The Mining Data Mart - Data Mining Techniques in CRM: Inside Customer Segmentation - page 179

A period of two years is proposed as the time frame for the data mart in
order to support the majority of the data mining applications, including predictive
modeling. Once again the main information tables should be combined with
descriptive lookup tables which can facilitate further categorizations of the data
dimensions. A lookup table of critical importance is the one used to describe and
organize the product codes (UPCs) into a multilevel hierarchy. This hierarchy
maps each product into corresponding categories and subcategories, enabling
the grouping of purchases at the desired product level. In a dynamic retailing
environment, the products offered change continuously. Therefore the product
categorization table should be well organized and the product taxonomy should be
frequently updated to accurately account for all such product changes.
TRANSACTION RECORDS
The data collected at the point of sale log all the details of each transaction, carrying
information about the type of products purchased, the value and amount of each
product bought, the store visited, and the exact date and time when the transaction
took place. Although the format and the exact contents of the transaction records
vary among retailers, a typical example of the information logged is given in
Table 4.29.
The above data form the basic information blocks for building the retailing
data mart proposed in the next section. The scope of this data mart is to provide
a good starting point for the majority of the analytical tasks, without having
to undertake rigorous data management of raw transactional records. For extra
information not covered by the data mart, data miners can always return to the
raw transactional data to make use of their detailed information.
CURRENT INFORMATION
Current tables contain information about current and past customers and only
carry the latest update of the relevant information. A typical example of a current
table is the one used to store the latest socio-demographic information of all
registered customers.
Customer Information
Apart from transactional data, the data mart should incorporate all the customer
socio-demographic and contact information. Information related to the customer's
life-stage, such as age and marital status, may determine at a certain level his or