Competitive Positioning Analysis with Learning Vector Quantization

Abstract

Positioning analysis aims at exploring the perceptual differences between the brands of a product class and the way they are linked to consumer preferences. Various sophisticated multivariate techniques have been suggested to analyze consumer perceptions and preferences and extract competitive dimensions. In this study, a completely different reduced space method based on Learning Vector Quantization is suggested for positioning analysis. The method of Learning Vector Quantization, belonging to the field of computational intelligence, is a special case of an artificial neural network that applies a supervised competitive learning-based approach. In comparison to the current multivariate techniques, the advantages of the proposed method can be listed as follows: (i) it allows to deliver decision support for positioning by examining multiple relationships simultaneously, (ii) its procedures are easy to implement, (iii) it does not impose rigorous distributional assumptions, (iv) nor does it require particular scaling properties of the raw data, (v) it effectively copes with samples of limited and unlimited size, (vi) it can be used in online mode, and finally (vii) it allows what-if simulations and predictions for new customers. The study outlines some major aspects of the methodical foundations of the LVQ-based positioning analysis and provides an illustrative example.