Sunday, October 30, 2005

Google ads

Nice article in the NYTimes describing Google's main revenue source: advertising. As a veteran of this era, I witnessed advertising as a revenue model go from very successful (circa 1999, mainly due to cost-per-view purchases by huge dotcoms willing to pay to aggregate eyeballs) to horrible (circa 2002, as supply outstripped demand and cost-per-view became cost-per-click, and then cost-per-action), and finally to success again, with auction models and a shortage(!) of prime ad space. It was one of the wildest swings in value of a commodity (Internet advertising) ever witnessed -- probably a swing on the order of a factor of 100 within a few years. (For veterans, from CPM of $10 to complete inability to sell CPMs at all, and an implied CPM value using click-through or action rates of less than 5 cents or so...)

Very few people know that it wasn't Google that pioneered the ultimately successful auction model (now used by Yahoo, Google and MSN). It was Bill Gross' Goto.com, which became Overture, which was acquired for billions by Yahoo. The Times article does a good job of clarifying exactly how innovation proceeded. Note how young (and smart!) the people involved were (are). Gross is a Caltech grad, Brin and Page were Stanford PhD students, Schmidt is a Berkeley PhD and Kamangar is a Stanford grad. It's yet another case of the non-linear value of brainpower in this century...

In early 2002, a Google employee, Salar Kamangar, now 28, convinced Mr. Schmidt and the founders to switch to an auction-based system like the one set up by Bill Gross, the head of IdeaLab. Mr. Gross had created Goto.com, a search engine made up entirely of ads, where advertisers paid only if their ad was clicked on, and the advertiser who bid the most per click was listed first. (Goto was later renamed Overture Services and then bought by Yahoo, an early Google backer that has become its fiercest rival.)

Mr. Kamangar, though, had an important improvement on the model. Rather than giving priority to the advertisers that bid the most per click, as Goto did, he realized that it was better to save the front of the line for ads that brought in the most money - a combination of the bid and the number of clicks on the ad. This was not only more profitable, but it also linked readers to ads that were more relevant to them. He also figured out that the system should use what is called a Vickrey auction - that is, to charge the winner only one cent more than the second-highest bidder. That gives advertisers an incentive to bid high, knowing that they will not be penalized if they are far higher than the rest of the market.

Mr. Page and Mr. Brin were suspicious of any system that put high-bidding advertisers at the top, Mr. Kamangar said. "They thought if someone was willing to pay more it was a negative," he recalled. But he was able to convince them that the site could be improved by incorporating how often users clicked on an ad.

Mr. Schmidt, who was still new as chief executive, was worried more that moving to an entirely auction-based system - amid a recession in online advertising - could be financially disastrous. "I said to Salar, 'Promise me the revenue won't go down,' " Mr. Schmidt said. "I was afraid people would realize these ads were worthless." In fact, revenue quickly increased tenfold.

As Google's audience took off, advertisers came running - many thousands of smaller ones at first, but soon large companies as well. Among Google's largest advertisers is eBay, which has long bought keywords for nearly every sort of merchandise it sells.

"The smartest thing that Google did was getting smaller advertisers to buy in," said Ellen Siminoff, the chief executive of Efficient Frontier, an agency that helps advertisers manage their campaigns on search engines. She estimates that Google has two to three times as many advertisers as Yahoo does, largely because Yahoo has a 10-cent minimum bid. This lets Google earn money on more obscure search terms for which rivals have no ads.

This growing advertising business gave Google the confidence to expand its audience. Most significantly, in 2002, America Online brought in Google to replace Overture, which provided both search and search ads; that deal enshrined Google as the premier search engine and ad network. Google won the deal by guaranteeing AOL a substantial sum, which it would not disclose. Google was willing to make that bid only because of its confidence in its advertising sales prowess. "If we were wrong," Mr. Kordestani said, "there were some scenarios that would bankrupt the company."

But by that point, Google had figured out that the same sort of computing and engineering skill that it used to find Web pages could also be used to improve the quality and, ultimately, the profitability of advertising. "Initially, we didn't understand how fundamental the computer science was in advertising," Mr. Schmidt said. "We didn't have enough staffing or focus on this area. I managed to fix that."

GOOGLE introduced its current system for determining which ad to show on which page late last year. It is a wonder of technology that rivals its search engine in complexity. For every page that Google shows, more than 100 computers evaluate more than a million variables to choose the advertisements in its database to display - and they do it in milliseconds. The computers look at the amount bid and the budget of the advertiser, but they also consider the user - such as his or her location, which they try to infer by analyzing the user's Internet connections - as well as the time of day and myriad other factors Google has tracked and analyzed from its experience with advertisements.

"If someone is coming from a particular location, a certain ad may be more popular there," explained Jeff Huber, Google's vice president for engineering. "The system can use all the signals available, and the system itself learns the correlations between them."

No comments:

Post a Comment