Optimizing Long-Term Revenue in Ad Auctions with User Response Modeling

The core message of this paper is to propose a Markov Decision Process (MDP) model to capture the user's response to the quality of ads, with the objective of maximizing the long-term discounted revenue for the ad auction platform. The authors characterize the optimal mechanism as a Myerson's auction with a notion of modified virtual value, and also propose a simple second-price auction with personalized reserves that achieves a constant-factor approximation to the optimal long-term revenue.