Assessing Dell Customer Value Based On Two-stage Model

With the increasing competition in the market, modern marketing theory experienced from the first "enterprise-centric" to "consumer-centric", both of which are based on the "mass market". But in the modern society, the demand of pluralistic heterogeneity is becoming more and more prominent, the demand of consumers is more and more differentiated, "segmented market" has increasingly become the main feature of the modern market, which requires companies’marketing strategies for each market segment more targeted, more "accurate".Customer segmentation is a core part of the precision marketing, which makes customer analysis become one of the core content of the theory of marketing research. In order to study customer segmentation, it is an important indicator that from the company’s point of view to assess the profit contribution brought by customers-customer value.Therefore, this paper is an actual business case study whose goal is to assess DELL corporation clients’customer value. The nearly 200,000 dataset used here is from the database of one consulting company. The dataset consists mainly corporation clients’basic information and consumer behavior information. There are nearly 600 variables, such as the corporation clients’age, industry back-ground, the number of employees, sales and so on. And continuous variables and category variables are all included here.To solve this business case, this paper using two-stage model to assess customer value. The first step, predicting customer response rate to DELL products for the next year based on the logistic regression model. The second step, making the estimated response rate weighted to the margin variable, then predicting the margin in the next year brought by responsible customers based on the multiple linear regression model. Meanwhile, to make the established model robust, here created two alternative datasets that are in-sample validation dataset and the out-of-sample validation dataset. Moreover the in-sample validation dataset and the modeling dataset are all randomly sampled from the same dataset, which can present very similar performance. While the out-of-sample validation dataset sampled from another dataset is different from the modeling dataset. After validation for the established two-stage model by using the decile analysis and lift chart, it seemed that the performance of the two alternative datasets fitted with the modeling datasets strongly, which suggested the established two-stage model is robust, and it is an effective model to assess customer value.