直接市场营销中以重复购买理论为基础地一种新地客户选择方法.docxVIP

rsometimes called a“gains table”(Jackson and Wang，1994)or a“decile analysis” rShepard，1 998))．The gains chart is computed by sorting the data in descending order ot scores and customers with similar score values are divided Into groups of equal size fusually deciles)Then，average and cumulative average response rates or dollar amounts per group are calculated．Also，the】m value—the decile response rate divided by the total response rate．and the cumulative lift value are calculated． To evaluate the scoring models and the gains charts generated，Malthouse(200t) examined the single．split gains chart method．To create the gains chart，direct marketers always split the available data into“training”and“test”sets．estimated their models on the training set and then apply it to the test set．But it was known that sampling variation across splits affect the judgment of the model．So，the author suggested two ways of reducing the variation across splits：Winsorization and s仃atified sampling．Empirical study on fl data set showed that both the methods could reduce the variation significantly． When using direct mail as a medium for communicating with potential buyers，deciding how many customers to mail is a complicated decision in the context of a broader business plan．The common practice is sending mail to those people in the mailing list， whose expected returns exce酣the marginal costs of the mailing．Or．when deciles are used，to send mails to all people thai belong to the deciles with average expected returns larger than the marginal costs． Bult．J．R．and Wansbeek，T．f1995)further refined the selection procedure by using maximized expected profit．The authors introduced a comprehensive methodology for the selection of targets from a mailing list．Thev determined which households should receive a mailing by equating marginal costs and marginal returns in order to maximize expected profit．And the authors claimed that at 1east theoretically．their method 1ed to more efficient selection proce