摘要
近年來資訊科技的發展，使得顧客收集資訊容易且購物網站的崛起讓消費方式改變，但也同時使得實體通路面對虛擬通路的競爭之下備感壓力。然而，也隨著科技的發展，使得資料庫行銷同時成為企業在作行銷策略時可利用之利器與未來趨勢，讓企業能夠主動地接觸顧客而非被動等待。因此，對於實體通路而言，如何運用最新科技來進行資料庫行銷，便是其未來存活的重要關鍵。 使用資料庫行銷常用到的是運用顧客過去的消費歷史培養與顧客未來的長久關係，而最常見的則是推薦系統的建置。透過推薦系統將新舊產品給新舊顧客，維繫顧客之間關係，甚至強化上下游廠商的聯結性。近年來，層級貝氏統計已被許多學者證實，透過該模型並利用資料庫中的會員個人資料與消費紀錄等，能夠準確地預測消費者的後續購買行為。相較於過去傳統的行銷手法，顧客常被視為是同質的，因此被推薦的產品品項或品牌等每個消費者都相同(傳統總合模式之推薦模型)，但如今行銷策略已盛行為一對一的觀念，將顧客視為異質性(Heterogeneity)並且擁有個人偏好與個人消費模式，因此，根據購買紀錄與會員資料所建置出來的產品推薦系統才能真正反映出顧客的需求。 本研究利用國內知名日系超市之資料庫資料，運用層級貝氏統計的普羅比模式，建立專屬於個人的品牌推薦模式，並將該模式用來預測顧客未來可能購買的品牌品項，最後，利用顧客的最後一筆保留消費資料，衡量此推薦模型的準確性。同時，也比較個人化之下的個人層級推薦模型與傳統總合模式之推薦模型的準確度。
並列摘要
With the advent of technology, it is easier for corporations to collect customer data and to develop virtual channel or online stores, which changed tremendously the way people consume today. Therefore, with computing technology, database marketing could help corporations to conduct efficient marketing strategies, to predict future trend and customer behavior, and to actively contact with target customers. However, vigorous virtual channel and online stores tread the neck of physical channel or physical retailers, keeping them barely survive today. Therefore, it is critical for physical channel and retailers to implement database marketing against low cost virtual channel. Database marketing use historical customers’ consuming data to apply one-on-one marketing strategies, attempting to reinforce relationship with customers and customers’ loyalty. The most prevalent execution of database marketing today is the recommendation system. Recommendation system is a platform to suggest customers to buy the products and the products are computed by the system and categorized in highest rating and preference for individual customer. While customers are heterogeneous, via implementing recommendation system, physical retailers could exactly predict the need of customers, control the inventory accurately and gain more bargaining power with branding manufacturers. This thesis used customer data of domestic noted supermarket and applied Hierarchical Bayesian Probit Model to build up recommendation system model. In this system model, each customer has his or her own preference to different brand (in the similar product category). In this way, each customer will receive personal shop suggestion for the next buying. Theoretically, personal suggestions are better than identical ones. The objective of this thesis is try to figure out whether the success hit rate of recommendation system via individual HB Probit model is more higher than the rate of traditional aggregate recommendation model.
參考文獻
7. 黃奕瑛(2000)，金融業導入顧客關係之關鍵成功因素，國立台北大學企業管理學系。
3. 任立中、陳靜怡，”顧客價值遷徙路徑分析：馬可夫鏈模型”，96年台大管理論叢第17卷第二期。
9. Shani, D. and Chalasani, S., “Exploiting Niches Using Relationship Marketing”, Journal of consumer Marketing, vol.9, no3
15. Kalakota, R.& Robinson, M. (1999) “E-Business :Roadmap for Success”
16. Arthur Hughes (1994), Strategic Database Marketing, McGraw-H-ill Inc.