Recommender System Rethink: Implications for an Electronic Marketplace with Competing Manufacturers

Recommender systems that inform consumers about their likely ideal product have become the cornerstone of eCommerce platforms that sell products from competing manufacturers. Using a model of an electronic marketplace in which two competing manufacturers produce substitutable products and sell the products through a common retail platform, we study the e ect of recommender systems on the marketplace, manufacturers, and consumers. Consumers are heterogeneous with respect to their preference for and awareness about the two products. The recommender system recommends the product that o ers a higher expected consumer net utility. Our analysis shows that the recommender system may bene t or hurt the retailer and the manufacturers, depending on the tradeo between two opposing e ects of recommender systems. The demand e ect of the recommender system increases the proportion of consumers that are aware of at least one of the products, thus increasing the overall market size and bene ting the sellers. The substitution e ect of the recommender system causes the demand function to be more price elastic through two forces one that shifts some consumers that are aware of a single product to ones that are aware of both products, thus making them comparison shoppers, and another that induces manufacturers to use price to gain informational advantage among consumers that are unaware of either product. The magnitudes of these two e ects depend on the relative sizes of consumer segments with di erent awareness levels and the recommender system precision. An increase in precision mitigates the substitution e ect of the recommender system. However, even perfect precision cannot guarantee that the recommender system bene ts the retailer and manufacturers. On the other hand, regardless of the precision, the recommender system bene ts the consumers and social welfare. We further show that by strategically committing to not recommend any product to some consumers those whose preference for a product is distorted by prices , the retailer can mitigate the substitution e ect and bene t more or lose less from the recommender system.