本研究嘗試以汽、機車顧客之加油行為作為資料庫行銷的實證研究對象,藉由資料庫中每位顧客的購買歷史紀錄,包括購買日期、每筆購買金額等,與人口統計變數資料,包括顧客編號、性別、年齡、職業與居住地區等,進行購買日期特性分析、平均購買期間分析、顧客活躍性分析、顧客價值分析與穩定性分析。 在購買日期特性分析中,本研究以兩種日期特性分類分別檢定其差異性:一周七日分類與週休、假日、平日分類。在推估顧客平均購買期間時,先以傳統的最大概似估計法與加權最大概似估計法為先,比較兩者差異以判斷整體顧客活躍性後,再與層級貝氏估計法(HB)做比較,HB的特性在於不僅考慮可觀察的每人購買率之不同,更考慮了無法觀察的顧客異質性,可修正因個別樣本交易紀錄過少而產生的偏誤。 再由購買次數與活躍性指標得出顧客活躍性分析、以購買次數與平均購買金額得出顧客價值分析,最後以購買次數與穩定性指數得出顧客穩定性分析,並透過不同的行為特性做為顧客分群之依據。 除了上述資料庫行銷分析外,本研究並對機車與汽車顧客的顧客行為特性差異做比較,結合各項分析與不同分群依據,針對企業總體利益,提供對應的策略建議,以作為企業發展行銷策略的參考。最後,結論與建議彚整各章之研究發現,並說明研究限制與未來研究方向。
The main purpose of this research is to help the company establish a better understanding of its customers, figure out the heterogeneity of each one, and adopt one-to-one marketing strategy rather than mass marketing strategy. Thus, we use customers’ past transaction data and some database marketing analysis techniques to identify each customer more precisely. First of all, we conduct the analysis of consumption behavior in variation of time according to two set of time frame. An estimation of the inter-purchase time of individual customer includes Maximum Likelihood Estimation(MLE), Weighted Maximum Likelihood Estimation(WMLE),and Hierarchical Bayesian Model(HB). The main advantage of HB is that HB considers the unobservable heterogeneity of individual customer and hence forecast its future behavior more accurately. Furthermore, from Customer Active Analysis, Customer Value Analysis, and Customer Reliable Analysis, we can segment all customers into groups by consuming behavior and make various marketing strategies corresponding to different groups. Employing the purchasing records of every gas station of a domestic leading petroleum brand, last but not the least, we compare the different fueling pattern between the motorcycle and car customers to help the company identify its target markets and do its promotion efficiently.