In the first part of this paper we will review and modify the cluster analysis procedure presented by Green, Frank and Robinson [Green, P. E., R. E. Frank, P. J. Robinson. 1967. Cluster analysis in test market selection. Management Sci. 13 (8, April) B-387–B-400.] in a recent issue of this journal. In revising their procedure, we will raise some very fundamental questions with respect to cluster analysis in particular and multivariate statistics in general. The scale that we use in measuring the input variables (e.g., age, income, education, etc.) will affect the results. The question is: “How do we scale the input variables so that the results are ‘meaningful’?” We will see that some of the usual methods that are used to give statistically meaningful results will not assure us of managerially meaningful results. Finally, a possible “solution” along with its advantages and disadvantages over standard techniques will be presented.