Influence of cluster analysis procedures on variation explained and consumer orientation in internal and external preference maps

Abstract

AbstractCommon statistical package clustering (SPC) methods may result in consumer segments that still are heterogeneous in product acceptance. Creating new products based on attributes selected from preference maps that use those clusters may result in product failure because the products are developed for consumers with some dissimilar preferences. Methods for clustering that produce more homogeneous clusters of consumers have been reported, that is, SPC with additional “manual” clustering used for gathering consumers with congruent liking of products. In this study, internal and external preference maps were created using two supplemental “manual” clustering approaches and are compared with maps from hierarchical and partitional clustering (i.e., k‐means) results. We observe how clusters with higher homogeneity in product liking patterns change spaces of consumers, descriptors, and product co‐ordinates in internal and external preference maps. Improvement in the preference mapping was exhibited, but the homogenous consumer clusters still miscommunicate some information, as the maps usually do. This was especially true for external preference maps with low percentage variance explained in the descriptors. Thus, it is essential to compare mapping results to the original data and to findings from the original descriptive study to gain the most appropriate information.Practical applicationsCorrectness of sensory information is very important for further analysis. The methods using some additional clustering procedures yield more exactitude in consumer liking patterns than only statistical package clustering. Statistical package clustering with supplementary manual clustering ensures that consumers who share both the most liked product and the least liked products are placed in same in the cluster. However, that does not ensure that the “maps” generated allow for unqualified interpretation. It is essential that preference mapping procedure are joined with common sense reviewing of the original data, key descriptive findings, and key information on most and least liked products within various consumer groups.