客户期望寿命的乘积限估计

Product limit estimator for the expected duration of a customer lifetime

采用生存分析中的非参数方法估计客户在其每个交易期内的维系概率、客户的期望寿命和客户终生价值。在将客户采购“日历”记录转变成“年龄”记录后,针对数据删失导致的流失、未流失和部分流失等3种客户生存状态,应用K ap lan-M e ier乘积限方法估计其生存概率,进而计算客户的期望寿命。在此基础上,提出结合净收益时间序列或加权平均计算客户终生价值。借助主流统计软件包(如SA S系统)进行乘积限估计,可以有效地实现该文提出的模型。

Nonparametric estimation in the survival analysis were used to estimate the retention probability of customers at every transaction period, and the expected duration of the customer lifetime, i.e., the expected value of the number of customer transaction periods, and the customer lifetime value. Purchase data in the ＂calendar＂ time extracted from customer databases was transformed into records in the ＂age＂ time. The survival probability of the customers （and, consequently, each customer＇s expected lifetime） was estimated by the Product Limit estimator for three survival states-lost, non-lost, and partly lost-caused by censoring of the data. The customer lifetime value was calculated by integrating the expected duration of a customer＇s lifetime with the time series or weighted average of the net revenue from the customer in each period. This model can be efficiently implemented using a Product Limit estimators in popular statistical software packages.