Consumers are diverse in their product perceptions. But within the consumer population there are often consumer segments whose product perceptions are relatively homogeneous. To discover these consumer segments, consumers\u2019 product-related responses are submitted to a cluster analysis. The particular cluster analysis is chosen by the researcher. But the choice of clustering algorithm can have profound consequences on the clustering solution that is obtained. This means that if a different clustering algorithm had been used, it might yield a different clustering solution. It is recommended that the researcher determine what is of primary relevance before conducting a cluster analysis, then to select a cluster analysis algorithm that aligns with the research goals.<\/p>\n\n\n\n\n\n\n\n
In this talk, we discuss the Ugly Duckling theorem and the importance of choosing the right clustering algorithm. For example, consumer clusters can be derived from product-related check-all-that-apply (CATA) responses. Similarity measures are often used to obtain consumer clusters. But clusters based on conventional similarity measures such can yield paradoxical clustering solutions: it can group consumers who check a lot of attributes but disagree together into a cluster that fails to discriminate any products at all. Recently, a new clustering algorithm, called b-cluster analysis, was proposed to cluster consumers according to product differentiation, rather than product characterization. Since this algorithm is designed to maximize product differentiation, it produces more discriminating clusters that, depending on the research question, could be more relevant.<\/p>\n\n\n\n
Next, we show that the idea that underlies b-cluster analysis can be extended to groups consumers based on their hedonic product responses. Again, clustering methods that are used routinely can be shown to produce paradoxical results. A clustering algorithm can maximally exploit the available data yet also fail to answer the correct research question. A new proposal for how approach the problem is proposed, focusing on relevance.<\/p>\n\n\n\n
Castura, J.C., & Naes, T. (2022). Clustering consumers based on their hedonic responses. Society of Sensory Professionals 2022 Conference<\/em>. 2-4 November. Savannah, GA, USA. (Oral Presentation).<\/p>\n\n\n\n