Something Approaching Science? Cluster Analysis Procedures In The Crm Era

The CRM industry is set to be worth $16.8bn by 2003 but almost 50% of firms are dissatisfied with the performance of their systems. Evidence suggests that CRM activity is currently concentrated on segmentation - a necessary precursor to fully developing differentiated customer relationship strategies. The time is ripe to examine in some detail the reliability and validity of cluster analysis, the statistical technique most widely used to identify segments. If ways can be suggested to improve the application of the technique, then this may render improved performance from CRM systems. This paper reports the findings of a study which subjected a data set designed for segmentation purposes to a series of rigorous validity and reliability tests and went as far as to randomise the data to ascertain whether current methods could detect "false" data. Our findings highlight that cluster analysis is a contentious methodology, exploratory in nature and relying extensively on subjective interpretation. Whilst the cluster solution based on the randomised data did, in fact, break down ultimately this was only after subjecting the procedure to a very rigorous series of tests. The question which arises is whether marketing managers faced with cluster solutions from CRM applications would test their data with such vigour. If they do not then they run the danger of seeking out relationships not with real groups of people but with meaningless computations of data.