DM Stat-1 Articles
Link to Home

Link to Articles

Link to Consulting

Link to Seminar

Link to Stat-Chat

Link to Software

Link to Clients

Retain Best Customers and Maximize
 their Potential:
A CRM Machine-learning Approach

Bruce Ratner, Ph.D.

A universal tactic of a customer relationship management (CRM) strategy is retaining best customers (i.e., who are now most responsive) and maximizing their future profits, namely, identifying high-value responders. The specificity of this tactic is as varied as the industries that implement the CRM paradigm. For example, in the catalogue sector where efficient target marketing is an ongoing plan, one wants to identify potential responders who will not return their purchased items; in addition, regarding harvesting the customer database, one desires to identify potential buyers of one product who will subsequently buy related products. In the credit card sector where assessing risk is essential for reliable investment projections, one requires to identify customers who are likely to be approved for credit line and are expected not to be write-offs. In the telecommunications industry where the revenue-retention ratio is a baseline performance metric, one seeks to model customer tenure in tandem with usage for identifying customers who are likely to have long tenure (continual responsiveness) and will have high usage of services.

The current two-stage approach for identifying high-value responders consists of building two models and then multiplying their corresponding model scores. One model is a logistic regression model for identifying responsive customers. The second model is an ordinary regression model for identifying high-profit customers. The two model scores are multiplied producing a single score that identifies individuals who are both most likely to respond and contribute large profit - the high-value responders. This widely used approach produces suboptimal results, and is cumbersome to perform and validate.

The purpose of this article is to introduce the machine learning CPR (combined profit and response) Model, which simultaneously addresses two important objectives facing database marketers: maximizing response and maximizing profit. The CPR Model balances the two objectives directly yielding a single score that identifies high-value responders. The CPR Model is theoretically optimal, and easy to build and validate. I discuss two real CRM case studies to compare and contrast the CPR Model and the current two-stage method.

Related Articles: Click here.

For more information about this article, call Bruce Ratner at 516.791.3544,
1 800 DM STAT-1, or e-mail at