JCPENNY CENTER FOR RETAIL EXCELLENCE

Research

JCPenney Center for Retail Excellence
RESEARCH AGENDA

LOYALTY AND STORE SWITCHING

Why do shoppers switch stores?
"Store Switching - Or Is It?"
Edward J. Fox and John Semple

Store choice data show that households seldom shop exclusively at a single retail store, but generally switch among stores. The purpose of this research is to investigate the factors that cause store switching. More generally, we determine how much of cross-shopping behavior is actually due to switching, and how much is driven by consumer's shopping strategies (e.g., cherry-picking). Using a random utility framework, we model switching as a function of (1) higher-order strategic behavior and (2) factors that change over time, including communicated retailer prices (advertising), shoppers' product needs, as well as opportunities to "chain," or link shopping trips together. In addition, we identify segments of shoppers that switch stores for different reasons. The resulting model has a hierarchical multivariate probit specification (Manchanda, et.al., 1999) and is estimated in a panel dataset for five grocery store chains.

How can a retailer identify its loyal customers using loyalty card data?
"Inferring Shopper Loyalty from Retail Loyalty Card Data"
Edward J. Fox and Jacqueline Thomas

There is little question that shopper loyalty is a critical driver of retailer success. Many retailers now gather detailed information on their customers using loyalty cards, in the hope of targeting offers to individual households. Such targeted offers would be made with the objective of retaining loyal customers, or increasing the retailer's "share of wallet" for customers who shop multiple stores.

This research lays out a methodology for using the retailer's loyalty card data, calibrated with syndicated panel data, to determine the loyalty of individual customers. Specifically, is a given household loyal to one retailer, or do they switch stores? This methodology involves (1) modeling household spending across stores and formats using the syndicated panel data, (2) applying those model estimates to the retailer's loyalty card data to predict spending across store chains, then (3) classifying shoppers as "loyals" or "non-loyals" using those spending predictions. We find that optimum loyalty classification criteria depend upon the offer under consideration. We then demonstrate how to maximize the profitability of different offers by modifying the loyalty classification criteria accordingly.

What makes shoppers loyal to a store chain or brand?
"Modeling Loyalty: A Share of Requirements Approach"
Edward J. Fox and Alan L. Montgomery

A commonly accepted definition of loyalty is based on the share of requirements of a type of product or service that a consumer meets with a particular brand. Surprisingly, loyalty as a share of requirements has not previously been modeled.

Cooper and Nakanishi (1983, 1985) developed a set of log-linear regression techniques for the analysis of market share data. These techniques apply for market shares on the unit interval, but are not appropriate for shares at the boundary points of 0 or 1. However, share of requirements data frequently include zero observations, indicating that the brand was not purchased. The authors propose a methodology for computing unbiased and consistent estimates for log-linear regression models when zero-shares are observed. A random utility framework is developed in which the utility must exceed a latent threshold value in order to generate purchases. The utility threshold, latent utilities, and parameter values are estimated using a Gibbs Sampling algorithm. We test the proposed model against other approaches to modeling zero-shares using store loyalty data. We also assess the biases inherent in these other approaches.