I'm Feeling Lucky_ The Confessions of Google Employee Number 59 - Douglas Edwards [98]
Larry and Sergey's insistance on seeing performance metrics for marketing redoubled with the addition of our ad buy on Yahoo. They began a drumbeat of demands for better measurement of our customer-acquisition techniques. What about the promotional text on our homepage? Which messages converted the most newbies to regular users? Testimonials? Promises? Comparisons? How many ads did they click? How many searches did they do?
The only way to answer these questions was to generate the homepage dynamically—essentially to implement code that would give us the ability to deliver variant versions of the homepage to users who came to our site. That would enable us to show different users different text and then track what they did after they saw it.
Larry gave me the task of writing the text to be displayed in April 2000 and assigned the coding of the dynamic homepage to Marissa. A logs team would generate the report on how many new users came back. While it took me a long time to get signoff on the homepage messages I wrote, dynamic homepage generation proved even more elusive. Soon every conversation I had with Larry turned into an inquisition about the conversion-rate test.
"Doug, when are we going to see those numbers?" he'd ask me. "We're wasting money because we're not effectively using our most powerful promotional medium."
The conversion-rate test was one of my main OKRs (Objectives and Key Results), and each time I had to tell Larry I had no data, a bit more of my credibility crumbled. I used every method at my disposal to jump-start the project, but I simply couldn't get the priority moved high enough. The only one who could move things along was Marissa, and it had been announced at TGIF that she and Larry were now a couple. Finally, in late August, I trudged upstairs to camp outside Larry's office. I waited until he was alone, then entered and closed the door behind me.
"Larry, I've begged, cajoled, and demanded," I said, "but the dynamic homepage code still hasn't been implemented so we can test conversion rates. Can you recommend some other approach?" I was frustrated and nervous and didn't hide it very well. I was admitting I couldn't get something done. At Google, that was not a career-enhancing move. And I felt uncomfortable telling the company president that the obstacle in my way was the engineer he was dating. Larry listened quietly to my concerns.
"Don't worry, Doug," he reassured me with a broad smile. "We'll work something out." Then he put his hand on my shoulder, gave it a gentle shake, and guided me to the door. It was a strange moment for me. Larry's earnestness emanated in waves, as if he wanted to let me know that he understood the unspoken dilemma I faced, that he and I were all right and that the situation would be all right, and that he would take care of things. I went back to my desk unsure what to do next. What if other issues came up with Marissa? I could already see that we had differing perspectives about our brand. Would I have to go to Larry for resolution each time we disagreed?
Coincidentally or not, within days the dynamic homepage coding was completed. In theory, anyway. We still needed a script so webmaster Karen could run the program on her Windows machine. That took several more weeks. Then the logs team had problems extracting the user data we needed. The first actual report wasn't ready until November.
I never believed that my engineering colleagues were intentionally neglecting my number one priority. Any of a thousand projects competing for their attention could legitimately take precedence