JOHN R. HAUSER and FRANK S. KOPPELMAN * Perceptual mapping is widely used in marketing to analyze market structure, design new products, end develop advertising strategies. This article presents theoretical arguments and empirical evidence which suggest that factor analysis is superior to discriminant analysis and similarity scaling with respect to predictive ability, managerial interpretability, and ease of use. Alternative Perceptual Mapping Techniques: Relative Accuracy and Usefulness Perceptual mapping has been used extensively in ty, and direct marketing strategy to the areas of marketing. This powerful technique is used in new investigation most hkely to appeal to consumers. product design, advertising, retail location, and many Perceptual mapping has received much attention in other marketing applications where the manager wants the literature. Though varied in scope and application, to know (1) the basic cognitive dimensions consumers this attention has been focused on refinements of the use to evaluate "products" in the category being techniques, comparison of alternative ways to use the investigated and (2) the relative "positions" of present techniques, or application of the techniques to market- and potential products with respect to those dimen- ing problems.' Few direct comparisons have been sions. For example. Green and Wind (1973) use simi- made of the three major techniques—similarity scal- larity scaling to identify the basic dimensions used ing, factor analysis, and discriminant analysis. In fact, in conjoint analysis. Pessemier (1977) applies discri- most of the interest has been in similarity scaling minant analysis to produce the joint-space maps that because of the assumption that similarity measures are used in his DESIGNR model for new product are more accurate measures of perception than direct design. Hauser and Urban (1977) use factor analysis attribute ratings despite the fact that similarity tech- to identify consumer perceptions and irmovation niques are more difficult and more expensive to use opportunities in their method for modeling consumer than factor or discriminant analyses. response to innovation. All of these researchers report In practice, a market researcher has neither the empirical applications in a number of product and time nor the money to simultaneously apply all three service categories. When used correctly perceptual techniques. He/she usually selects one method and mapping can identify opportunities, enhance creativi- uses it to address a particular marketing problem. The market researcher must decide whether the added insight from similarity scaling is worth the added *John R. Hauser is Assistant Professor of Marketing. Graduate expense in data collection and analysis. Furthermore, School of Management, and Frank S. Koppelman is Associate if the market researcher selects an attribute-based Professor of Civil Engineering and Transportation, Technological method such as factor analysis or discriminant analysis Institute, Northwestern University. The authors thank Bruce Bagamery. Yosy Prashker. Steve Shu- he/she wants to know which method is better for gan. and Ken Wisniewski for their extensive technical assistance. perceptual mapping and how such maps compare with Richard Johnson, Dennis Gensch, William Moore, Edgar Pessemier, Glen Urban, Paul Green. Allan Shocker, Subrala Sen, Brian Sternthal. Lou Stern, Steven Lerman, and Philip Kotler provided See for example Best (1976), Cort and Dominguez (1977), Day. useful comments, criticisms, and suggestions. Deutscher. and Ryans (1976), Etgar (1977), Gensch and Golob (1975), The data for the study were collected under a University Research Green and Carmone (1969). Green. Carmone, and Wachspress Grant, DOT-OS-4001, from ihe US. Department of Transportation, (!977), Green and Rao (1972), Green and Wind (1973), Hauser Peter Stopher and Peter Watson, principal investigators. The authors and Urban (1977), Heeler. Whippie, and Hustad (1977), Jain and Ihank ihe Transportation Center at Northwestern University for Pinson (1976). Johnson (1970, 1971), Pessemier (1977), Singston the use of these data. (1975), Summers and MacKay (1976), Urban (1975), and Whippie (1976). 