不孕症自費市場銷售預估與顧客價值分析之研究 - 以A藥廠為例

Abstract

由於科技的進步，市場競爭日趨激烈，許多企業已逐漸意識到顧客關係管理(Customer Relationship Management)的重要性。資料庫行銷(Database Marketing)是顧客關係管理的一種分析工具，可將龐大的顧客交易資料，藉由資訊科技的輔助，來找出既有顧客與潛在顧客的重覆購買行為。在資料庫行銷的分析中，最普遍使用的三個指標為RFM (Recency, Frequency & Monetary Amount)。利用RFM分析則可將顧客區隔為不同價值群的顧客，則企業就可針對不同的顧客群制定適當的行銷策略，以避免顧客的流失。故本研究的主要目的為預測出顧客重覆購買行為模式及辨認出活躍度高的顧客，以提供適合的行銷資源配置。

本研究以一家藥廠的顧客交易資料為對象，透過Beta-Geometric(Fader and Hardie, 2001)的混合分佈模型，在試算表軟體EXCEL上建立完整購買銷售模型，以對顧客每週購買量做重複購買行為的實證研究。另外，本研究亦採用考量顧客的交易與流失行為的BG/NBD (Fader et al, 2005)模型，依顧客過去購買行為(購買次數、最後一次購買時間與購買時間間隔)，使用機率模型來計算出顧客活躍率及未來時間的期望交易次數。在研究中，我們成功推導出原模型並未導出的個別顧客活躍率公式及預測出個別顧客的活躍程度。

由於Beta-Geometric與BG/NBD兩者都能很容易在EXCEL上建構出預測模型，而模型的參數利用EXCEL軟體的”規劃求解”功能就能快速計算獲得，故行銷人員不需再學習其它複雜的電腦軟體程式。本研究的結果顯示出模型可掌握到購買行為的變化及預測出重複購買行為模式，並以預測出的個別顧客活躍率可來做出顧客價值分群，以做為行銷決策的參考。

With the development of technology, market competitions become increasingly intense. Enterprises begin to know the importance of Customer Relationship Management (CRM). Due to the richness of customer information, database marketing analysis of CRM plays a role in generating integrated and accessible customer information to help marketers predict purchasing behavior of existing customers or prospects. One of the well-known Database Marketing tools is the recency-frequency-monetary (RFM) formula, which can segment the customers into different value groups to choose profitable customers. The purpose of this research is to predict repeat-purchase behavior of customers and identify high-value customers to appropriately allocate marketing resources on right customers.

In this research, we use Beta-Geometric model (Fader and Hardie, 2001) in conjunction with the transaction data of a pharmaceutical company to conduct an empirical analysis of stochastic model in buying behavior. We model weekly purchases using a finite mixture of beta-geometric distributions with a time-varying component to catch the underlying trend in repeat-purchase behavior. In addition, we also make an empirical study on BG/NBD model (Fader et al., 2005) using the characteristics of the purchase and dropout process of individual customers to predict active probability and expected transaction numbers in the future. In the model, we successfully derive the equation of active probability for a random-chosen customer.

Aiming at Beta-Geometric and BG/NBD model, all of them can easily be implemented within a standard spreadsheet environment of Microsoft Excel and their parameters can be obtained quite fast in Microsoft Excel by using ‘Solver” function. The research results demonstrate the Beta-Geometric model is capable of describing the underlying sales patterns as well as produce a good medium-term forecast, and BG/NBD model provides the ability to predict active status & future purchasing patterns of individual customers to build customer segmentation for a reference of marketing decision.