Hing-Po Lo (Department of Management Sciences, CITYU)

Title: A Scoring Model for Loyal Customers in Direct Marketing - 
A Practical Use of the Repeat Buying Theory

Abstract: The selection of potential customers to mail an offer or catalog 
is of key importance to direct marketers and catalogers. Who should receive 
which catalogs and when to send out the catalogs are difficult questions 
constantly challenging the management of direct marketing companies. Catalogs 
are never mailed out at random. They are always mailed to a list deliberately 
chosen in order to achieve some purposes. Since the databases of customers 
of these companies are usually very large, different data mining tools
have been proposed to develop the mailing list. The most commonly used method 
in this industry is the RFM, the Recency, Frequency, Monetary segmentation. 

Alternatively, scoring models can be built using logistic regression,
CHAID/CART and ANN. Scoring models assign to each customer a score that 
enables the direct marketer to rank all customers in question.  When using 
the scoring models, deciding how many customers to mail is a difficult 
decision. The common practice is to send mail to those people in the mailing 
list, whose expected returns exceed the marginal costs of the mailing. 
However, this is not the best solution. The objective of this paper is
to propose a new approach of customer selection that makes explicit use of 
one very important feature of a customer: the loyalty of the customer. The 
theory of loyal customers (or repeated buyers) has been extensively studied 
by Ehrenberg (1988). By using this theory, we are able to better predict the 
behaviors of customers and argue that the mailing list could be divided into 
two groups for separate model developments. Also, separate estimates of 
response rates can be obtained, which can be used to set the cut-off points 
for direct marketing. Empirical results based on a large data set are used to 
demonstrate our new approach of customer selection.

(Joint work with Xiaoling Lu, Department of Statistics, 
Nankai University, China)