隨著資訊科技的提升與資料庫分析技術的發展,加速與普及企業對顧客消費行為資料的應用,企業若能由顧客消費行為中,篩選具有高價值的顧客,找出潛在顧客,區分不同顧客群之消費特性,就可發展適宜的行銷與服務策略。在企業有限的資源應用中,依顧客特性做最適的資源分配,以有效提高顧客購買頻率與消費金額,延長顧客關係,達到提升企業營運績效、降低營運成本的功效。基於近年來成人教育訓練服務業的興起,各項職場專業能力訓練機構林立,其經營方式趨向企業化,教育訓練逐漸被視為一項服務商品,這個趨勢尤其在職場專業能力訓練中特別明顯。因接受訓練服務是以顧客個人為單位,其消費行為記錄適合作為顧客關係管理之分析。本研究即以教育訓練機構個案之顧客消費行為紀錄作為分析研究資料,利用最近購買日(recency)、購買頻率(frequency)、購買金額(monetary amount)三個變數以兩階段集群法(Two-stage clustering)與自組織映射圖網路(Self-Organizing Map, SOM)分群方法分析,比較不同分群方法之優異,找出較適之顧客分群結果。本研究個案中,發現兩階段集群法(Two-stage clustering)分群效果較佳,自組織映射圖網路(SOM)分群方法則需結合K-means進行兩階段分群,可有效改善自組織映射圖網路(SOM)分群結果之發散現象,分群效果與兩階段集群法(Two-stage clustering)類似。在研究個案中,將顧客消費行為資料運用上述統計分析技術能有效進行顧客分群,共將個案資料分為五個群組,分別為高價值頻率型顧客、頻率型顧客、潛在型顧客、高價值新進顧客、高價值單次型顧客,並發現此個案顧客消費特性為消費頻率高者,消費金額偏低;消費金額高者,消費頻率則較低;消費金額高於全體平均之群組其年齡亦高於全體平均等。最後利用區隔之顧客群組,找出同群組間共通的特性,對於不同消費特性的顧客群,提供不同的差異化行銷及服務策略建議,作為顧客關係管理的基礎,以利提高企業利潤、降低營運成本。
The development of information technology and database analyzing skill has stimulated and popularized the application of customer consumption behavior data on enterprise. If the company can recognize high-value potential customers and group customers by their characteristic of consumption behavior, they can further develop suitable marketing and service strategies for their customers. The enterprise can conduct optimal allocation based upon differentiated customers’ characteristics to raise customers’ purchase frequency, increase their consumption amount, and extend the relationship with customers as well under the limitation of enterprise’s resources. That can also increase company’s operating profit and reduce the operating cost. Along with the developing of service in adult education training, many kinds of professional occupation training mechanisms have been established. Education training, especially in professional occupation training, is treated as a services product. Because the object of training service is for individual customer, the records about consumption behavior of customers are very suitable for the research on management of customer relationship. In this thesis, two-stage clustering and self-organizing map (SOM) methods are used to classify customers and to find out the optimal customers’ groups based upon customers’ consumption behaviors in the case of education training based upon three variables, namely, recent purchasing date, purchasing frequency, and purchasing amount. The study results show that the two-stage clustering has the best performance , while the self-organizing map (SOM) working with K-means could effectively improve the grouping performance of the self-organizing map (SOM) and its grouping efficiency is very similar to that of two-stage clustering method. The study results also show that the proposed statistical methods by this thesis can effectively grouping customers into five groups based on customers’ consumption behavior. The classified five groups are named as high-value frequency customer, frequency customer, potential customer, high new customer and high-value one times customer. In our study case, we also find that customers’ consumption behaviors have the characteristics that high frequency (amount) customers have lower consumption amount (frequency). Besides, the customers with higher consumption amount have higher age. Finally, we extract the common factors for differentiated groups and provide suggestions about marketing and service strategy for each group based upon their consumers’ behavior characteristics. The study results could provide the basis for consumers’ management to increase company’s profit and also reduce the operating cost as well.