but Sergey reluctantly agreed, most likely so he could rejoin the festivities without fear of another interruption.
It took forever to reach Susan, and it was midnight before all the foreign-language text was stripped off the site.
"A not-insignificant fraction of our users are complete idiots," groused one engineer, "if they can't figure out how to use our site, just because it's all in Portuguese." Google had clearly crossed the gap from serving the tech elite to playing in the mainstream market—an online segment he knew to be densely populated with the clueless.
"I'm more worried that we got spooked by a little negative feedback," said Howard, the iconoclastic easy-riding engineer. "We backed off the playfulness that's an important part of Google. We watered down our April Fools' joke to make it less invasive. I guess that's what happens as we grow up—we become a more conservative company." He did not see that as a positive development.
Sergey tacitly agreed that the problem was not on our end. German results were quite funny. Besides, far more user feedback about MentalPlex was positive than negative.
At two a.m. I crawled into bed. I dreamed a monochrome dream about Germans and Lisbon and a police captain who looked at me quizzically and asked, "Shut it down? But everyone is having such a good time!"
The next morning I felt emotionally hung over. I had launched a joke that worked. MentalPlex perfectly fit Google's unconventional and just-a-bit geeky style. Sergey had been right all along—it was okay to play with our brand. But the mistake of translating the results interface embarrassed me. I worried that it didn't embarrass any of my colleagues.
Why had it required so much effort to make a change once complaints started coming in? Why had the decision been made to switch to Portuguese? Who had made that decision anyway? What was my role as brand manager, if not to manage the brand? I now understood where true power resided at Google. It lived in the keyboards of those authorized to push a new GWS, to actually check in the code that drove the site. Those engineers were the ultimate gatekeepers. Even when I had a green light to move a project down the tracks, someone with a different idea and a hand on the switch could change our course in the middle of the night—leaving me to awaken a hundred miles away from where I'd thought I'd be when I went to bed. MentalPlex had almost derailed us.
As soon as I got to work on Monday, I wrote up a postmortem memo to keep it from happening again. I included all the data I could glean from the six hundred email responses we had received. Seventy percent liked MentalPlex. Most of the rest complained about the translation of the interface into another language. My note codified the lesson that we should never, ever intentionally make it harder for people to search. I drafted a letter of apology to users who had complained and offered to send them free Google t-shirts. The nightmare was over. We had taken our lumps and learned from our mistake, and now we could roll on.
Not necessarily so, Marissa argued.
"The problem is not that we translated the interface into German," she said, "but that we called attention to the change by including a 'tip' about MentalPlex detecting German thoughts." In her view, if we had just put a disclaimer on the page that it was a joke, no one would have been upset.
"Most people would likely not have even noticed the translation without the tip," another engineer agreed. "Only the navigational text on the page was rendered in German—the web results themselves were still in English."
How could users not notice, I wondered, that all the text except the actual results was no longer in English? And adding a line at the bottom explaining it was all in jest would not only have failed to solve that problem, it would have been an admission that the joke was too weak to sustain itself. Jonathan Swift rarely used asterisks to explain he was being satiric.
Marissa and I never saw eye-to-eye on MentalPlex. She questioned my categorization of the user feedback even when I sent the original emails to her. I had cut-and-dried, irrefutable, objective facts at my disposal, yet we fundamentally disagreed on how to interpret the data. I wasn't interested in placing blame, but I did want to make sure we learned from our mistake and didn't repeat it. To do that, we had to reach consensus on what the mistake had been.
Omid Kordestani, the head of our sales and business development group, offered his opinion. "Really amateurish!" he complained. He meant MentalPlex from start to finish. Omid hadn't seen MentalPlex coming, and when it smacked him in the face on April 1, he was furious. Calm, smiling, even-keeled Omid derided our joke as likely to alienate the very advertisers his team had been working so hard to befriend. And had we even considered international users who simply wouldn't get our all-American attempt at humor?
"I hope we never repeat it," he admonished us. "Thank God this happened on a Friday night." There was only one silver lining Omid could see. Portal sites like Yahoo, who thought we might compete with them, would no longer worry about us. We clearly wouldn't be stealing their users with this kind of immature grab-ass idiocy. We weren't a Stanford dorm-room project anymore, Omid reminded us. We were running a serious business, and this came off like a beer-fueled freshman prank. I think he flashed back to his days at Netscape—a company high on its own hubris until a giant jackboot ground its face into reality.
