Comparison of different clustering methods for investigating individual differences using choice experiments

Abstract

Different strategies for investigating individual differences among consumers using choice experiments are compared. The paper is based on a consumer study of iced coffee in Norway. Consumers (n = 102) performed a choice task of twenty different iced coffee profiles varying in coffee type, production origin, calorie content and price following an orthogonal design. Consumer factors, such as socio-demographics, attitudes and habits, were also collected. Choice data will be analysed using two different clustering strategies. Strategy one is the most classical approach called Latent Class Logit (LCL) model, while Strategy two uses Mixed Logit (ML) model combined with Principal Component Analysis (PCA) for visual segmentation or with automatic clustering detection using Fuzzy C Means clustering (FCM). The clusters obtained can be interpreted using external consumer factors by using the Partial Least Square – Discrimination Analysis (PLS-DA) model. The different approaches are compared in terms of data analysis methodologies, modelling, outcomes, interpretation, flexibility, practical issues and user friendliness.