 |
Daily bidding for top billing on the Internet
 |
EECS professor Umesh Vazirani says he never noticed the ads on Google until he started working on the “pagerank algorithm” two years ago. The focus of his work is quantum computation, which, he says, involves “sitting around and thinking about what computers would look like if they had quantum mechanics.”
PEG SKORPINSKI PHOTO
|
Every day, the hammer goes down on millions of online auctions. And we’re not just talking about eBay. Search engines like Google and Yahoo! have created an advertising revolution and become billion-dollar businesses by auctioning off popular keywords to businesses that want their ad to rise to the top of the page.
Berkeley EECS professor Umesh Vazirani and colleagues (one of whom is his brother and Georgia Tech computer science professor Vijay [Ph.D.’84 CS]) have developed a novel computer algorithm that could further refine the advertising auction model, making the process more competitive, optimizing income for the search engines, even allowing small businesses to get in on the action.
If you search the word refinance on Google, for example, you’ll see a list of mortgage brokers on the right side of the page. The top spot usually goes to those willing to pay the most each time someone clicks their ad. The clickthrough rate—how often an ad is clicked when linked to a specific keyword—is also factored into ranking the ads. Each day, advertisers specify their budget allotment for the automated auctions; once that budget is drained, the ad drops off the page.
Under the current system, an advertiser with a small budget who bids on an expensive keyword could run out of money early in the day, dropping out of the auction and reducing competition for that keyword. Indeed, a savvy opponent with a larger budget could purposefully bid high early in the day, raising the price of the keyword and forcing the low-budget advertiser to make an early exit.
Such antics are not good for Google's bottom line, however, because Google earns more when there are many advertisers competing for slots. The researchers’ new mechanism addresses this problem by lowering the ranking of advertisers who are short on cash, forcing them to spend more slowly and stay in the auction a little longer. “Our mechanism is more resilient than traditional methods for addressing this kind of gaming,” Vazirani says.
The researchers devised new mathematical tools that determine the optimal tradeoff between bid and daily budget for determining the ranking of advertisers. They have filed for a patent but hope the research will remain in the public domain for others to build upon. For more on the story, go to the College’s online research digest, Lab Notes.
|
|
FOREFRONT takes you into the
labs, classrooms, and lives of professors, students, and alumni
for an intimate look at the innovative research, teaching, and
campus life that define the College of Engineering at the University
of California, Berkeley.
Published twice a year by the College of Engineering Office of Marketing & Communications. Have a comment about Forefront? E-mail
your letter to the editor. Click here
to learn more about the magazine. |
|
