Segmentation Applications in Banking - Data Mining Techniques in CRM: Inside Customer Segmentation - page 243

Figure 6.6 Bubble-plot illustrating the percentage of purchases at home stores, for
traveling, and for entertainment per cluster.
attributes which had not participated in the building of the cluster model, including
general usage fields and demographic information.
The separation of the clusters is reflected in the main usage KPIs presented
in Table 6.9. Cluster 8 has the lowest purchase amount as opposed to clusters
4, 7, and 1. Clusters 3, 6, and 7 carry the highest balances. The international
usage of the cards is generally low, except by clusters 4 and 7. Those two clusters
also have the highest e-shopping usage. With respect to their demographic profile
(Table 6.10), clusters 2, 4, 6, and 7 concern a large number of younger customers
as opposed to cluster 5 which seems to contain mostly mature people. Women
form the majority in cluster 6 and men in cluster 3. Clusters 3 and 5 are mainly
composed of married customers. The smaller percentages of married customers
can be found in clusters 2, 4, and 7.
The profiling phase concluded with the application of statistical hypothesis
tests on the percentage purchases fields. These tests compared the mean values of
each cluster to the overall population means and identified statistically significant
differences. The test results have the format of profiling charts. The direction of
the bars indicates the relation between the cluster and the overall population. The