Profit Driven Business Analytics

CHAPTER 5Profit-Driven Analytical Techniques

INTRODUCTION

In Chapter 4, uplift modeling was introduced as a means to further optimize marketing campaigns by estimating the net effect of a campaign on customer behavior. More generally, uplift modeling allows users to optimize decision making by estimating the net effect of operational decisions. When applied for developing retention campaigns, uplift modeling allows to select only those would-be churners that can effectively be retained, as well as to customize the campaign so as to maximize the probability for individual would-be churners to be retained by the campaign. The returns of such a campaign can even be further optimized by accounting for customer value and by adopting profit-driven analytics as introduced in this chapter. Customers with a high CLV that are about to churn may be given a stronger incentive to remain loyal, and are more important to be detected when about to churn than a customer with a low CLV. Acknowledging customer value when developing predictive and descriptive analytical models, as well as when making decisions based on these models, is what profit-driven analytics are about.

In this chapter, we introduce various profit-driven analytical techniques. The first section of this chapter motivates the use of profit-driven predictive analytics and discusses a number of key concepts, such as the cost matrix and the cost-sensitive classification cutoff. Next, a cost-sensitive classification framework is introduced ...