Dirichlet implications for portfolio management

Abstract

AbstractWe extend the utility of Goodhardt's negative binomial distribution‐Dirichlet model to demonstrate how it can be used to support portfolio decisions relating to retail assortments. The approach is based on analysing polarisation of loyalty, , obtained via a transformation of the model's S parameter, to investigate loyalty to attributes. Prior research on loyalty to attributes suffers from limitations of method or application, so we develop a full treatment of the statistical basis and analytical methods required for ongoing research in this area. We then demonstrate the utility of our approach through application to four subcategories of coffee in the USA over 2 years. Importantly, we reveal the specific attribute levels (features) that generate the least and most loyalty, and show how this, when combined with market share, can be used to respond to the strategic context. Specifically, those attribute levels with low loyalty may be more suitable for seasonal or limited time offers for which maximising short‐term custom is more important than repeat purchase; conversely, attribute levels with high loyalty may be more suitable for achieving long‐term growth of brand share. In both cases, attribute levels with higher market share are preferable.