Position Auctions for Sponsored Search in Marketplaces

In recent years, online shopping has become increasingly concentrated in marketplaces, which now have a similar number of daily users as searching platforms and social networks. As a result, marketplaces have started adopting monetization mechanics from search platforms, with sponsored search being the main example. However, compared to searching platforms, where “search” is the main source of income and the only decision point (to show sponsored results or not), marketplaces have more complex mechanics. Each seller has at least two decisions that affect marketplace income: what price to set for their goods and how much to spend on search boost. This complexity makes the development of position auctions for sponsored searches more challenging. Currently, most marketplaces still use auctions that were designed for searching platforms, like VCG and GSP auctions. In this work, we consider the properties of VCG and GSP auctions for a simplified marketplace model. In the marketplace setting, those auctions are well studied, but all previous works focused on auction fairness, search costs, and so on, which are usually important in the late stages, i.e., when companies already use auctions and want to make them work better. On the other hand, there is a lack of research on possible upper and lower bounds on the revenue of auctions for marketplaces when information is limited. The question “how much revenue can it generate?” is critical in the early stages of auction implementation. In this work, we try to answer this question and get upper and lower bounds on revenue generated by VCG and GSP auctions for marketplaces.