The Mining Data Mart - Data Mining Techniques in CRM: Inside Customer Segmentation - page 160

• Number of months with minimum savings balances below a specific
threshold, for instance $50.

• Number of months with bucket

<

1 during the examined period.
• Maximum bucket during the examined period.
• Savings balances delta. Compares the savings balances of the most recent
period, for example, the last three months, to those of the previous three
months and denotes the respective percentage difference.
• Flag of usage of alternative channels (phone, Internet, SMS, etc.).
• Percentage (%) of teller (branch) transactions: teller transactions' amount
as a percentage of total transactions.
• Number of different product categories that each customer owns.
• Number of different channels used by each customer.

>

THE MINING DATA MART FOR MOBILE TELEPHONY
CONSUMER (RESIDENTIAL) CUSTOMERS

The main principles for the design of an efficient mining data mart presented in
the previous sections also apply in the case of mobile telephony. Therefore, the
mining data mart should contain consolidated and pre-summarized information
tailored to cover the majority of the analytical needs. Its purpose is to provide rapid
access to pre-aggregated data and to support the key analytical and marketing tasks
without the need for additional implementations and interventions from IT.
The selected level of aggregation should ensure reduced query execution
times and good performance without sacrificing important information.
The data mart should integrate data from all relevant data sources in order to
cover all the information elements required for analyzing and predicting customer
behavior. Therefore it should incorporate:

• Traffic information, such as number and minutes/volume of calls by call type,
source/destination network, and so on. This information is stored in CDRs (Call
Detail Records) providing a complete usage profile of each customer.
• Information about the customers and the line numbers (i.e., MSISDNs).

More specifically, it should cover, at a minimum, the following data dimen-
sions:

• Registration and socio-demographic information of each customer.
• Core segmentation information (consumer postpaid/contractual, consumer pre-
paid customers) and information on segment migrations.