"Not to worry, Omid," Sergey offered reassuringly. "We do foolhardy things from time to time,
but not at random, and they usually have positive results."*
I pointed Omid to all the positive feedback from our users, but Cindy let me know I had blown it big time. Omid was a stakeholder and should have been in the loop. Another lesson learned the hard way. If nothing else, I consoled myself, at least this would give impetus to a more formal sign-off process. That would reel in the late-night GWS pushers and stop their off-key improvisations. As if.
Though no one else seemed to take a lesson from the MentalPlex mishap, I found an epiphany in the dust of the dying hubbub. I could build Google's brand from inside the product. I didn't need to rely on banner ads or postcard programs. Google's personality would shine through in the way we—the way I—talked with users. I could try anything, because the only ones looking over my shoulder were two guys noticeably lacking inhibitions. It was a liberating moment.
I owned Google's words. Now I would give them a voice.
Disorienting in the Extreme
Google time folded in on itself like a tempered samurai sword. So much activity took place simultaneously that the linear narrative of this book flattens the immersive 3D experience into an unrecognizable shape. Maybe if you tear out the pages, throw them in the air, and read from them randomly as they flutter down around you, you'll get a better feel for what it was like. I'd been at the company five months, and every hour of every day another neural pathway in my brain uprooted itself and traced a new route to an unexpected destination.
"I've drafted some lines we could put on the homepage," I told Larry and Sergey at a product review. "They should drive repeat traffic by reaffirming our quality to first-time users. Here's the list."
Sergey didn't look up from the new phone he appeared to be breaking into pieces.
"How are you defining first-time users?" Larry asked. "What if they have cookies turned off?" He glanced at the dozen rows of my text the projector pinned neatly against the screen. "Why don't you use the testimonial quote from Time magazine?" he asked before I could answer his first question. "I don't think any of these are going to be very effective."
Sergey raised his head to stare at Larry, who raised an eyebrow in response and spoke.
"You need to make these much more compelling. Like, 'If you printed out all the data Google searches, the stack of pages would reach from here to the moon.'"
Sergey smiled and leaned farther back in his chair.
"No," he said.
Larry's eyebrow shot up again and hovered at its apex.
"What we need to do," Sergey went on, "is test things that are totally random to see what has the biggest effect." His eyebrow called Larry's and raised it a quarter inch.
They stared silently at each other for a moment, their eyebrows dancing a pas de deux in a closed loop of telepathic communication. I sometimes thought I grokked Larry or understood where Sergey was coming from. Observing them together, though, was like trying to catch a glimpse of light in the thin patch of space between twin neutron stars. There was enormous energy being exchanged, but at a spectrum range I could not detect.
Sergey returned his attention to the disassembled phone. "You know what we should do?" he asked his lap. "We should make the homepage hot pink and see how many people come back."
Larry's face lit up as if all the eyebrow bouncing had earned him a free ball and a bonus round.
I smiled too, but when I stopped, they didn't. A doubt crept in. They weren't ... they couldn't be ... surely they didn't mean that as a serious suggestion? I began the long trip from incredulity to denial to sputtering refutation based on logic, only to be sent packing until I could come up with a data-based argument proving their idea lacked merit.
Staking out extreme positions and making staffers fight their way to a safe haven of sanity seemed to amuse our founders. They got to watch forehead veins throb and palms drip sweat as PMs, ops people, and engineers probed carefully, attempting to keep a grip on logic while cutting through a thicket of distractions to defuse explosive ideas.
"Google was the first company I worked at where I had that experience," project manager Deb Kelly recalls. "Where you're thinking, 'That's nuts. That's crazy. That's dumb,' and then thinking about it for a while and going, 'Actually, that might be really, really smart—hard, but really, really smart.'"
The madness was not without method. Not only did Larry and Sergey's hyperbolic proposals force us to reason more tightly, but starting at the ideological antipodes exploited the full value of the intelligence in the room. After Larry or Sergey made one of their outrageous suggestions, nothing that followed would seem inconceivable. To sort the improbable from the impossible, we needed to pay attention and to argue facts, not suppositions or conventions. If we began with only "acceptable" solutions that had been tried before, we would never uncover the state-changing breakthroughs that destroyed worlds and raised new ones from their rubble.
