Knowledge that Transforms

Commentary to Farris, P., J. Olver, C. De Kluyver. 1989. The relationship between distribution and market share. Marketing Sci. 8 107–128.

If buyers face transaction costs, a workable pricing scheme might require door prizes in addition to a per-unit price. The door prize compensates customers for the transaction expense, and the seller earns positive profit because the price exceeds the cost of inframarginal units.

Researchers have used a variety of measures for the assessment of conjoint analysis reliability. In this note we critically evaluate two different measures and identify potential shortcomings for each one. We identify the context for which a given measure is suitable, and recommend the use of multiple measures for other cases.

Factor Analysis is commonly used to reveal the structure underlying the multiple attributes that describe marketing stimuli. This paper provides a theoretical investigation and an empirical comparison of two approaches to Factor Analysis that are based on two different sources of variation in the input data. The “Among” Analysis is based on variation across marketing stimuli, while the “Within” Analysis is based on variation across individuals responding to the same marketing stimulus. We identify conditions under which one approach is superior to the other. If both approaches are applicable, we recommend the use of the “Total” analysis that pools the variation across stimuli and individuals. An empirical study, in the context of consumers' cognitive response to ads, shows that the Among Analysis results can be seriously distorted by differential familiarity with the ads so that it is important to partial out the spurious effects of familiarity.

In the food industry, taste tests are an important element in product development, test marketing and advertising. The primary factor detracting from the reliability of taste tests is guessing by subjects; it confounds true discrimination and preference with random responses. In this paper simple modifications of discrimination and preference tests are suggested. Easily implemented, these changes do not increase the costs of the test while greatly reducing the probability of a subject guessing correctly. Two experiments to check the effect of these changes on task difficulty were carried out. Results show that there may be some increase in respondent's task difficulty, but this increase is overshadowed by the increased reliability of the suggested test design.

The expectation formation process has been extensively studied by economists, particularly for macroeconomic variables. However, no prior research has examined how expectations are formed and used in the context of marketing forecasting. In this paper, we use data from a simulated competitive environment (1) to examine the expectation formation process for important marketing variables such as market size, number of competing products, and average industry price, and (2) determine how inter-firm variation in the formation process affects performance. Using the Rational Expectations Hypothesis as a framework, we find that decision-makers' forecasts tend to be efficient, i.e., utilize all relevant available information, but are biased. Other key findings suggest that degree of firm rationality in the forecasting process as well as the level of forecasting accuracy are positively related to performance.

This note derives the Colombo/Morrison ‘hard-core loyal/potential switcher’ model without assuming the existence of either hard-core loyals or potential switchers. We only assume that the relative probability of buying a product other than what one purchased previously is independent of this previous purchase. This result significantly extends the applicability of the Colombo/Morrison model.

Two issues that have implications across a firm's marketing activities are the estimation of the market response functions, and the determination of the structure of the market in which its products compete. In a recent paper, Clements and Selvanathan (Clements, K. W., E. A. Selvanathan. 1988. The Rotterdam demand model and its application to marketing. Marketing Sci. 7(Winter) 60–75.) describe how the Rotterdam Demand System can be used to study these two issues in a joint framework. In this note we extend their model to test hierarchical market structures using data at the brand/form level. We also incorporate idiosyncratic response coefficients that capture the effects of advertising and/or other nonprice marketing mix variables separate from the effects of a price change. These extensions provide additional insights into the nature and extent of competition among brands within a product class.

This paper presents Elimination-By-Cutoffs (EBC), an individual-level model of choice which derives choice probabilities from the location of choice alternatives in a perceptual space. EBC is a multidimensional operationalization of Tversky's (Tversky, A. 1972. Elimination by aspects: A theory of choice. Psych. Rev. 79(4) 281–299.) feature-based set-theoretic Elimination-By-Aspects (EBA) model of choice. It employs monotonic and noninteractive utilities of perceptual attributes in a sequential elimination strategy to determine choice. EBC like EBA is free from the property of independence from irrelevant alternatives. Unlike the independent multinomial logit model (IML), EBC takes into consideration the effect of product similarities on the computation of choice probabilities. An empirical application is also presented in this paper. The EBC model is parameterized from readily available data and is computationally tractable. A comparison between predictions from EBC and IML is drawn using a holdout sample. Approaches for assessing the implications of alternative product positioning strategies are also developed. Finally, extensions to nonmonotonic utility functions, interactive utility functions and other directions for further work with the EBC model are discussed.