Analysis of Customer Value and Product Recommendation System-Transactions of Locks for Example

[[abstract]]在行銷領域中，客戶之間永遠存在著異質性。企業80%的銷售利潤是來自於20%的客戶，而其餘20%的銷售利潤，卻花了公司80%的行銷費用。有鑑於此，如何找出具有價值的客戶，進行產品的行銷對企業是多麼重要。具有價值的客戶係指其未來為公司帶來的利潤大過於公司花在其身上的成本。本研究是以資料庫行銷的技術來分析不同價值群客戶的真實消費狀況，以鎖匣廠商的的交易資料庫為例，推測建設公司(即客戶)的未來採購行為。本研究以兩種模型分析客戶價值，第一是RFM分析模型，第二是客戶活躍性的分析，並利用卜瓦松?歸(Poisson regression)來建構出一套產品推薦系統。本研究之推薦系統模式為「內容導向模式」(Content Model)，以客戶對「產品屬性的偏好」為基礎作推薦。在此模式下，不同客戶價值群具有不同的產品偏好結構，廠商可據此進行產品的推薦。另外，對於無交易紀錄的新客戶，本研究利用舊客戶的偏好結構與顧客特質做交叉分析，做為新客戶產品推薦的依據。冀望廠商在競爭激烈之環境中，不但可以藉由資料庫找尋「最有價值之客戶」，利用產品資料庫中萃取「最適產品屬性組合」給合適之客戶。對不同價值群客戶做有效的推薦，節省行銷成本，且能為建設公司複雜的採購流程做最好的產品之推薦，並提升經營效率以期建立一套完整的「產品購買推薦模式」。本研究亦針對各不同客戶價值群建議適當的策略主軸，並研擬適當具體方案提供五金門鎖經銷商的參考，期後能幫助企業贏得最大的利潤。[[abstract]]There is heterogeneity between customers in the marketing field. Eighty percent of selling profits of a firm comes from twenty percent of customers; however, the rest twenty percent of profits costs eighty percent of marketing expenses. Therefore, how to find those valuable customers and develop product marketing becomes much more important. A valuable customer is defined as bring more profits to a firm rather than a firm which invests on a customer. This study intends to analyze purchase situation of different valuable customers by database marketing. We use the construction company purchase history from a locker firm’s transaction database as the data analyzed by our model to predict their purchase behavior in the future. The study adopts the RFM approach, the active analysis, and the Poisson regression model to build a set of recommendation system on specific products. One of recommendation systems is the content model which is based on customer’s preference of products. Under this model, different customers are assumed to have their own preference structure and the firm should build the appropriate product recommendation according it. Besides, to the new customers who have no transaction records, we use a cross analysis about the relationship between the preference structure and characteristics of the old customers as the basis of product recommendation to the new customers. We hope that firms could not only find the most valuable customers in such competitive surroundings but also extract the most appropriate combination of product properties for different customers by database marketing. Consequently, firms can improve their relationship with customers by helping them make complicated purchase decisions and profit from the long-term relationship with customers.[[note]]碩