Urs understood this better than most. "Larry and Sergey had an expectation that things would be watered down along the way," he explained. "Starting with something that's more ambitious will get you something that's reasonable. But if you don't put the goal post way out there, people are already taking fewer risks and are less ambitious about how big the idea should be."
It was another reason Google valued intelligence over experience. "In the best case," Urs said, "you had someone who was very excited and didn't know what was impossible and got really far. The big question was always, 'Can you do it cheaply enough for this to actually be affordable?' Maybe, but without trying it and measuring and doing the next iteration, you're not going to find out."
I felt sorry for new Googlers who stepped into the reality abattoir for the first time and looked around the room with naked-at-school-nightmare eyes, trying to see if this was some cruel initiation rite. If you were a goal-oriented, eager-to-please new staffer, what did you say when Sergey insisted, "We should really forget about [whatever topic was under discussion] and focus on building space tethers?"*
Meetings weren't the only place ideas went to die. If the founders didn't like a proposal, they might consign it to what PR director David Krane called the "dead letter office": "You'd ship something into the ether and there was no response. That was your way of knowing you'd probably missed."
It didn't help that when Larry and Sergey did respond to our ideas, their responses were often ambiguous. "If they liked something they'd say, 'It doesn't seem too sucky,'" recalls facilities manager George Salah. "When they pushed back on something, they'd say, 'Hmmmm, that seems suboptimal' or use some technical way of saying something between yes and no. They never gave a clear decision."
When I took copy to Sergey for approval, he would say, "It's cute. I like it" or "No. That's not very Googley."
Once I spent days with my team developing a full rationale for an ad campaign, based on what we understood about the target audience and their motivations. Sergey glanced at the layouts, frowned, and said, "I think you need to think about it some more."
"Is it the concept you don't like? The art? The wording?" I asked, looking for something more definitive so we could eliminate what was offensive. He simply shook his head and turned back to his monitor.
"Think about it some more," he repeated infuriatingly.
Eventually I concluded that because Sergey was an engineer, he wanted our ads to depict an idealized world in which everything was optimized, including the people. I discovered that he disliked images that didn't fit a classical notion of beauty and thus could be seen as outliers in the data set of fully realized human potential. One trade ad for our advertising program included a picture of a sweaty sumo wrestler. Sergey wrinkled his nose when we showed it to him. "I don't think we should run ads with unattractive people in them. Our ads should always be aesthetically pleasing so people will think happy thoughts when they think of Google."
Devin Ivester, who left his own ad agency to join us as a creative-team leader,* remembers another such occasion, when he showed Sergey an ad with an exhausted man asl
eep on a couch, surrounded by books, notes, and papers. Sergey's response was, "No, we can't use that guy. He's ugly."
"You don't mean ugly, right?" Devin asked, somewhat taken aback.
"Well, no, but he's obviously not organized. He's just not a good person."
"How is he not a good person?"
"Well, look at him. He's just asleep on a couch."
"But if you read the headline, he's asleep because he's actually been working all night. He's almost tireless, but at some point he's had to give up."
"Well, obviously, he's disorganized then—if he can't keep up with his workload."
The lesson for Devin was that "Sergey had a very Disney-like idea of what we should show in our ads. It didn't have to be so real life. It was his idea of the perfect Googler and how people should be."
Perhaps because they viewed the world through a polarizing filter of "ideal" and "suboptimal" (or "good" and "evil," if you will) and were so confident about which position they occupied in this binary system, our founders displayed a fondness for hyperbolic vilification of those who disagreed with them. In almost every meeting, they would unleash a one-word imprecation to sum up any and all who stood in the way of their master plans.
"Bastards!" Sergey would mutter if a competitor signed a client we were pursuing.
"Bastards!" Larry would exclaim when a blogger raised concerns about user privacy.
"Bastards!" they would say about the press, the politicians, or the befuddled users who couldn't grasp the obvious superiority of the technology behind Google's products.
It was a little intimidating until you got used to it, but it wasn't long before "bastards" became corporate nomenclature for any individual or institution that didn't see things the Google way. Ratings services undercounted our traffic? Bastards! Hard-drive vendors refused to cut deals below wholesale cost? Bastards! It became so prevalent that someone proposed that Sergey's five-word acceptance speech for an online award should be "The Webbys are for bastards."
I'm Feeling Lucky: The Confessions of Google Employee Number 59 Page 13