Segmentation Applications in Banking - Data Mining Techniques in CRM: Inside Customer Segmentation - page 271

if the bank wants to retain and win them back it should initiate immediate action
to stop them from turning to the competition.
The remaining segments were assigned to individual business units as a first
step in building specialized marketing strategies. The efforts of the bank to gain
deeper insight on its customers did not stop there. As part of a more thorough
investigation, the bank decided to initially focus on the ''Pure Mass'' segment and
to apply a clustering technique to identify the underlying distinct customer types
and behaviors. The following sections present the results of this project and the
behavioral sub-segments revealed for the ''Pure Mass'' customers.
SEGMENTATION USING BEHAVIORAL ATTRIBUTES
The aim of the sub-segmentation process was to further break down the ''Pure
Mass'' customers into distinct smaller groups of differentiated needs and behaviors
that could support the design of customized marketing efforts.
A clustering technique was applied to reveal the natural groupings of the
customers with respect to their behavioral characteristics. The required informa-
tion to support the training of the model was retrieved from the organization's
mining data mart and MCIF. It covered six months of data and contained sum-
marized information, aggregated at a customer level, on the following aspects of
behavior:
• Product ownership and utilization: Each customer's balances (monthly
average balances) with respect to the main product categories.
• Types of transactions: Relative number and volume (monthly average) per
main transaction type.
These were the main segmentation dimensions. An extended list of relevant
fields was prepared for the needs of the analysis, to provide a complete view of
each customer in regard to the above characteristics. This list included derived
attributes such as monthly averages and ratios (percentages of total balance and of
total transactions) which denoted the dominant product categories and the most
common types of transactions for each customer. The complete list of input fields
can be found in the next section.
Selecting the Segmentation Fields
As noted above, the list of clustering inputs contained only fields related to the
specific marketing objectives of the segmentation process (Table 6.14). Other
informative, yet not directly relevant, fields, such as customer demographics