Ad Auction Design with Coupon-Dependent Conversion Rate in the Auto-bidding World

Online advertising has become a dominant source of revenue of the Internet. In classic auction theory, only the auctioneer (i.e., the platform) and buyers (i.e., the advertisers) are involved, while the advertising audiences are ignored. For ecommerce advertising, however, the platform can provide coupons for the advertising audiences and nudge them into purchasing more products at lower prices (e.g., 2 dollars off the regular price). Such promotions can lead to an increase in amount and value of purchases. In this paper, we jointly design the coupon value computation, slot allocation, and payment of online advertising in an auto-bidding world. Firstly, we propose the auction mechanism, named CFA-auction (i.e., Coupon-For-the-Audiences-auction), which takes advertising audiences into account in the auction design. We prove the existence of pacing equilibrium, and show that CFA-auction satisfies the IC (incentive compatibility), IR (individual rationality) constraints. Then, we study the optimality of CFA-auction, and prove it can maintain an approximation of the optimal. Finally, experimental evaluation results on both offline dataset as well as online A/B test demonstrate the effectiveness of CFA-auction.