A lot has been said and published on the topic of segmentation, typology or cluster analysis. One possible reason for this activity is that the approach is very plausible and even simple at the first glance. On the one hand the concept of "type" is well known in everyday language (e.g. "a modern type") and on the other hand, even the mathematical (numerical) part of this approach seems to be rather simple at the first view: one only has to find homogenous groups of objects (e.g. persons) that are maximally separated. This paper will snow that there are a lot of difficulties in the realisation of clusterings and give some hints for finding acceptable solutions.
The aim of this contribution is the evaluation of clustering techniques from the viewpoint of practical marketing application. Clustering techniques (segmentation methods, grouping methods) are procedures that search for and detect natural groupings of objects (eg. persons) which are described by their values on variables. The requirements that must be met by such methods and the criteria for their evaluation which will be discussed have been shown to be important in the repeated application to practical marketing problems.