Home | Business News | Browse by Publication | I | International Journal of Market Research

Combining revealed and stated preferences to forecast customer behaviour: three case studies.

Article Excerpt
Many companies collect stated preference (SP) data, such as intentions and satisfaction, as well as revealed preference (RP) data, such as actual purchasing behaviour. It seems relevant to examine the predictive usefulness of this information for future revealed preferences--that is, customer behaviour. In this paper we address this issue by considering three case studies. Our results indicate that adding SP data to RP data for predicting future customer behaviour does not result in better forecasts.

Introduction

At present, many companies collect stated preference (SP) data among large samples of their customers. Examples are annual surveys concerning stated loyalty, satisfaction and future needs. At the same time, many companies also keep track of actual behaviour of their customers by collecting revealed preference (RP) data in their customer databases. Examples of these data are records of past purchases, additional purchasing behaviour, and number of contacts with the company.

SP data sources provide the manager of a company with important metrics, such as brand preference and satisfaction (Kaplan & Nortan 1992; Wilson 2002). Moreover, SP data can also be used to reduce risk when making marketing decisions. Examples include research on new product development, ad messages, etc. In particular, from a Balanced Score Card perspective it is assumed that SP data on satisfaction and loyalty intentions should be considered as important indicators of future firm performance. Whether this is true and useful is a topic for ongoing research. For example, some managers complain that satisfied customers still defect (Jones & Sasser 1995). As the collection of SP and RP data is costly, natural questions to ask concern the predictive usefulness of current SP data for future revealed preferences and the potential relevance of past behaviour (measured by the RP data) for predicting future behaviour.

Recently, a number of studies have developed models that used these revealed preference data as well as stated preference data from a survey to describe customer behaviour (Bolton 1998; Bolton & Lemon 1999; Mittal & Kamakura 2001; Verhoef et al. 2001). In general these models show significant in-sample effects of the self-reported data on metrics, such as relationship duration, service usage and cross-buying. It is of course also interesting to examine how this in-sample fit carries over to out-of-sample forecasting. While there are several studies that investigate the forecasting performance of SP variables for sales at the aggregate or product level (Bemmaor 1995; Morwitz et al. 1997; Armstrong et al. 2000), the forecasting performance of stated preferences at the consumer level has almost been ignored. We are aware of only a single study that addresses a similar issue: Van der Poel et al. (2000) find little additional predictive value of SP data in a direct mail response model. In this paper, we aim to...

View this article FREE - Now for a Limited Time, try Goliath Business News
Free for 3 Days!