| Summary: | 碩士 === 南台科技大學 === 企業管理系 === 99 === There is increasingly intensive competition in IT software market of Taiwan, and customers’ loyalty is a major profit support of a software company. In the software industry, the rule of thumb for selecting target customers is the experience of sale representatives. However, the use of experience rules is not a guarantee for sale strategy for a software company. Therefore, this study tried to construct a predictive model based on historical purchase transation records to help identify the target customers who possibly re-purchase software products. The contructed predictive model and extracted experience rules will help capture possible sale opportunities and establish proper sale strategy for the company.
This study followed a retrospective research method, and collected 281 complete historical sale records from 2000 to 2009 from a well-known software company in Taiwan. The collected cases are described with RFM measures used in RFM customer value model. This study employed two decision tree methods, including C4.5 and PART, and a CAR method (classification association rules) to construct the predictive model for identification of target customers. According to empirical results, three variables, including "Transaction situation in latest three years (R1)", "Amont of a specific decision-support software transaction (M2)", and "Areas of customers (AREA)", are the main elements of extracted experience rules and the constructed model.
According the results of this study, we found that the data mining approach is able to extract useful experience rules from historical sales records. If we combine the sales’ experiences and extracted association rules to build the integrated target customer identification rules, the built rules will be the important basis to draft sale plans for the sale persons.
|
|---|
