資料庫行銷是落實關係行銷與一對一行銷不可或缺的利器，廠商透過資料庫的分析得以衡量個別顧客之價值。傳統的顧客價值衡量方法多偏重於顧客於特定時點的平均價值，並未深入動態探討。廠商若能從動態的角度觀察顧客價值成長或萎縮的過程，事前預判顧客價值的遷移路徑，則更能達到開發顧客價值潛力以及預防顧客價值流失的效果。本文以馬可夫鏈模型為基礎，根據顧客過去之價值狀態遷移路徑，搭配層級貝氏方法，如貝氏多項式模型、層級貝氏Probit模型以及層級貝氏Logit模型，建立個人化的移轉機率矩陣，進而預測顧客的價值遷移路徑，據此採取適當的顧客關係行銷之策略。

Database Marketing is a requisite and efficient weapon for firms to put one-to-one relationship marketing in practice. The major task of this practice is to evaluate or calculate individual customer value through proper analyses of customers' purchase history data. Conventional approaches to measuring customer value mainly focus on the average value from past to present, lacking the dynamic viewpoints. What firms need to do is to be able to capture the migration patterns of a customer's value in advance, in order to explore potential customers and prevent inactive customers. This paper employs the Markov chain model and the hierarchical Bayesian approach to construct individual customer's transition probability matrix for forecasting the customer value migration process.