Does Collaborative Filtering Technology Impact Sales? Empirical Evidence from Amazon.Com

Abstract

This paper investigates the impact of collaborative-filtering-based recommendation systems and consumer feedback mechanisms, two technologies commonly adopted by online retailers, on sales. With data from Amazon.com, we find evidence that more collaborative-filtering-based recommendations are associated with higher sales at Amazon.com, while sales also contributes to recommendations. On the other hand, consumer ratings do not directly correlate to sales. Instead, high ratings that are validated by a large number of reviews have a positive effect on sales. Furthermore, we find that the effects of collaborative-filtering-based recommendations on sales are stronger for books that have fewer reviews and are less popular. Because less-reviewed, and less-popular books generally have less information available and are usually less accessible to buyers when compared to well-reviewed best sellers, recommendations based on collaborative filtering technology may be a more important source of information for buyers. This finding suggests that collaborative filtering technology has the potential to make up for the inefficiency of consumer feedback mechanisms by supplementing information for consumers to make decisions. Our findings also suggest that an efficient and effective selling strategy is to promote early sales, especially to customers with a long history of book buying because sales to someone with no history provides little data for collaborative filtering. Overall, our results provide important strategic implications for vendor strategies to design and use these technologies.