Customer portfolio analysis: Crisp classification versus fuzzy classification - Based on the supermarket industry

Abstract

Acknowledging that traditional matrix-form customer portfolio models that result in crisp clusters are clouded with ambiguity, we propose the use of fuzzy clustering in customer portfolio analysis. This has been done in order to assist managers in better understanding their overall customer portfolio and reducing the effect of descriptive indicators. Our approach is tested on a supermarket data set of 3076 customers and its results are compared with a conventional customer portfolio matrix. A qualitative and quantitative assessment of the categorization generated both by our fuzzy clustering approach and the conventional matrix-based crisp clustering has been carried out along the following parameters: substantiality and balance of portfolio. The results show that the use of fuzzy clustering yields more substantial clusters, as well as a more balanced portfolio of customers. Although a particular portfolio matrix has been chosen for this research, the approach proposed here could be modified for use with other portfolio matrices. © 2012 Macmillan Publishers Ltd.