Carrot or Stick? An Economic Analysis of Online Ad Fraud Deterrence

Abstract

Ad fraud is increasingly becoming a major concern in online advertising, with publishers being one of the key sources of fraudulent ad traffic. Although ad fraud deterrence is a critical problem from both technical and economic perspectives, past research has not considered their interplay. Our paper fills this critical gap by building a game-theoretic model wherein an ad network (an intermediary between publishers and advertisers) strategically leverages its technological tool (the configuration of a given fraud detection technology) and economic tool (the payment to publishers) to deter ad fraud and maximize profits effectively. Our analysis generates several interesting findings. For example, as the fraud detection technology and fraud generation techniques are getting better, we show that although the ad network needs to respond by making the technology configuration stricter to admit less ad traffic (acting as a "stick" to dampen publishers' fraud motives), they may sometimes need to increase the payment (acting as a "carrot" to induce publishers' participation). Furthermore, although many stakeholders advocate levying strict policies such as banning publishers from an ad network, we show that this prevalent practice may sometimes fail to reduce fraud traffic and hurt an ad network's profit. In addition, our results provide other useful implications that present a new theoretical perspective on the incentive problems in ad fraud generation and detection. Our study also draws valuable insights for ad networks into implementing effective ad fraud deterrence policies and for advertisers into auditing and monitoring their ad campaign performance.

Keywords: Ad fraud, online advertising, ad networks, publishers, game-theoretical modeling