Combining guaranteed and spot markets in display advertising: selling guaranteed page views with stochastic demand

While page views are often sold instantly through real-time auctions when users visit Web pages, they can also be sold in advance via guaranteed contracts. In this paper, we combine guaranteed and spot markets in display advertising, and present a dynamic programming model to study how a media seller should optimally allocate and price page views between guaranteed contracts and advertising auctions. This optimisation problem is challenging because the allocation and pricing of guaranteed contracts endogenously affects the expected revenue from advertising auctions in the future. We take into consideration several distinct characteristics regarding the media buyers’ purchasing behaviour, such as risk aversion, stochastic demand arrivals, and devise a scalable and efficient algorithm to solve the optimisation problem. Our work is one of a few studies that investigate the auction-based posted price guaranteed contracts for display advertising. The proposed model is further empirically validated with a display advertising data set from a UK supply-side platform. The results show that the optimal pricing and allocation strategies from our model can significantly increase the media seller’s expected total revenue, and the model suggests different optimal strategies based on the level of competition in advertising auctions.