Optimal Dynamic Auctions for Display Advertising

We investigate the optimal dynamic auction design for the display advertising industry. Currently, display advertising is sold through two markets side-by-side. In the traditional guaranteed market, the publisher commits to deliver a pre-specified number of impressions within a fixed time frame through a guaranteed contract. In the spot market, the publisher runs an auction to allocate the impressions, and the supply of heterogeneous impressions is highly uncertain and non-storable. We characterize the precise tradeoffs between extracting the revenue from the spot markets, materializing the instantaneous benefit shared with the guaranteed advertisers, and releasing the pressure of paying the penalty eventually. Furthermore, we identify the dual role of the publisher as a system designer and as a bidder on behalf of the guaranteed advertisers. With heterogeneous due dates of guaranteed contracts, we demonstrate the inherent scheduling issue and solve the joint scheduling and capacity allocation problem.