Combining guaranteed and spot markets in display advertising: selling guaranteed page views with stochastic demand
share
This paper proposes an optimal dynamic model for combining guaranteed and spot markets in display advertising. We assume a media seller can well estimate the total supply and demand of page views from a specific advertisement (in short ad) slot in a specific future period. The model helps the seller to determine how to distribute and price those future page views between guaranteed contracts and advertising auctions. The former is sold algorithmically in advance while the latter happens in a few milliseconds after a user visits the Web page in the future. Therefore, the former is called programmatic guarantee (PG) and the latter is called real-time bidding (RTB). This is one of a few studies that investigate the RTB-based posted price PG for display advertising. The optimization problem is challenging because the allocation and pricing of PG affect the expected revenue from future RTB campaigns. Several assumptions are made on media buyers' behavior, such as risk aversion, stochastic demand arrivals, and effects of time and guaranteed contract price. We use dynamic programming to solve the optimization problem and our solution is relatively scalable and efficient. We validate the proposed model with an RTB dataset and find it increases the seller's expected total revenue by adopting different pricing and allocation strategies according to the level of competition in future RTB campaigns.
READ FULL TEXT
