An Expressive Auction Design for Online Display Advertising

Abstract

We propose an expressive auction design that allows advertisers to specify the kinds of demographics and websites they wish to target within an advertising network. The design allows the network to differentiate impressions according to relevant attributes (e.g., geographic location of the user, topic of the webpage). Advertisers can place bids for different kinds of impressions according to their attributes, and can also specify volume constraints to control exposure. The novelty of the design is a bidding language that admits scalable allocation and pricing algorithms. We discuss the incentive properties of different pricing approaches. We also propose a bidder feedback mechanism to mitigate the complexity of expressive bidding.