How Do Recommender Systems Benefit Online Retailers in the Long Run? Evidence from a Field Experiment

Résumé fourni par la source

Previous research on recommender systems primarily focuses on their short-term effects on customer search and purchase behaviors. This study investigates the effect of a recommender system on customer loyalty and long-run retail sales using a randomized field experiment. We manipulated the presence vs. absence of product recommendations from an item-based collaborative filtering recommender system at an online retailer. The results reveal that displaying recommendations increases consumers’ purchases of recommended products but at the cost of reduced sales of non-recommended ones, which may not increase total sales in the short run. In spite of this, consumers’ shift of focus (to recommended products) due to the presence of recommendations plays an important role in enhancing customer loyalty. When recommendations are disabled, long-term sales decrease significantly due to reduced customer loyalty. The results show that the loyalty effect is primarily driven by a preference effect, in which displaying recommendations induces consumers to view more recommended products, enhancing their shopping experience and increasing customer returns. For returning visitors, who may deem product recommendations as built-in elements of a desirable store environment, a one-time disablement of recommendations can directly lead to defection. The findings provide insights on the true value of recommender systems.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.