Bayesian Applications for Marketing

• Activity Number: 229
• Type: Invited
• Date/Time: Tuesday, August 5, 2003 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and Marketing
• Author(s): David Bakken*+
• Companies: Harris Interactive
• Keywords: marketing ; Bayesian ; conjoint

Abstract

Statistical methods have been applied to a wide variety of marketing problems. For the most part, statistical methods are limited to the post hoc analysis of data. Recently, however, the use of Bayesian methods has led to greater integration of statistics and problem formulation. Despite the expanding number of applications for Bayesian estimation methods, only one application has made significant inroads into the practitioner world. This is the use of hierarchical Bayesian estimation of stated preference discrete choice models. Practitioner adoption of new statistical methods is determined by the extent that they improve our ability to reduce uncertainty around the loss functions for different managerial choices. Bayesian methods can reduce this uncertainty in two important ways. First, the methods capture unobserved heterogeneity, which in turn leads to models that more accurately reflect consumer responses in the market. Second, these methods have the potential to reduce uncertainty arising from survey nonresponse. This paper describes a few common marketing problems with case study examples of Bayesian methods applied to those problems.