Bid Shading In First-Price Real-Time Bidding Auctions

Abstract

Online advertisements can be bought through a mechanism called real-time bidding (RTB). In RTB the ads are auctioned in real time on every page load. The ad auctions can be second-price or first-price auctions. In second-price auctions the one with the highest bid wins the auction, but they only pay the amount of the second highest bid. In this paper we focus on first-price auctions, where the buyer pays the amount that they bid. The buyer should bid more than others to win the impression, but only as little amount more as possible and at maximum what they consider the impression to be worth. This research will evaluate how multi-armed bandit strategies will work in optimizing the bid size in ReadPeak’s first-price real-time bidding environments. ReadPeak is a demand-side platform (DSP) which buys inventory through ad exchanges. We analyze seven multi-armed bandit algorithms on offline data from the ReadPeak platform. Three algorithms are tested in ReadPeak’s production environment. We discover that the multi-armed bandit algorithms reduce the bidding costs considerably compared to the baseline. This has potential to bring significant savings for the advertiser. More research is required to get a decisive result on which algorithm performs the best in the production environment.