Knowledge that Transforms

A formal analysis is presented for three general classes of discrete attribute models of brand switching. The analysis focuses on the role of feature importance in specifying transition probabilities. A number of formal properties based on ordinal relations between transition probabilities are defined and each class of models is shown to satisfy a unique subset of the properties. The analytic results reveal important relations between the functional forms of model equations and the managerially-oriented interpretations that are given to the variables in those equations.

This paper presents a new ideal point probabilistic choice model. Unlike the model suggested by Cooper and Nakanishi (Cooper, L. G., M. Nakanishi. 1983. Two logit models for external analysis of preferences. Psychometrika 48 (4) 607–619.) which attempts to capture choices via a single ideal point, the proposed model, though based on aggregate data, allows for heterogeneity in preferences by estimating a distribution of ideal points. The model accounts for substitutability among choice alternatives and alleviates one of the major sources for the violation of the “Independence from Irrelevant Alternatives” property. It is demonstrated that the final form of the model is a Multinomial Probit, with a covariance matrix that depends on the relative position of the choice alternatives. An empirical application is provided and the resulting parameters are compared to the distributions of ideal points and attribute weights obtained via LINMAP (at the individual level) and via both the Logit and Probit versions of the model proposed by Cooper and Nakanishi (at the aggregate level).

Formulas are derived which estimate the accuracy of conjoint analysis in predicting preferences in a validation sample. This accuracy turns out to depend on (among other things) which model of preference is used (e.g., whether interactions are added, whether partworth or linear functions are used). I first show a paradoxical result that simpler models often yield higher predictive accuracy, even when a more complex model is the true one. The reason for this is that the additional parameters of the complex model are estimated with larger variance, which tends to overwhelm the benefits of using the true model. I then shift my criterion from predicting an individual's preferences, to predicting market share, which is of most interest to managers. Under this criterion my conclusions reverse, and I show that a true model (even when complex) is much more likely to yield higher predictive accuracy than a simpler incorrect model. This reverses some previous conclusions in marketing, and confirms that finding the correct model of consumer preference is important in improving prediction. Results from four previous empirical papers are correctly predicted by these formulas, as well as results from additional Monte Carlo studies.

This paper considers the problem of pricing a new product in a market having competing products of different qualities and market penetration levels, as measured by the cumulative number of units sold. Each customer type selects his optimal product based on maximizing consumer surplus. Pricing policies for a new product are determined for the seller based on cumulative profit maximization without discounting. An example is solved in detail for two demand function forms.

A probabilistic multidimensional scaling model that estimates both location and variance parameters for proximity and preference data is described and compared to a deterministic scaling model. Simulated and empirical choice data are used to compare models. Variance estimates from the probabilistic model are used to test a hypothesis about the homogeneity of stimulus perception under alternative modes of stimulus presentation.

This paper discusses the evolutionary development of an implemented, regression-based forecasting system used in planning and managing a schedule of performing arts events. In particular, the changing usage and refinement of this system is examined over a five-year period. One issue addressed is whether the manager, who can accept a regression forecast or revise it, is more accurate than the estimates produced by the regression model alone. The literature on bootstrapping and behavioral decision theory is used in examining the impact of managerial judgment on forecast accuracy. In addition, the paper presents a detailed review of the approaches used over time to revise the original forecasting system and an evaluation of their accuracy.

Theoretical models are developed for two strategies consumers use to choose among “noncomparable” alternatives. The models view consumers as trading off decision error and processing effort when selecting a decision strategy. The models predict the use of choice strategies, and implicit abstraction processes, for noncomparable alternatives in a number of different circumstances.

Household level panel data are the input for many types of marketing studies. An interesting, but until now unaddressed, question is what is the effect of aggregating individual members' purchases to the household level. Under “standard” assumptions the answer is unambiguous: the household brand switching pattern looks more zero order than the typical individual family member's behavior. When the standard assumptions are relaxed the overall spirit of the results do not change. These conclusions give comfort to those who use brand switching data to partition product categories. Those looking for variety-seeking behavior from household data are given some cause for concern—as well as reasons for reinterpreting previous studies.

This paper estimates a nested logit model for space heat and water heat choice using data from the 1980 Pacific Northwest Energy Survey (PNW). The estimated structure involves six alternative space heat systems and three alternative water heat fuel types. Operating and capital costs are predicted for each residence using an energy thermal model. To relax the Independence of Irrelevant Alternatives assumption we employ a family of nested logit models. We illustrate the two-step estimation method and use a Lagrange multiplier test for nested logit structure. Finally, we use the estimated models to forecast the market shares of alternative space heat systems under an assumed scenario for the course of relative fuel prices.

This paper is based on a series of studies undertaken for “V-8” Cocktail Vegetable Juice over a five-year period. The studies were a consequence of management questions regarding the effectiveness of a new advertising campaign, the best media mix for this campaign and the apparent “wear-out” of the advertising copy. The studies included controlled experimentation, estimation of advertising response functions, and exploration of price sensitivity following major price increases by “V-8.” It was found that the short-term response to advertising was S-shaped, and that the new creative approach rather than expenditure levels alone accounted for by far the largest component of the impact of advertising on sales. It was also found that for a short period following the major price increases, “V-8” became extremely price sensitive. However, after about six months, the marginal impact of price returned to historical levels. The application of results at Campbell Soup Company, their relationship to the literature and their implications for both researchers and practitioners are discussed.