随机RFM模型及其在零售顾客价值识别中的应用

Stochastic RFM Model and Its Application in Customer Values Analysis

对顾客进行价值识别是以顾客价值为导向进行差异化营销管理的基础和前提。本文通过建立随机RFM模型,提供了一种对未来顾客价值进行识别的模型方法,并利用一家购物中心的顾客交易数据对该方法进行了实证分析。结果表明随机RFM模型不仅是进行顾客价值识别与分析的有效工具,也是研究分析顾客购买行为的有效模型方法。该模型方法对于零售企业加强顾客价值分析,提高基于顾客价值的管理水平具有积极的实践意义。

Customer relationship management programs have the merits of targeting valuable customers and distinguishing between valuable customers and non-valuable customers.The RFM（Recency-Frequency-Monetary） model is the most common method used in CRM programs.RFM often asserts that customers with higher shopping frequency and larger shopping volume are the most valuable customers.However,the traditional application of RFM ignores the nature of customer heterogeneity.A deeper understanding of customer heterogeneity can lead to more accurate predictions of customer demands.Customer heterogeneity is a valuable source of ideas for creative marketing strategies.One effective way of overcoming current RFM problems is to model buyer behaviors based on the stochastic method.The goal of this paper is to propose a stochastic method to estimate customer values and conduct empirical validation.This research consists of three steps.Step 1 is to propose the underlying probability distribution of customers＇ purchasing behaviors and construct a stochastic method to model customers＇ purchasing behavior.Step 2 is to use the stochastic method to model customer R（Recency）,F（Frequency） and M（Monetary） behaviors,predict customer future responses and estimate customer value.Step 3 is to validate the proposed model using empirical data collected from a panel of customers in a shopping center in Beijing.Data collected are split into two sets： 6 months of training data and 6 months of testing data.A cluster analysis is conducted to cluster the participants into four homogeneous groups,followed by customer＇s value analysis.Analysis results are then compared between the model＇s predicted value and the actual value.Several conclusions are made in this study based on the analysis results.The proposed model outperforms the traditional methods and achieves a very accurate estimation of customers＇ values.The modeling method has great value to retailers in enhancing customer value analysis and improving customer relationship management.The model presented in this study is a good method of analyzing customer behavior because it can use observations of past responses to accurately predict future responses.In addition,the modeling and estimating processes are straightforward and the required data can be easily collected.The proposed model is also a great tool to dynamically monitor or track the development of customer value.Managers can modify their customer segmentation and marketing strategy according to the development of customer values.The proposed modeling method is not only an effective tool to estimate future customer values but also can be used to estimate the performance of special marketing tools（such as advertising and large-scale promotion）.The model can use data from the last period to estimate the theoretical norms of customer behaviors in subsequent periods.This estimation is not affected by the marketing tools.We can then compare the actual customer behaviors with the theoretical behavior norms.The effect and performance of the marketing tools can be assessed based on the discrepancy between theoretical values and actual values.These findings provide theoretical and practical implications.