Authors: Douglas Edwards
Larry recognized that the bidding aspect of the GoTo system had value. The flaws, though, he didn't know how to fix, and he remained skeptical that we needed a CPC pricing model to compete at all. If we did, we certainly didn't want an imperfect system to deliver it. It was hard to believe GoTo would be very successful anyway—their results were nowhere near as good as Google's. Larry put off dealing with the question. For now, anyway, Google had more important things to worry about.
Having banned banners, we looked for other sources of revenue to supplement our plain text ads. A company called RealNames fit into that plan, but not comfortably.
Foremost among the moneymaking strategies our competitors employed in early 2000 was pay-for-placement, the practice of allowing advertisers to pay a fee to receive a more prominent position in organic search results.
*
GoTo's model was entirely pay-for-placement since it blatantly sold the top position to the highest bidder. Other sites simply included listings in the body of their search results without identifying that money had changed hands to put them there.
Pay-for-placement made Larry angry. It was unethical to confuse users about what was objectively useful and what showed up because someone paid to jump to the front of the line. Besides, it ruined a good algorithm. If you were generating useful results, why intentionally corrupt them with less valuable data? In the paper they had written together as grad students at Stanford, Larry and Sergey had identified pay-for-placement as "more insidious than advertising, because it is not clear who 'deserves' to be there, and who is willing to pay money to be listed."
†
The problem with religious principles is that believers employ different standards of orthodoxy. For us, RealNames was the shibboleth that separated purists from pragmatists. RealNames sold keywords to companies and then worked with search engines to make sure that links to their clients' homepages showed up at the top of search results. For example, Ford might buy the keyword "Ford Explorer" to ensure that searchers for that term would see a link to Ford's own Explorer page. If we didn't already display that page as our first result, RealNames would pay us to include the link above and separate from our results. MSN, AltaVista, and
Go.com
were doing it, but we struggled with how to display RealNames links while making it clear they were neither Google-generated results nor paid advertisements.
"I feel as though we've sold out with this RealNames thing," engineer Paul Bucheit complained to a large group of Googlers including Larry and Sergey. "RealNames is a real source of junk and I think it is only going to get worse." Paul cared deeply about engineering quality and about ethics and especially about the intersection of the two. He summed up his attitude some time later with the phrase "Don't be evil." It caught on.
"This is even worse than auctioning off results," Paul complained. "Just for fun, I registered the RealNames keyword 'Craig Barrett.' Now when you search for 'Craig Barrett' on Google, the top result links to my hideous little web page." I tried it and found myself staring at a glittering pink screen full of animated unicorns and rainbows. At the time, Craig Barrett was the CEO of Intel.
It had never even occurred to me that search engines had integrity to protect. An Internet search engine was just a business, after all, not a public watchdog. Yet the RealNames debate took me right back to 1998 and the boardroom of the
Mercury News.
I had developed guidelines for accepting advertising on
Siliconvalley.com
, and the publisher grilled me on every worst-case scenario. "Will Microsoft be able to buy ads next to coverage of their antitrust trial?" he asked. "people will think we've compromised our editorial objectivity." It took weeks to iron out all the permutations to his satisfaction.
In Google's initial implementation in early 2000, we displayed only a small "RN" next to each RealNames result. That didn't test well. Or rather, it tested too well. In usability groups, Stanford students thought that the RealNames link was an extra-special search result that we gave prominence because of its superiority to the rest of the dreck we'd found.
Omid Kordestani was our VP of business development and responsible for Google's revenue generation. He started off defending the RealNames deal, but just when I expected him to dismiss all the research conclusions as flawed, he said something odd: "There should absolutely be no reason for Google to sell out its quality, even for great revenue."
What? A business guy turning down money already in hand? That kind of blew my mind. Who was this "salesman" who could see past his commission checks?
In the spring of 1999, Larry and Sergey had hired Omid—a Stanford MBA and a former Netscape exec—as Google's eleventh employee. His black Armani suit slimmed his silhouette, despite the travel weight he packed as a souvenir of globe-hopping sales calls. He smiled often and laughed easily. He listened when other people talked. I worried that he was too nice to be our dealmaker. Where was the forced sincerity I saw modeled in professional selling seminars? Where the thrashing, slashing, primordial-reptile-brained need to crush, devour, and destroy our competitors? Omid crushed you with kindness and devoured only desserts. Over time, Google ate away his extra pounds, but not the cuddly-teddy-bear persona, which masked an extremely intelligent and analytical mind. Omid could see where deal terms would settle and would shepherd both sides to that resolution point without snapping at their heels along the way—though he did on occasion unleash others who were perfectly willing to bare their fangs.
"When RealNames results are bad, they're really bad," Sergey noted, "but that doesn't mean we shouldn't have them. We should clearly differentiate them from Google results so that users can tell they're paid for."
Larry didn't want to change anything. "I don't think it's a problem if people confuse RealNames with our search results," he said. "If the result is better, they should probably click on the RealNames link. As long as it's clear to someone that it isn't exactly like a search result, that seems fine to me." Besides, he didn't want to give RealNames any free branding by identifying links with the RealNames term "Internet keyword." It could help them grow into a competitor.
"Slightly better for business, slightly worse for UI," Larry acknowledged. Most likely, he thought, RealNames wouldn't survive the year anyway. Society would be better served if we used their capital to improve Google before they lost it to bankruptcy.
It surprised me that Larry took the issue so casually. He seemed to be backtracking on the notion of keeping our integrity intact. Fortunately, someone was willing to point out the danger.
