Eliciting preferences of sponsored search advertisers: implications for mechanism design

Sponsored search advertising attracts hundreds of thousands of advertisers, many with dozens or even thousands of campaigns, leading to tens of millions of distinct keyword bids. Advertiser objectives are heterogeneous. Some advertisers primarily focus on making immediate sales that are referred by clicks, while others want to promote their brand with a top-position placement. In this letter we demonstrate how one can use the empirical bidding data to recover the values of bidders in a sponsored search marketplace when the type of bidder preferences is known (i.e. whether a given bidder values clicks). We also show how one can use the history of bid changes for a given bidder to recover both the type of preferences for this bidder and the value at once. This methodology has direct implications for mechanism design making the case for combining the empirical work and auction design to avoid the optimization of the auction mechanism for the wrong preference type of the bidders.