An empirical investigation of user and system recommendations in e-commerce

Despite the popular use of user and system recommendations by online retailers to drive product sales in e-commerce, empirical research investigating their relative effectiveness and interactions still lags. We attempt to answer: (1) What is the relative impact of user and system recommendations on product sales in e-commerce? (2) Are user and system recommendations substitutes or complements in affecting product sales in e-commerce? Using data on the digital camera category from the largest business-to-consumer platform in China, Tmall.com, we use linear panel data models to examine the impact of user recommendation volume and valence, and system recommendation strength on product sales, controlling for relevant factors at the recommended product, product recommender, product category, and time unit levels. Importantly, we account for implicit sales correlations among products and potential simultaneity between recommendations and sales. We uncover several notable findings. Specifically, a 1% increase in the volume (valence) of user recommendations on a product increases the product's sales quantity by 0.013% (0.022%), whereas a 1% increase in the strength of system recommendations on a product increases the product's sales quantity by 0.006%. Overall, user recommendations are more effective than system recommendations in driving product sales. Furthermore, we find that there is a substitute relationship between user recommendation volume and system recommendation strength. Our findings provide important theoretical contributions and implications for recommendation-based product marketing and e-commerce platform design.