異常事件對顧客價值之干擾效果-個體層次層級貝氏模型之運用

The Disturbance of Abnormal Events to Customer Value: An Application of Individual-level Hierarchical Bayesian Models

汪偉霖(Wei-Lin Wang)

指導教授 ： 任立中

國立臺灣大學/管理學院/國際企業學研究所/碩士(2007年)

https://doi.org/10.6342/NTU.2007.02603

摘要

顧客價值為企業制定行銷策略的重要依據。從廠商的角度出發，顧客終生價值衡量個別顧客為廠商創造的利潤之折現值。然而受限於資料取得不易，顧客終身價值模型通常忽略異常事件對顧客終生價值所造成的干擾，導致估計值與實際值的偏離。
本研究以事後評估的觀點，設定受到異常事件干擾之顧客為實驗組，運用個體層次層級貝氏模型，以事件前資料估計模型參數，模擬在未受事件影響之下實驗組顧客的顧客終生價值。其次假設另一群未受事件干擾的顧客為控制組，使用同樣方法估計其終生價值，再以控制組的估計偏誤調整實驗組之模擬值。最後比較實驗組顧客之調整後模擬值與其受干擾之事後實際值，進而估算出異常事件對該群顧客價值之干擾效果。
實證研究以台灣某線上文具供應商為例，該公司之顧客資料庫為離職員工所盜竊，該名員工亦在離職後對資料庫所紀錄之顧客進行行銷活動。針對此異常事件，本研究根據事件前已列入資料庫之顧客，挑選其中在事件前後皆有兩筆以上交易資料者為實驗組樣本，並以相同標準挑選事件後才列入資料庫之顧客為控制組樣本。研究發現此異常事件造成實驗組顧客終身價值之減少，以累積銷售而言，廠商在事件後第228天之損失金額為新台幣3,377,181元。
本研究在兩組樣本皆發現，儘管全體顧客在事件後仍持續購買行為，然而部份顧客於事件後延長其購買期間，另一部份顧客則縮短其購買期間。另就群體行為來看，延長購買期間的顧客在實驗組所佔比重較高。針對此發現，本研究運用個體層次層級貝氏混合模型，討論實驗組中個別顧客是否因事件影響而造成購買期間之結構性改變。然而研究過程中面臨模型認定不足之問題，故本研究並未對前述命題做出結論，唯待未來研究對模型進行改良並對此一命題提出解釋。

關鍵字

顧客價值 ； 顧客終生價值 ； 異常事件 ； 層級貝式模型 ； 購買期間 ； 層級貝式混合模型

並列摘要

Customer value is a key reference for businesses to formulate marketing strategies. From the suppliers’ perspectives, customer lifetime value is a managerial tool which is used to calculate the present value of the profit contributed by each customer. However, customer lifetime value models usually ignore the impact of abnormal events due the lack of data, and therefore overestimate/underestimate the true customer lifetime value.
Since the occurrence of an abnormal event is identified, this study regards customers who have been affected by the event as the experimental group. Besides, this research uses individual-level hierarchical Bayesian models to simulate the true customer lifetime value of customers belong to the experimental group based on data previous to the occurrence of the event. Furthermore, another group of customers who have not been influenced by the event are regarded as the control group which is used to adjust the simulated customer lifetime value of the experimental group. Finally, by comparing the simulated value after adjustment and the realized value, the disturbance of the abnormal event to customers from the experimental group could be estimated.
In empirical study, this research studies an online stationary vender in Taiwan. The customer database of this company was stolen and leveraged to run a new business by an ex-employee of the company. To study this abnormal event, this research select customers who had been recorded in the database before the event occurred with at least two transactions both ex ante and ex post as the experimental group. In addition, other customers who was recorded in the database after the occurrence of the event and have at least two transactions are chosen as the control group, After the analysis, this research finds that the abnormal event does decrease the customer lifetime value of customers from the experimental group, and the loss value in terms of cumulative sales at the 228 day after the event occurred is NT$ 3,377,181.
Another finding in this study is that although all customers keep purchasing after the occurrence of the event, some of them lengthen their inter-purchase time, while others shorten their inter-purchase time. This phenomenon is detected in both groups, but comparing to the control group, more customers from the experimental group lengthen their inter-purchase time. In order to identify whether each customer belonging to the experimental group has a structural change in his or her purchase behavior in terms of inter-purchase time after the event occurred, individual-level hierarchical Bayesian mixture models are introduced. Unfortunately, these models face the question of under-identified. Thus, future researches are needed to improve the method and to study the tapped but unsolved issue.

並列關鍵字

customer value ； customer lifetime value ； abnormal event ； hierarchical Bayesian model ； inter-purchase time ； hierarchical Bayesian mixture model

參考文獻

Allenby, G. M., Leone, R. P., and Jen, L. (1999), “A Dynamic Model of Purchase Timing With Application to Direct Marketing,” Journal of the American Statistical Association, 94(446), pp. 365-374.

Ansari, A., Essegaier, S., and Kohli, R. (2000), “Internet Recommendation Systems,” Journal of Marketing Research, 37(3), pp. 363-375.

Blattberg, R., Getz, G. and Thomas, J. S. (2001), “Customer Equity: Building and Managing Relationship as Valuable Assets,” Harvard Business School Press, Boston.

Bolton, R. N. (1998), “A Dynamic Model of the Duration of the Customer’s Relationship with a Continuous Service Provider: The Role of Satisfaction,” Marketing Science, 17(1), pp. 46-65.

Christopher, M. (1996), “From Brand Values to Customer Value,” Journal of Marketing Practice, 2(1), pp. 55-66.