Intelligent Classification-Based Methods in Customer Profitability Modeling

Abstract

The expected profits from customers are important informations for the companies in giving acquisition/retention decisions and developing different strategies for different customer segments. Most of these decisions can be made through intelligent Customer Relationship Management (CRM) systems. We suggest embedding an intelligent Customer Profitability (CP) model in the CRM systems, in order to automatize the decisions that are based on CP values. Since one of the aims of CP analysis is to find out the most/least profitable customers, this paper proposes to evaluate the performances of the CP models based on the correct classification of customers into different profitability segments. Our study proposes predicting the segments of the customers directly with classification-based models and comparing the results with the traditional approach (value-based models) results. In this study, cost sensitive classification based models are used to predict the customer segments since misclassification of some segments are more important than others. For this aim, Classification and regression trees, Logistic regression and Chi-squared automatic interaction detector techniques are utilized. In order to compare the performance of the models, new performance measures are promoted, which are hit, capture and lift rates. It is seen that classification-based models outperform the previously used value-based models, which shows the proposed framework works out well.