Monetizing Rating Scales

• Activity Number: 192
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Marketing
• Author(s): Nino Hardt*+ and Greg M. Allenby and Alex Varbanov
• Companies: The Ohio State University, Fisher College of Business and The Ohio State University and The Procter & Gamble Company
• Keywords: Scale Usage Heterogeneity ; Customer Satisfaction ; Bayesian Estimation

Abstract

Determining the value of product attributes is difficult in real-world studies because of the presence of many product attributes and the need to get respondents to provide evaluations on a common scale. A common scale is needed whenever data are combined across respondents to conduct inference. Moreover, it is common to find a large number of product attributes in commercial studies, making techniques such as conjoint analysis difficult to implement. Difficulties also arise when attributes are not directly related to the physical formulation of the brand, but are instead related to psychological and experiential features that do not lend themselves to experimental manipulation. In this paper we develop a method for monetizing scales for product analysis and demonstrate its application in a large-scale brand positioning study of a packaged goods manufacturer. The adjusted ratings data are shown to increase the accuracy of purchase predictions by about 20% relative to existing methods.