Hitting the Jackpot with a Greedy Bidding Strategy

"Greed is good," uttered Gordan Gekko in the quintessential 1980’s flick Wall Street. Though no corporate raider, Michael Schwarz couldn’t agree more—at least when it comes to algorithms.

Schwarz, a Yahoo! research scientist specializing in microeconomics, recently co-authored a paper examining so-called "greedy bidding strategies." These are algorithms that have the potential to increase efficiencies for advertisers and broaden the overall advertising pie for search providers.

The paper set out to answer a range of important questions, including: How can advertisers more effectively participate in keyword auctions to maximize revenues and boost their bottom line?

Today, for example, a travel agent can bid on search terms like "Hawaii" or "getaway". When a user searches on any of those words, a sponsored link to the travel agent’s website appears on the results page. If the user clicks on this link, the advertiser pays the search engine. The process of determining which ads get assigned to which keywords and how much each advertiser pays is done via keyword auctions. For each keyword, advertisers submit a bid stating the maximum amount they are willing to pay per click.

But the question for advertisers is: How much should they bid on these keywords, and is that very best price they can pay? "If you get a lot of clicks but your profit per click is zero, that is bad," explains Schwarz. "If you have huge profits per click, but the number of clicks is miniscule, that also is not very good. You need to find the sweet spot where profits and clicks are maximized."

In the paper, Schwarz and his colleagues reference what they call the "envy-free" point.
This is the point where Advertiser A doesn’t care if Advertiser B undercuts his bid. The concept was first introduced by Schwarz in an earlier paper he co-authored with Michael Ostrovsky and Ben Edelman, two of his former students at Harvard.

On the one hand, Advertiser A is happy to get undercut because the position of his ad on the results page improves and he gets more clicks. But on the other hand, he now has to pay more per click. "Envy-free" is the point at which the extra clicks perfectly compensate the additional fees the advertiser must now pay.

"The interesting fact is that if everyone lives at the envy-free point, it is equilibrium," says Schwarz. "But if someone believes he is no longer at the envy-free point and changes his bid, everyone else will adjust their bids in random order and eventually they will converge again on the same equilibrium."

The paper shows that the envy-free bidding strategy performs very well against standard bidding strategies, making it useful for search engines and advertisers. In fact, if a bidding agent based on this strategy were available, advertisers would not have to strategize about how much to bid, because it would be done for them automatically.

"If everyone used these robots, we would arrive at a stable equilibrium," say Schwarz. "There would be no profitability deviation and collectively both advertisers and publishers would be doing as well as they possibly could."

Of course, this means more money for everyone, which is the ultimate win-win scenario.