Accommodating the Effects of Brand Unfamiliarity in the Multidimensional Scaling of Preference Data

This paper presents a multidimensional scaling (MDS) methodology (vector model) for the spatial analysis of preference data that explicitIy models the effects of unfamiliarity on evoked preferences. Our objective is to derive a joint space map of brand locations and consumer preference vectors that is free from potential distortion resulting from the analysis of preference data confounded with the effects of consumer-specific brand unfamiliarity. An application based on pre/'erence and familiarity ratings for ten luxury car models collected from 240 consumers who intended to buy a luxury car within a designated time frame is presented. The results are corupared with those obtained from MDPREF, a popular metric vector MDS model used for the scaling o~ preference data. tn particular, we find that the consumer preference vectors obtained from the proposed methodology are substantially different in orientation from those estimated by the MDPREF model. The implications of the methodology are discussed. Multidimensional scaling models for the analysis of preference data have been widely usedin marketing for more than two decades (Green and Carmone 1970). These models provide a spatial representation of stimulus (brand) locations as well as subjects (consumers), typically represented either as vectors or as ideal points. The input data are usually preference judgments from consumers for the various brands in the evaluation set. These consumers may not be equally familiar with the different brands in the set; indeed, they are likely to be more familiar with certain brands (e.g., those previously purchased) than others. Traditionally, MD$ models have not taken into account the consumers' degree of familiarity with the different brands. If, however, a consumer's preference judgment of a brand is influenced by the degree of familiarity with the brand and both intuition and formal evidence suggest that this is so then the joint space representing brancl locations and subject preferences derived from MDS preference models may re*The authors gratefully acknowledge helpful comments from the editor and two anonymous reviewers, and also from Michael D. Johnson and Robert J. Meyer. They also thank Michael Kusnick and Robert Kleinbaum for their assistance in conducting the survey. 86 RAB1KAR CHATTERJEE AND WAYNE S. DESARBO flect some confounding of preference and the degree of familiarity. Our objective is to propose a procedure that attempts to disentangle this confounding so that the effect of unfamiliarity is, in effect, "filtered out" of the derived joint space. Product familiarity is defined by Alba and Hutchinson (1987) as "the number of product-related experiences that have been accumulated by the consumer" (p. 411). Product related experiences are defined "at the most inclusive l e v e l . . . [to] include advertising exposures, information search, interactions with salespersons, choice and decision making, purchasing, and product usage in various situations." Our focus is on familiarity with a specific brand within a category; in out context, the degree of familiarity with a brand corresponds to the extent of knowledge the consumer has about that brand. We develop a methodology that seeks to derive a MDS joint space representation of brand locations and consumer preference vectors under the (unobserved) condition of complete familiarity of the consumers with all the brands in the evaluation set, while simultaneously estimating the effects of the degree of familiarity on brand preference for each consumer. In terms of input data, the procedure calls for each consumer in the sample to provide two ratings preference and familiarity for each brand in the evaluation set. Further, replications can be accommodated, so that the input data may be three-way (consumer x brand x replication) matrices of preference and familiarity measures. 2 0 u r methodology extends the basic vector preference model underlying MDPREF (Carroll and Chang 1964) by incorporating the impact of unfamiliarity on preference. Thus, MDPREF, which is orte of the most widely used MDS procedures for analysis of preference data, serves as a natural "baseline" modël a potential benchmark for evaluating our procedure. Among probabilistic MDS models that have been developed both for proximity data (Ramsay 1977; Takane 1978) and for preference data (De Soete and Carroll 1983; De Soete, Carroll and DeSarbo 1986; DeSarbo, De Soete and Jedidi 1987), those proposed by Zinnes and MacKay (1983, 1987; MacKay and Zinnes 1986; MacKay 1989) incorporate probabilistic stimulus locations. The distribution describing the brand locations is estimated from data consisting of either dissimilarity or preference ratio judgments between brand pairs. Brand unfamiliarity must be inferred from the estimates of the standard deviations of the brand coordinates. In contrast, our approach separates out the effect of unfamiliarity in deriving the joint space by using a direct measure of brand familiarity (along with preference ratings) to estimate a preference-familiarity relationship. We first present the model and describe the estimation procedure, followed by an actual application of the proposed methodology involving luxury car preference. We conclude with a summary of the implications of the procedure and a brief discussion of its limitations and directions for future research. BRAND UNFAMILIARITY IN THE SCALING OF PREFERENCE DATA 87