Online Guaranteed-Delivery Advertising under Frequency Capping

As the number of network users grows, advertisers are more likely to display their ads to users on webs. Motivated by this trend, we consider a practical form of Ad contract which allows advertisers to specify the number of ads to display (Guaranteed-Delivery), and the maximum number of times each user is exposed (Frequency Capping). In this paper, we propose a two-phase framework to minimize under-delivery with the constraints of frequency capping. As it is hard to accurately predict the user frequency, our algorithm does not rely on this prediction. Moreover, we take other hard but actually demanded features into account by using a pattern-based delivery mechanism. For example, our algorithm allows advertisers to implement any desired pacing of ads over time at the user level, and try to deliver as many varieties of ads as possible to each user. Numerical experiments with real system data show that our algorithm produces high-quality solutions and has promising running time and scalability.