“Search Design on E-commerce Platforms” by Ms. Fei Long
Ms. Fei Long
PhD candidate in Decision, Risk & Operations
Graduate School of Business
Columbia University
E-commerce platforms, such as Amazon, Alibaba and Flipkart have transformed the retail sector by matching sellers and consumers at an unprecedented scale. These platforms operate their internal search engines to help buyers find relevant products from the large number of sellers, and also allow sellers to advertise to consumers through ad auctions for positions in the search listing. Determining an optimal ranking of products in response to a search is a difficult problem for the platform because the sellers have private information about products that the platform does not have. However, we note that the platform can use the sellers' bids in the ad auctions to obtain information about their types, which it can use to refine the ranking of products in the organic listing. While doing so may help consumers find an appropriate product match (information effect) thereby also helping sellers and the platform, it may negatively impact sellers' profits (competition effect) leading to reduced seller participation. Consequently, it is not clear how search engine design will affect the platform's advertising revenues, commission rates and sellers' pricing decisions, all of which directly affect the platform's performance. We explore how a platform's internal search engine should be designed by modeling this ecosystem using a Bayesian game with private information where a product's match with consumers is a priori only known to the seller. The platform's decision is whether to use this information to improve the ordering of its organic listings (strategic case) or to rank organic links independently from sellers' bids (independent case). While we find that while the strategic case tends to dominate, it also has a major impact on the platform's choice of commission as well as on sellers' pricing and participation. In particular, we find that strategic listing of organic links reduces price competition between sellers but increases advertising competition between them, thereby also reducing seller participation. These effects are stronger when the uncertainty of product fit with consumers is high. We also prescribe the optimal commission rate that a platform should charge by balancing between commissions and advertising revenue.