Chapter 4. How can I characterize my customers from the mix of products that they purchase? 49
If you do not have existing business rules, or a set of customer characteristics
that you prefer to use, then we have attempted to describe a set of customer
attributes; we call them characteristic variables, that all retail organizations
should be capable of deriving from information captured at the point of sale. We
are going to use these characteristic variables throughout this book and hopefully
show you how data mining can be used to generate the business value from
them. If you do not already derive these, or a similar set of characteristic
variables about your customers, then you should consider doing so.
4.2 The data to be used
You clearly cannot do data mining without having the data about your customers
to mine. But what data do you need?
The second stage in our data mining
is to identify the data required to address the business issue and where
we are going to get it from.
In this section we look at what types of data are typically available to retail
organizations and how these can be used to determine customer characteristics.
We then suggest how to construct a relatively simple data model that can be
used as a starting point for discovering similar groups of customers. The data
model we describe should be derivable by all retail organizations from data
routinely collected at the point of sale. The model can then easily be extended to
include other types of data collected about your customers.
4.2.1 The types of data that can be used for data mining
Depending on the size of the your organization, you already have a wide variety
of data about your customers. This may range from transaction data, collected at
the point of sale, to the storage of all types of derived information about your
customers, stored in some form of data warehouse.
To perform data mining for customer characterization, there are essentially five
types of data that can used to develop a customer segmentation model:
Customer relationship data
Although all are desirable, only the first two are essential. The five types of data
are detailed below.