以改良式RFM模式結合資料探勘建立客戶分群並提升行銷效益

Abstract

依據義大利經濟學家帕列托所提出所謂80/20法則(1897)，意指在原因和結果、努力和收穫之間，存在著不平衡的關係，而典型的情況是：80%的收穫，來自20%的付出，也就是80%的結果，歸結於20%的原因。而在實際的情形下也是如此，20％的重要客戶貢獻了利潤的150％；而最差的40％客戶，使利潤縮減50％，可見客戶區隔的重要性。 所謂「客戶分群」，係指利用完整消費者的客戶基本資料、交易分析資料、客戶的互動資料，將客戶劃分為數個不同消費行為模式的客戶群，並建立客戶的分群，以期能讓企業分離中出高價值客戶、主力型客戶、成長型客戶、無價值客戶等，有助於讓企業針對不同的使用群制訂適當的行銷策略。 本研究利用改良型RFM模型建立客戶分群的系統雛形，並利用資料探勘之資料分群之技術-SOM的運算法則將不同的各客戶資料主動依客戶的交易資料歸類四群，並將此四群映射於高價值客戶、主力型客戶、成長型客戶、無價值客戶等四個客戶群組中，同時針對映射的客戶群組資料作一分析與探討，並與市場及業務部門的功能結合，期望利用此一決策系統幫助市場及業務相關人員規劃並審視現行市場及業務策略。

According to the Pareto’s 80/20 rule, there were unbalance relationship between cause and effect. It means that 80% gain may came from 20% contribution in customers. Thus, how to obtain the maximun benefit between customers and enterprise has become a critical issue in market analysis. Customer segmentation is one of approach which is using marketing database, including analysis transaction data, customer data, to cluster the customers into 4 groups: High value customers, Maior Customers, Valueless Customers and Growing Customers. In this paper, we proposed to use refined RFM model and Self-Orgnanization Map(SOM) to build customer segmentation prototype, and then to verify this prototype using transaction data of customers in semiconductor manufacturing. Further, we propose the respective marketing policy for the 4 groups in order to make great contribution to markets and sales.