"I couldn't disagree more," said Urs. "While a relatively mild form, it's still a paid placement, and making it look like an 'objective' result is deceptive. The
New York Times
didn't become a leading newspaper by blurring the distinction between paid ads and independent news."
It was Marissa who found a solution. She examined how other search sites implemented RealNames results and concluded that our UI didn't meet the same standards. She sent around mockups that more clearly separated RealNames listings from our search results, and Larry selected one. The partnership lasted twelve months, and a year after we ended it, RealNames was no longer in business.
Once the issue had been resolved, I reflected on what I had seen. At the
Mercury News,
approvals took months and required written proposals in scheduled meetings. Opinions were formally presented, debated, evaluated, and carefully revised. At Google, quick mockups ruled, data persuaded, and decisions were made in hours. Yet Google was no less intent on setting an industry standard for incorruptibility. I embraced the notion that speed need not be the enemy of ethics. And I began to view competitors who took payment for placement in their results as duplicitous—if not outright evil.
My corporate branding experience had not been an asset so far, but with the imminent launch of our first major product since the search engine, I prayed that would change. G.O.D. would provide. And by G.O.D., I meant the Google Open Directory.
Based on Netscape's Open Directory project, G.O.D. would siphon users from Yahoo, though it would have been impolitic to admit that publicly in January 2000. We wanted to maintain our friendly relationship with Yahoo—they owned the galaxy in which we were but a small rising star. Mostly, though, G.O.D. was an annoying, if necessary, distraction.
"Everyone should just use search," Larry complained. Directories were dead. To find something in a directory, you had to understand the structure of the categories within it and then drill down through layer after layer to the hidden pool of knowledge you sought. With search, you just threw everything into a big ol' sloppy data bucket and fished out answers using words describing what you wanted to find. It didn't matter where your data swam, because a good search engine would catch it, review it, rank it, and serve it to you before your fingers stopped twitching from clicking the mouse. Search? Efficient. Directories? Not.
People, though—irrational, illogical, and idiosyncratic as they are—sometimes preferred to troll the data themselves instead of dispatching a helpful search engine with an enormous net. Perhaps they wanted to see what other information fell into the same category. Perhaps they felt more comfortable worming out the perfect little fact or file or figure for themselves. If people insisted on wasting time with a data-retrieval method other than search, Google would oblige by offering an alternative—a far less useful alternative, but an alternative nonetheless.
We couldn't incorporate the best-known directory into Google, because it belonged to Yahoo. So Google hooked up with the Open Directory, a project under the stewardship of Netscape (and its parent AOL). Thousands of volunteers reviewed websites, then added them to the appropriate directory categories. That human involvement also offended search purists. Who could be trusted to categorize sites accurately? Even if a semi-intelligent agent looked at every site, figured out what it was about, then slotted it into its appropriate cubbyhole, there was no chance a manual operation could keep up with the Internet's growth. Robotic software tirelessly scouring websites at the speed of light—that was the future.
Once Google had found a directory, we couldn't just duct tape it onto the search engine. You don't put a rust-stained camper shell on a Ferrari. So Paul Bucheit added a search component to our version of the Open Directory. Now people who insisted on browsing for information could at least use search to get to the appropriate category to browse. It was an ugly baby, but it was our ugly baby.
Now our ugly baby needed a name.
Product naming is a tricky process. Do you give a product a new name that suggests it's completely separate from your core business? Do you incorporate your primary brand name to show that it's equal to, but distinct from, your existing product? Or do you simply use a generic term to describe it so that it becomes just another service you offer, rather than a distinct product that stands alone?
"Let's label it with a lower-case
d
and make it a feature," I said to Salar. "And let's use that as the model for future product launches. We want all our brand equity to reside in the Google name, not in each sub-product." I sketched out a brand architecture showing Google in a separate box at the top, with "directory" and other potential services in smaller boxes underneath. "That makes Google more powerful as a primary brand," I explained, "so it can add credibility to any new service we introduce."
Salar agreed. "Let's stay with this strategy as we build out the information types accessible from our site," he said, "like Google images. We'll brand them as extensions of Google's service rather than new products." That decision, which took a few minutes to make, ensured that all of our engineers' awe-inducing achievements would accrue to the benefit of the Google name, rather than be parsed out to a hundred different product lines.
I look back on that as a quarter-hour well spent.
A brand is the sum of all the "touch points" you have with a product or service—your interactions, your impressions, your expectations, your unplanned casual encounters. Brand managers strive to make those touch points consistent. An obsessive brand manager at Google would have reviewed every pixel on our web page, every punctuation mark in our customer emails, every word in our sales literature, and every note of our switchboard's hold music. That's pretty much what I did. It absorbed an inordinate amount of energy, and I knew I couldn't apply myself at that level to managing a thousand Google sub-brands.
"Why don't we create a thousand Google sub-brands?" Sergey asked me a short time later. "That way we can give users all the things they like about our competitors, except with really good search results. We could do a sweepstakes site like iWon or a portal site or natural language–style queries, only with Google on the back end running the actual searches."
His logic appeared sound: if you could get all the benefits offered by competitors
plus
a search engine that actually worked, why would you use anyone else? I pointed out that with our limited resources, promoting a dozen or a hundred other brands would be counterproductive. Unless he wanted to hire more brand managers? Sergey backed down, but he revisited the idea anytime a competitor appeared to be gaining market share with search technology inferior to ours.