Modelling consumers' heterogeneous preferences: a case study with Chilean wine consumers

Abstract Background and Aims Understanding consumers' preferences is key to making a successful product, but preferences are heterogeneous. We compare three approaches to consider preference heterogeneity in discrete choice models: (i) systematic preference variations based on socio‐demographic characteristics; (ii) latent classes; and (iii) hybrid choice models with latent variables measuring consumers' attitudes. Methods and Results Data from a stated choice survey of Chilean wine consumers were analysed using three different approaches; these agreed on average trends but differed in fit and implied different trade‐offs. For example, socio‐demographic characteristics correlate poorly with preferences. Latent classes offer a good fit but do not link preference heterogeneity to consumer characteristics. The hybrid choice model provides the best fit but requires more data, making it more difficult to use this approach in forecasting. Conclusions The best approach might depend on the research objectives. Using latent classes on a representative sample is the best approach if forecasting is paramount. Modelling attitudes is helpful when more insight into consumers' preferences is sought. Systematic preference variations based on socio‐demographic characteristics are a good choice when only average trends are relevant. Significance of the Study We make recommendations on how to model preference heterogeneity when studying wine preferences, an issue often overlooked.