A Fuzzy Set Model of Consideration Set Formation Calibrated on Data from an Online Supermarket
Authors: Jianan Wu, Arvind Rangaswamy
Choose brands using a decision process that is dynamic, and sometimes complex. Generally stated, this decision process involves consumers searching for information, eliminating unacceptable alternatives, and making a final choice from among the acceptable alternatives. To represent this process analytically, researchers have developed several stylized two-stage models: 1) a consideration set stage where the consumer first forms a set of brands that she would consider buying, and 2) a choice stage in which she evaluates each alternative in the consideration set to select the best alternative (e.g., Fortheringham 1988, Roberts and Lattin 1991, Andrews and Srinivasan 1995, Siddarth, Bucklin, and Morrison 1995, Bronnenberg and Vanhonacker 1996).
The existing two-stage choice models have contributed significantly to our understanding of how marketing mix elements influence consideration set formation. However, these models have been silent on how consumer-search influences consideration set formation. In this paper, we offer an axiomatic framework to address this issue. To parsimoniously represent the dynamic nature of the consumer search and consideration process, we use a fuzzy set modeling approach in which the degree of consideration of various brands can change with additional information that consumers obtain through search. Specifically, we propose that as consumers gain more information about the alternatives, it reduces fuzziness, i.e., the consideration sets become crisper. We show that the proposed model is a generalization of several existing two-stage choice models.
To test the validity and value of our model, we operationalize it in the context of data obtained from Peapod, an online grocery store. In Peapod, we can track consumers' search process, and therefore, are able to investigate the impact of consumer search on consideration set formation. For our example application, we use data from the liquid detergent category. We also develop a latent mixture specification to allow for the possibility that different segments of consumers may have different search behaviors and response sensitivities to marketing instruments. We demonstrate that our model performs better than competing models on several criteria (e.g., model fit, predicted choice probability), even when all the models use identical information sets.
Our results suggest that consumers not only rely on their internal memory search, but also sometimes engage in external information search to reduce the fuzziness of their consideration sets. We also find there is heterogeneity in consumers’ capability to process external information. For some consumers, searching the external information (in the online store) dramatically increases the sizes of their consideration sets, but for others, external search does not impact the consideration set much at all. Thus, online stores need to be designed using personalization technologies to offer superior shopping experiences to both types of customers.
The contributions of this paper are two-fold: 1) We conceptualize, axiomatize, and operationalize a model of consumer choice behavior, which is a generalization of several existing two-stage choice models. This model offers an improved approach to represent the process by which consumers make choices. 2) We empirically demonstrate that existing models are inadequate for fully capturing the richness of the choice processes that are increasingly becoming feasible to observe in online markets. Models, such as the ones proposed here, are necessary for deriving managerially relevant understanding of choice behavior in online markets.
