Segmentation Applications in Banking - Data Mining Techniques in CRM: Inside Customer Segmentation - page 226

In the next sections we will follow the efforts of a bank to segment consumer
credit card holders according to their usage behavior. The bank's objective was to
reveal the underlying customer groups with similar usage characteristics in order
to design differentiated marketing strategies.
The first decision that the involved marketers had to make concerned the
main segmentation criteria by which they would group the credit card customer
base. Their primary goal was to use segmentation in order to develop new card
types and reward programs tailored to the different customer characteristics.
Therefore they decided to start by focusing on the customers' purchasing habits
and the products/services that they spend their money on. Their plan was to mine
purchasing preferences and assign customers to segments according to the type of
products that they tend to buy with their cards.
DESIGNING THE BEHAVIORAL SEGMENTATION PROJECT
The goal set by the organization's marketing officers was to group customers
with respect to their purchasing habits and, more specifically, in terms of the
mix of products/services that they tend to buy. This marketing objective was
translated into a data mining goal of behavioral segmentation by using a clustering
algorithm to automatically detect the natural groupings in the customer base.
The final outcome of this project would be the identification of the different
customer typologies and the assignment of each customer to a corresponding
behavioral segment. The data chosen for segmentation purposes included the
relative spending per merchant. This information was available in the mining
data mart of the organization where credit card purchase transactions were
differentiated by merchant category, enabling identification of the purchasing
preferences of each customer.
The implementation procedure for the project also included the application of
PCA to identify discrete data dimensions before clustering. Moreover, it involved
a phase of thorough exploration of the derived clustering solution in order to
interpret the clusters and evaluate the derived solution.
The organization's marketers wanted not only to identify the underlying
customer groupings, but also to disseminate this information throughout the
organization, including the front-line systems, to ensure its use in everyday
marketing activities. Additionally, they wanted to be able to monitor the revealed
segments over time. Therefore the project also included the development of a
deployment procedure for easy updating of the segments that would also allow
tracking of segment evolution over time.
Additional methodological issues that the project team had to tackle before
beginning the actual implementation included: