Economics of Recommender Systems in Online Marketplaces

Abstract

Electronic marketplaces such as Amazon Marketplace and Ebay deploy recommender systems as sales support tools to help consumers a nd their ideal product among the vast variety of products sold in these platforms. Recommender systems affect consumer decision making by informing consumers about products they may not be aware of and enlarging the consumers' consideration set ("informative role"). In this dissertation, we study the impacts of recommender systems on different players in an online channel structure where a dominant e-commerce platform sells competing products from different manufacturers, and simultaneously recommends a subset of these products to consumers. The dissertation consists of three essays. The rest essay highlights how recommender system design affects the upstream competition between manufacturers and the consequent implications for the recommendation strategy to be adopted by the retail platform. In our setting, consumers are differentiated with respect to their preference for the two products and awareness about the two products. A recommender system is designed to select the recommendation based on a weighted sum of expected retailer pro t and expected consumer value. We fi nd that the recommender system may benefit or hurt the retailer and manufacturers, depending on its design and market characteristics. We show that the retailer's optimal recommendation strategy is mildly pro t oriented in the sense that it assigns a larger but not too-large a weight to retailer pro t than consumer value, and that under the optimal strategy, the price competition is less intense and the retailer pro t is higher compared to when there is no recommender system. The informative role of the recommender system deployed by an electronic marketplace functions as a medium for targeted advertising for sellers, analogous to traditional advertising media such as TV, newspaper, and the Internet. In the second essay, we examine how a recommender system affects competing sellers in electronic marketplaces regarding their advertising and pricing decisions. We nd that sellers advertise less (advertising effect) on their own and decreases product prices (competition effect) in the presence of a recommender system. As a result of these two effects, sellers are more likely to benefi t from the recommender system only when it has a high precision. While the first two essays consider a recommender system that recommends competing products to help consumers fi nd a better alternative, the third essay considers a mixed recommender system where both competing products and complementary products are included. In particular, we consider four sellers that sell four products in two categories via a common retail platform. Products in the same category are substitutes, and products in different categories are complements. We show that the recommender system does not necessarily benefi t the marketplace; on one side, the recommender system increases the total sales, but on the other hand, the recommender system alters the competition in each category. In the presence of the recommender system, the price and profi t of each seller critically depends on the degree of complementary among products.