Customer Segmentation Based On Structural Equation Modeling

The research content in this paper is customer segmentation, and basic method isStructural Equation Modeling. To improve the classification, description andprediction effect of classification method based on the Structural Equation Modeling(SEM), the author conducts the following studies.1. It recalls the development process of customer segmentation theory, summarizethree different classification methods according to the different characteristics:segmentation based on demographic characteristics, segmentation based on customerbehavior and segmentation based on customer psychology, and briefly describes theClustering Analysis and Automatic Interaction Detection methods (AID) as the simplemarket segmentation methods. By comparing the advantages and disadvantages ofClustering Analysis, AID and SEM, it comes to a conclusion that classificationmethod based on the Structural Equation Modeling is the most promising andeffective method. The few existing classification methods based on the ESM areconducted a detailed study and comparison.2. By comparing the few existing classification methods based on the ESM, itfinds that the advantage of the REBUS-PLS method is to find people who have thesame behavior characteristics and psychological characteristics as a homogenousgroups,but it has difficult to describe and predict the characteristic of a group usingthe demographic information. On the contrary, the PATHMOX overly usedemographic information to classify the customers, therefore it can not guarantee thehomogeneity within the members. As a starting point, the author tries to find a methodthat can finish both classification and prediction work. The author designs a new―distance‖concept in which the demographic information is also added to thecalculation of the distance based on the REBUS-PLS, and gives a comprehensiveconsider of the internal and external characteristics when conduct the classification Inorder to prevent demographic information surpasses the role of the internalcharacteristics in the classification progress, the author sets up a less than1adjustableweight for demographic information.3. Because application of SEM in customer satisfaction is very successful, so it’snaturally to choose the classic customer satisfaction case as the very beginning of the practical application. According to the2008Spanish financial institution customersatisfaction survey results, the author builds a customer satisfaction model, thevalidity, reliability and stability analysis further confirms the effectiveness of thestructural equation model. The new approach is conducted during the customersegmentation, comparing to the result of REBUS-PLS, the new method has certainlyachieved both classification and forecasting purposes. There are many problems in theprocess of customer segmentation, the author lists them honestly, to provide areference for future research.