存貨消耗行為之研究─層級貝氏潛藏行為模型之運用

Abstract

隨著資料庫行銷的發展，擁有顧客交易資料庫的廠商得以藉由分析顧客之購買行為，協助行銷策略之擬定。顧客交易資料庫記錄之購買行為包含購買量、購買期間、購買頻率、購買金額與購買發生等資料。其中，購買量與購買期間對於製造商與零售通路商具有重要之管理意涵。購買量之預測有利於製造商的生產規劃；而零售商亦可從購買量之預測預估需求，進行貨架空間陳設之改善與其存貨規劃，以避免存貨過剩或存貨不足之風險。購買期間之預測則有助於品牌製造商策劃行銷活動執行時間點；零售通路商則得以預估顧客購買時點，並針對目標客群進行行銷活動，以利行銷成本之降低。 本研究以陳靜怡（2005）建購之層級貝氏潛藏行為模式，分析國內知名超市的顧客資料庫中白米消耗行為。過去有關購買量與購買期間之研究，大多假設這兩個行為變數彼此獨立。該層級貝氏潛藏行為模式經由存貨消耗行為整合購買量與購買期間。本研究之實證分析驗證該模式之效度，並與傳統統計之最小平方法估計結果作比較，證實層級貝氏潛藏行為模式之穩健。本研究之實證分析發現人口統計變量並未能顯著解釋顧客購買白米之行為，其可能原因於最後之研究限制與未來研究方向中說明。

With the development of database marketing, firms owning customer transaction database are able to analyze customer purchasing behaviors that are helpful in the formulation of marketing strategy. Such purchasing behaviors as purchase quantity, inter-purchase time, purchase frequency, purchase amount and purchase incidence are comprised in customer transaction database; wherein, purchase quantity and inter-purchase time have significant managerial implications to manufacturers and retailers. Prediction of purchase quantity is helpful for manufacturers in production planning while it provides an estimation of customer demands to improve the display of shelf space and inventory control that avoids the risk of excess inventory and shortage in inventory. Moreover, the forecast of inter-purchase time is favorable for manufacturers in scheduling marketing activities whereas predicting purchase timing is advantageous for retailers to select target customers of a given marketing campaign, leading to effective cost reduction. Latent hierarchical Bayes model developed by Chen (2005) was employed to analyze the scanner data from a noted supermarket in Taiwan in this research. In past decades, researches related to the modeling of purchase quantity and inter-purchase time assume that these two variables are independent. However, the latent hierarchical Bayes model integrates purchase quantity and inter-purchase time through inventory consumption behaviors. The empirical research in this study verified the excellence of the model along with a comparison with the performance of OLS estimators. The results of empirical research showed that demographics did not significantly explain the behaviors of customers purchasing rice and the possible explanations are presented in the last chapter.