Data Mining Techniques: For Marketing, Sales, and Customer Relationship Management, Third Edition

Chapter 18

Building Customer Signatures

This chapter is about finding customers in data and building customer signatures to describe them. It is the first of three chapters that deal with various aspects of preparing data for mining. The theme that runs through this chapter is finding customers in data, which means gathering the traces that customers leave when they make purchases, visit websites, contact call centers, pay bills, respond to offers, and interact with the company in other ways. These scattered traces of data must be shaped into a useful customer signature that can be used to train data mining models.

Quite often, the customer signature is the data in data mining. As discussed in Chapters 3 and 5, the signature determines whether data mining techniques produce profiling models or predictive models. The signature is the representation of customers that will be used to make decisions about how they are treated — what offers they receive, how their requests are handled by the call center, and how much effort is made to retain them or win them back. In addition to its primary role supporting data mining, the customer signature is a data source for reporting and ad hoc queries.

A customer signature is a representation of a customer as of a particular date. The code that generates the customer signature should be able to create such a representation as of any date for which data exists. It should also be capable of supporting different “as of” dates for different customers ...