Segmentation Applications in Banking - Data Mining Techniques in CRM: Inside Customer Segmentation - page 256

classification rules can be used for scoring and the assignment of new records
to segments.
Another useful approach is the development of a decision tree using
as inputs only those fields that are available upon customer registration,
for instance demographic information and contract/account details. This
proactive model can then be used immediately after the registration process
of new customers to estimate their future behavior and assign them to the
appropriate segments. Obviously this type of assignment provides a short-
term, early-stage estimation and should be replaced with the actual behavioral
segmentation scheme for established customers.
Figure 6.15 presents a subset of the decision tree rules derived for
the credit card holder segments. The model was trained using the set of
fields denoting the percentage of purchases by merchant categories as inputs.
Note how the decision tree has identified the profiling rules for each of the
eight segments. In the displayed set of rules, we can see the association of
segment 2 with telecommunication services, while segment 3 is associated
with supermarket purchases, segment 4 with use for traveling, and cluster 7
with spending on entertainment.
USING THE SEGMENTATION RESULTS
Once the segmentation scheme was finalized and approved, it was then deployed.
An update process was designed and implemented. It was scheduled to run on a
monthly basis in order to keep the segmentation scheme informed and to monitor
the possible movements of credit card holders across segments. The deployment
procedure ensured the storage and distribution of segment information to decision
makers and authorized marketing users and the loading of the segments onto the
operational CRM systems.
One evident conclusion of the segmentation analysis was that most of the exist-
ing customers seemed to be using their cards for specific purposes. The majority of
credit card holders were characterized by non-diverse usage, a sign of a non-mature
market. Only ''Typical users'' seemed to use their cards to cover various types of
needs. Thus, a first marketing action that was decided was to motivate customers
in the remaining segments to start using their cards in all aspects of their everyday
life. To this end, a marketing campaign was set up, which offered incentives to
customers in order to encourage them to expand and diversify their card usage.
The derived scheme identified different customer typologies and revealed
different needs and preferences of the customer base. This information was then
used as a guide for the development of new specialized products/reward
programs and tailored offers/incentives in order to achieve more specialized
customer handling.