David A. Schweidel, Peter S. Fader, & Eric T. Bradlow
Executive Summary
Service churn and retention remain central as constructs in marketing activities such as valuation of customers and resource allocation. Although extant research has linked several external factors (e.g., satisfaction, quality) to service retention, managers must often contend with situations in which they have limited information. In this research, the authors use subscription data from a major telecommunications provider to develop a statistical model of service churn without relying on any external factors—a limited information scenario that many firms regularly face. In their analysis, the authors consider a set of five different factors that may affect service retention: duration dependence, promotional activity, cross-cohort effects, calendar-time effects, and subscriber heterogeneity. The authors find that their framework yields accurate forecasts of the number of subscribers from each cohort who will discard service each month. They are able to predict the pattern of service churn for subscribers who already have service, as well as future subscribers who have not yet acquired service. Across seven different services offered by the service provider, they find that the inclusion of promotional activity always yields improved forecasts of retention behavior, whereas the inclusion of cross-cohort effects do not improve forecasts. A combination of subscriber heterogeneity, calendar-time effects, and duration dependence are necessary to yield the most accurate forecasts, but they find that the exact mix varies from service to service.
Furthermore, the authors demonstrate how a subscription can be valued within their framework, allowing it to be used for resource allocation decisions. They also highlight the importance of capturing the right mix of factors by computing the expected value of a subscription for every combination of factors and find considerable variance across model specifications, which can adversely affect managerial decisions, including promotional activity and resource allocation, when the wrong specification is chosen.
Biography
David A. Schweidel is currently Assistant Professor of Marketing at the University of Wisconsin–Madison. He earned a Bachelor of Arts in Mathematics in 2001 from the University of Pennsylvania and an AM in Statistics in 2004 and a PhD in 2006 from the Wharton School of the University of Pennsylvania. David has served as a consultant for companies in the pharmaceutical, telecommunications, and entertainment industries. His research interests include stochastic modeling for advertising and customer relationship management applications.
Peter S. Fader is the Frances and Pei-Yuan Chia Professor of Marketing at the Wharton School of the University of Pennsylvania. He is delighted to have another article in the Journal of Marketing—his first after an 18-year hiatus in the journal. In the interim, he has been busy playing with two kids, managing a popular Web site (www.coolnumbers.com), and trying to keep people in the Marketing Department at Wharton from taking themselves too seriously. He is pleased with his progress on all of these projects and hopes it will not be another 18 years before he is back in JM again.
Eric T. Bradlow is currently the K.P. Chao Professor, Professor of Marketing, Statistics, and Education, and Academic Director of the Wharton Small Business Development Center in the Wharton School of the University of Pennsylvania. He earned a Bachelor of Science in Economics from the Wharton School in 1988 and an AM in Mathematical Statistics in 1990 and a PhD in Mathematical Statistics in 1994 from Harvard University. He joined the Wharton faculty in 1996. Eric received the MBA Core Curriculum teaching award and Miller-Sherrerd MBA Core Teaching award in 1999, 2000, 2001, 2002, and 2007 and the 2003, 2004, 2005, 2006, and 2007 Excellence in Teaching Award. Eric was recently named a fellow of the American Statistical Association, is past chair of the American Statistical Association Section on Statistics in Marketing, is a statistical fellow of Bell Labs, and was named DuPont Corporation’s best young researcher in 1992. His research interests include Bayesian modeling, statistical computing, and developing new methodology for unique data structures with application to business problems.
Journal of Marketing, Vol. 72, No. 1, January 2008
View Table of Contents