Solving Problems in Industry - Java Data Mining: Strategy, Standard, and Practice - page 28

could have many reasons: the person is on vacation, lost his job for
which the phone was largely used, started using other means of
communication such as Internet, or switched to another service
provider for work-related calls, but still uses the service for personal
calls. Figure 2-2 illustrates a pattern in minutes of usage that a
possible churner may exhibit before terminating his account.
However, different groups of individuals may be exhibiting this
behavior, for example, teenage girls with large family and friend
circles, 30-something single male professionals, etc. Understanding
the particular characteristics of each of these groups enables busi-
nesses to develop campaigns to retain such customers or to increase
their service usage. Data mining can identify the important factors or
attributes that lead to a specific behavior, as well as group individu-
als according to their behavior.
A customer retention or loyalty strategy can revolve around
several techniques. Customer loyalty can be increased when that
customer purchases more products. Identifying which other prod-
ucts existing customers are likely to purchase, called cross-sell (see
Section 2.1.5), can meet this objective. The data mining technique
association , as discussed in Section 4.5, can help here.
We have already noted that some customers are more valuable
than others, that is, they purchase products in greater quantity or
purchase more profitable products. Helping to prevent the loss of
such high-value customers is another area where data mining can
help. First, one needs to be able to identify high-value customers.
Being able to classify individuals efficiently as high , medium , or low
value, or as representing a specific dollar amount to the business, is a

Possible
Churner

More likely
Churner

Account
Terminated

Time

Figure 2-2 Customer J. Doe's cellphone usage pattern.