The Tale of the Tail: Inference for Customer Purchase Behavior in the Long Tail

39 Pages Posted: 3 Sep 2021

See all articles by Bruno Jacobs

Bruno Jacobs

University of Maryland - Robert H. Smith School of Business

Didier Nibbering

Monash University - Department of Econometrics and Business Statistics

Date Written: September 1, 2021

Abstract

A large product assortment is typically characterized by many products that are rarely purchased: the long tail. Combined, these products make a sizable contribution to the total purchase volume. A retailer that better understands the purchase behavior in its long tail can increase the value of these products. Yet, analyzing tail purchases at the customer level is challenging: The available purchases per product in the tail are limited, while the number of customers and products are large. We develop new methodology that overcomes these challenges and sheds light on customer-specific purchase behavior for the long tail. The idea underlying our approach is a dimension reduction that uses latent product groups to summarize tail purchase behavior. We rely on variational inference to apply our method to a large-scale purchase history dataset with almost 50,000 products and over 3 million shopping trips. We are able to identify the customers that are likely to purchase in the tail of the assortment, how this varies across product categories, and how tail purchases relate to purchases of other products in the assortment. These insights can be used to improve recommendations of tail products, facilitate navigation through the assortment, and inform assortment management decisions.

Keywords: long tail, large-scale assortment, purchase history data, dimension reduction, latent product groups, variational inference

JEL Classification: M31, C55, C11

Suggested Citation

Jacobs, Bruno and Nibbering, Didier, The Tale of the Tail: Inference for Customer Purchase Behavior in the Long Tail (September 1, 2021). Available at SSRN: https://ssrn.com/abstract=3915874 or http://dx.doi.org/10.2139/ssrn.3915874

Bruno Jacobs (Contact Author)

University of Maryland - Robert H. Smith School of Business ( email )

College Park, MD 20742-1815
United States

Didier Nibbering

Monash University - Department of Econometrics and Business Statistics ( email )

900 Dandenong Road
Caulfield East, 3145
Australia

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
206
Abstract Views
754
Rank
269,995
PlumX Metrics