Recommendations Systems: Beyond Matching Products to Buyers

Abstract

The digital revolution has allowed sellers to make large assortments of products available to consumers. Recommendation systems have played a central role in this dynamic. At the core of these systems is the use of data and sophisticated algorithms to predict match values between products and buyers.

By analyzing consumer search data and product recommendations of an online used car seller, we find that there is scope for value creation by recommendation systems beyond their primary matching role. More specifically, our analysis leverages search consumption: The fact that consumers enjoy inspecting at least some of the products on sale. We identify an engagement effect such that recommending some products with high hedonic value induces additional customer engagement while keeping baseline conversion rates unchanged. The engagement effect is economically significant in our data: It explains 55% of the potential value available to recommendation systems, the remaining 45% made up by the traditional product matching mechanism.

JEL Classification: D83, L86, M31