The Mining Data Mart - Data Mining Techniques in CRM: Inside Customer Segmentation - page 187

Examples of Derived Measures Used to Enrich the MCIF
An indicative list of more complex KPI examples for retailers is as follows:
• Basket size: average transaction amount and average number of items per
transaction.
• Percentage (%) of purchase amount spent on private labels.
• Percentage (%) of purchase amount spent per product group/subgroup.
• Monthly average number of visits within the period examined.
• Total number of distinct product groups/subgroups purchased.
• Average time interval between two sequential visits.
• Flag of order over the Internet.
SUMMARY
Data mining can play a significant role in the success of an organization, but only if
it is integrated into the organization's operations. Data mining models need data as
input and produce valuable results as output. A standardized procedure is required
for the deployment of the derived results as well as for obtaining the input data.
In this chapter we have dealt with the issue of input data and suggested
the main data dimensions typically required to cover the analytical needs of the
banking, retailing, and mobile telephony industries. We also proposed a way to
organize the data for the needs of data mining applications. Customer attributes
assessed as necessary for marketing and mining purposes should be extracted from
various data sources, transformed, and loaded into a central mining repository, the
mining data mart, which should be designed to cover the majority of the analytical
needs of the organization. The mining data mart should enable analysts to monitor
the key customer attributes over time, without delays for ad hoc data acquisition
and preparation, and without additional requests for implementations from IT.
The contents of the data mart are the main information blocks used to build the
MCIF, a flat table, typically at a customer level, which integrates the information
of each customer into a single record, a customer ''signature'' record. The MCIF's
purpose is to be an almost-ready input file for most of the upcoming analytical
applications.