Reputation Effect, Business-stealing Effect, and Income Effect in Amazon’s Co-purchasing Product Networks

초록

This paper is an empirical examination of the reputation, business-stealing and income effects that pervade co-purchasing product networks. Co-purchasing between a pair of products refers to a situation where the two products are frequently purchased together (but not necessarily at the same time). Online retailers frequently use co-purchasing relationships as a basis for recommending products to their customers. (For example, Amazon has the section“Customers Who Bought This Item Also Bought” for its products.) How effective is this strategy? What features of co-purchasing product networks can be used to promote the sales of a given product? These are the empirical questions of our study. To this end, we use the data collected from Amazon.com, which has three major components: (1) sales ranks (as a measure of relative sales), (2) online reviews by customers (as a proxy for product quality), and (3) co-purchasing product networks (from “Customers Who Bought This Item Also Bought”). We construct two sets of features for co-purchasing networks: One contains the number, the average product rating, and the average sales rank, of co-purchased products, and the other contains the same variables for co-purchased products with a two degrees of separation (co-purchased products of co-purchased products). We hypothesize that there can be various effects of co-purchased products: They may enhance the reputation of a given product by association if they are well received by consumers (Reputation effect). They also compete with the given product by being substitutes (Business-stealing effect). Finally, these products pursue the same, limited consumer budgets (Income effect). Our results show that the reputation effect might be much smaller than the business-stealing effect of co-purchased products for co-purchased products with a two degrees of separation (co-purchased products of co-purchased products) than for co-purchased products with one degree of separation. Our results shed light on how to use co-purchasing information to design an effective product recommendation strategy.

키워드