Uncovering Consumers' Choice Processes

The varied and qualitative nature of means-end decision process information makes it difficult to identify prototypical behaviours without imposing arbitrary constraints on the number and content of the steps these processes encompass. The proposed segmentation method (MPC) does not require such constraints, thus allowing the use of a wide range of data collection methods and means-end processes. First, a dynamic programming approach is used to partial out insignificant transitory effects and to translate means-end chains into vectors of a common probability space. Then, these vectors are clustered using a robust heuristic. The MPC method is illustrated and discussed with data initially published by Reynolds and Gutman (1988) and later used in other papers with different quantitative methods. MPC appears to discriminate processes whose paths are embedded, with two main results: the identification of dominant processes that do not necessarily end with a terminal value, thus contributing to the debate on goals and values, and the possibility to choose among chains of varied length for matching specific media strategies.