Modeling for Churn

As the economy has grown, more companies have learned to use database marketing to retain customers. Competition in some products, particularly credit cards, cellular phones, long-distance service, and health care, has become intense. The attrition rate, or the rate at which people switch to another provider, is very high. At the same time, statistical modelers have found that regression analysis can help in predicting which customers are most likely to drop the service. The models use a combination of appended data and transaction history to pick out the variables that precede a customer’s dropping the service. Then advanced database marketing techniques are used to concentrate on these potential defectors and get them to change their minds. Defection prediction plus lifetime value analysis further concentrates the communications on those whom it is most profitable to serve. The result is risk/revenue analysis (see Table 1-2).

Table 1-2 : Risk/Revenue Matrix

			Probability of leaving soon
LTV	High	Medium	Low
High	priority A	priority B	priority C
Medium	priority B	priority B	priority C
Low	priority C	priority C	priority C

In this analysis, which is discussed later in this book, a concentration on those customers who are priority A or priority B reduces the task of lowering churn by 56 percent, saving resources and boosting profits.

Bruce McDoniel of Summit Marketing used this system with a large regional bank to find the right price for priority A customer renewals. The goal was to get customers to renew, but also to maximize revenue. Bruce tried three price offers that netted the bank from $8 to $10. Table 1-3 shows the result of the test.

Table 1-3 : Price Offers

Number tested	Number renewing	Price offer	Revenue
1000	800	$10	$8000
1000	900	$ 9	$8100
1000	950	$ 8	$7600

This simple test enabled the bank to fix the offer price at $9.