495 Journal of Marketing Research Vol. XVI (November 1979), 495-506 JOURNAL OF MARKETING RESEARCH, NOVEMBER 1979 496 those from similarity scaling. To answer these ques- description, we have indicated the appropriate refer- tions, one must compare the alternative mapping ences. techniques. Similarity scaling develops measures of consumers One way to compare these techniques is theoreti- perceptions from consumer judgments with respect cally. Each technique has theoretical strengths and to the relative similarity between pairs of products. weaknesses and the choice of technique depends on Though consumers are asked to judge similarity be- how consumers actually react to the alternative mea- tween products, the definition of similarity usually is left unspecified. The statistical techniques select surement tasks. Theoretical hypotheses are established relative values for two, three, or four perceptual in the next section. dimensions such that distance between products best Another comparison approach is Monte Carlo simu- corresponds to measured similarity. Green and Rao lation. This useful investigative tool has been employed (1972) and Green and Wind (1973) provide mathemati- by researchers to explore variations in similarity scal- cal details. ing and other techniques (Carmone, Green and Jain For empirical studies with large sample sizes, com- 1978; Cattin and Wittink 1976; Pekelman and Sen 1977). mon space representations are developed such that In perceptual mapping Monte Carlo simulation can each consumer's / perception x,^j of product J along compare the ability of various techniques to reproduce dimension d is a. "stretching" of the common repre- a hypothesized perceptual map, but it requires that sentation x,.^. The technique, INDSCAL (Carroll and the researcher assume a basic cognitive structure of Chang 1970), simultaneously estimates the common the individual. Monte Carlo simulation leaves unan- space Xj^ for all / and estimates individual weights swered the empirical question of whether the analytic v,.j to "stretch" these dimensions. Effectively, i,^^ technique can adequately describe and predict an =' v]d^ ^ d- For details see Carroll and Chang (1970). actual consumer's cognitive structure. The dimensions are named by judgment or by a The comparison procedure we use is practical and regression-Uke procedure called PROFIT (Carroll and is based on theoretical arguments and empirical analy- Chang 1964). In PROFIT, consumers are asked to ses to identify which procedures yield results most rate each product on specific attributes, e.g., "atmo- useful for marketing research decisions. If the theoret- sphere" for shopping centers. The ratings are depen- ical arguments are supported, researchers can continue dent variables in a regression (possibly monotonic) to subject a technique to empirical tests in alternative on the perception measures, i,^^, which serve as product categories. In this way, one gains insight about explanatory variables. The regression weights, called the techniques by learning their strengths and weak- directional cosines, indicate how strongly each per- nesses. If and when the hypotheses are falsified new ception measure relates to each attribute rating. For theories will emerge. details see Carroll and Chang (1964). We chose the following guidelines for the compari- Factor analysis begins with the attribute ratings. son. The assumption is that there are really a few basic 1. The marketing research environment should be perceptual dimensions, x,^.^. Many of the attribute representative of the way the techniques are used empirically. ratings are related to each perceptual dimension. 2. The sample size and data collection should be large Factor analysis examines the correlations among the enough to avoid exploiting random occurrences and attributes to identify these basic dimensions. For should have no relative bias in favor of the tech- statistical details see Harmon (1967) and Rummel niques identified as superior. (1970). Because concern is with the basic structure 3. The use of the techniques should parallel as closely of perception, the correlations between attribute rat- as possible the recommended and common usage. ings are computed across products and consumers 4. The criteria of evaluation should have managerial (sum over / andy). The perceptions of products are and research relevance. measured by "factor scores" which are based on the To fulfill these criteria, we chose an estimation attribute ratings. The dimensions are named by exam- sample and a saved data sample of 500 consumers ining "factor loadings" which are estimates of the each, drawn from residents of Chicago's northern correlations between attribute ratings and perception suburbs. The application area is perceptions of the measures. In applications, attribute ratings are first attractiveness of shopping areas in the northern sub- standardized by individual to minimize scale bias. urbs and the criteria are the abihty to predict con- Discriminant analysis also begins with the attribute sumer preference and choice, interpretability of the ratings, but rather than examining the structure of solutions, and ease of use. attribute correlations, discriminant analysis selects the (linear) combinations of attributes that best discrimi- HYPOTHESES: THEORETICAL COMPARISON nate between