ALTERNATIVE APPROACHES TO CLUSTER-BASED MARKET-SEGMENTATION

The common practice of factor analysing data matrices, followed by a c lustering or rotated, standardised factor scores has recently been que stioned by several researchers in marketing. This paper presents compa rative analyses of nine large-scale data sets in which two different t ypes of transformations are employed prior to k-means clustering. The study results suggest that there is some empirical evidence for the re cent claims. Clusterings based on rotated factor scores generally show poorer associations with exogenous (background) variables than those based on either the original data or a less severe preliminary transfo rmation. Implications of these findings for post hoc market segmentati on are discussed.