Likely to purchase
Identifying your customers.
support line interactions) than profits for the business. If we use
traditional query techniques, we may select a subset of these possible
customers and target those of a certain age or in a particular income
bracket or household size. This type of segmentation is often based
on intuition or business experience. Yet, only a subset of these
customers are likely to purchase such a product or respond to a
campaign, whether contacted by mail, phone, or e-mail.
Where customer acquisition costs are high, either due to the
means of customer contact or incentives offered to them, knowing
which customers are likely to respond (see Section 2.1.3, “Response
Modeling”) and the potential value of those customers can greatly
reduce costs. Yet still, only some of the customers who purchase or
respond will be loyal. A goal for customer acquisition is to target
those customers who will have the greatest probability for response,
loyalty, or lifetime value.
For a moment, let's go back to the simple query approach. Form-
ing precise boundaries to determine which customers to target based
on intuition or business experience may be accurate some of the time,
but likely leaves out some customers, perhaps many, who may prove
to be highly valuable, simply because they didn't meet a precon-
ceived set of constraints.
Data mining is a key component of any modern customer acquisi-
tion strategy and revolves around several techniques. Given a set of
potential customers, perhaps obtained as data acquired from a third
party, data mining techniques such as clustering and classification
can be used to identify the various customer segments that exist
among those customers. After analyzing each of those segments, we
can determine the likelihood that each customer segment, and each
individual customer, will purchase specific products. To achieve this