A Brand-Switching Model With Implications for Marketing Strategies
Jan 1, 1987
Use of switching data to gauge performance of different marketing strategies.
Type of Report
Provides general background on use of switching data; proposes a modeling technique that extracts much information about loyal customers and potential switchers at relatively low cost.
To arm marketing managers with a measurement tool that gives them additional insights from data already in hand. The technique provides an inexpensive “second opinion” of traditional attitude tracking measures; it furthermore provides a more sophisticated picture of a brand’s consumers than can ordinarily be “read” from switching matrices.
Develops a two-class latent model for brand-switching. The two classes are Hard-Core Loyals and Potential Switchers. Provides an example of the model in use by presenting an interpretation of previously unpublished automobile data. Additionally, data from a project using frequently purchased packaged goods, already described in the literature (Zufryden, 1986), are re-analyzed. Further applications are suggested for managerial implementation.
Researchers and analysts with technical proficiency should find the method and its discussion stimulating; non-technicians should be interested in the implications.
Customarily, researchers confronted with a switching matrix assume that the customers represented in it are homogeneous–that one is as likely as another to be a brand switcher. This traditional assumption, among other things, has led to a common practice of “eyeballing” switching matrices to read some obvious patterns.
This practice, although hallowed by custom, is potentially misleading. In the real world, some consumers are predisposed to purchase the same brand they bought previously, while others are more susceptible to brand-switching. If a brand’s customers could be modeled into even these two highly simplified categories, customary analysis of readily available data could help reveal whether current marketing strategies are having the desired effect.
The paper describes a model accommodating only these two classes of consumers: Hard-Core Loyals and Potential Switchers. The probabilities of brand purchase are given for each.
By allowing for only these two classes of consumers, it is possible to “read” from successive switching matrices just how successful a given brand is in winning Potential Switchers from its competing brands–how well it “conquests” its competition.
Estimating such a model is not a trivial undertaking. But it can be done with readily available data and requires less “number-crunching” than some measures of purchase activity. Indeed, the method can provide an inexpensive gauge of convergent validity for analyses done with “better” loyalty measures.
Although switching data are most often collected for frequently purchased packaged goods, the Hard-Core Loyal/ Potential Switcher model is at least as relevant for durables. For example, data such as “which brand of dishwasher did you buy and what brand did you previously own?” are perfect input for the model. Similarly, asking an investor his most used brokerage firm and his most-used firm a year ago will give valid input for the model.
The paper describes the development of an unobservable latent model to explain an observable switching matrix. As such, the technique is not at all new; it has been employed in social sciences for a number of years. However, latent class models of this type have not been widely used in marketing to date. This model can introduce marketers to the technique with relative ease.
In particular, managers should benefit from using the model whenever behavioral switching constructs are meaningful inputs for the problem under consideration, data are available which appropriately measure repeat purchase and switching probabilities, and the questions to be studied can be cast in terms of “loyalty” versus “conquesting.”
This model partially addresses the whole strategic issue of niching. The switching matrices analyzed by the Hard-Core Loyal/Potential Switcher model will indicate whether the brand of interest has a small but highly loyal following (e.g., it is a niche brand), or a wider audience of consumers who buy the brand on occasion (i.e., perhaps it is a variety seeking brand).
The authors offer the following observations on the research:
“The modeling technique accomplishes two very important things at once: (1) it provides a more-sophisticated analysis of the problem using data already in hand (usually), and (2) it costs very little to implement. Managers should be attracted to the idea that this is a way of working ‘smarter, not harder.’
“This Hard-Core Loyal/Potential Switcher model is a special case of the well-known Mover-Stayer model used by sociologists to investigate labor mobility. To estimate the Mover-Stayer model, three consecutive purchases are needed (the HCL/PS model needs just two). Clearly, this is not an onerous data requirement.
Future research should take the HCL/PS model to the next logical step. These types of discrete latent class models have been used in marketing, but not nearly so often as their continuous latent variable counterparts such as the beta-binomial and the negative binomial. The managerial implications are often more direct and easier to communicate for the discrete models. We hope this current effort will stimulate additional work that better addresses important managerial issues such as loyalty, conquesting, niching, and variety-seeking behavior.”
About the Authors
Richard Colombo is Assistant Professor of Marketing, New York University. Donald G. Morrison is William E. Leonhard Professor, John E. Anderson Graduate School of Management, the University of California, Los Angeles.
This resource is not available online.