营销科学学报 ›› 2011, Vol. 7 ›› Issue (3): 102-110.

• 论文 • 上一篇    下一篇

基于Beta-geometric/NBD模型的客户流失预测分析

王谢宁   

  1. 王谢宁,东北财经大学管理科学与工程学院讲师,E-mail: wangxnlm@yahoo.com.cn
  • 出版日期:2011-09-01 发布日期:2013-11-01
  • 基金资助:

    感谢匿名评审专家对本文所提出的宝贵的修改意见和建议,作者文责自负。

Prediction and analysis of customer loss based Beta-geometric/NBD model

Wang Xiening   

  1. Wang Xiening, institute of Management Science and Engineering, DongBei University of Finance and Economics
  • Online:2011-09-01 Published:2013-11-01

摘要:

客户交易行为预测和研究对于企业营销有着较大的应用价值, Beta-geometric/ NBD模型能够清晰地依据客户历史购买行为对未来购买行为进行预测,其较为严格的假设条件使得它在实践应用上存在一定的局限性。本文主要围绕模型在营销领域的实践应用展开,主要对实证过程中出现的问题进行分析和对模型参数的优化,并对模型的关键结果进行二次扩展利用,构建出基于计算客户流失率的设置客户流失预警点的方法;实证结果表明模型在中国BtoC营销相关领域对预测客户购买行为有较高的应用价值;模型更适合于购买频率较高的情况,对于交易次数很少而且距离第一次购买时间不长的客户数据,其预测结果对比真实值有一定的偏差。最后,讨论了在实践应用时模型对于其条件假设限制下的适用范围和研究的不足。

关键词: Beta-geometric/NBD模型, 客户价值, B2C网站, 流失预警

Abstract:

It is of great practical and applicable value for corporation marketing to forecast and research the behavior of customer transactions. Beta-geometric/ NBD model can forecast customer’s potential purchasing according to his/her past buying behavior. However, there are some limitations in using this model due to its restricted assumptions. This paper focuses on the model’s practical application in the field of marketing, empirical analysis and model parameters optimization, as well as utilization of intermediate results to build an early loss warning methods. The empirical results show that the model have a high application value in forecasting customer’s buying behavior in China BtoC marketing related fields. This model is more suitable for the customer who has high frequency purchasing history. Relatively, the prediction outcome has certain bias for those who have low transactions frequency and stay not a long time from the first purchase. Finally, the paper discussed the applicable field and scope of the model under its assumptions and indicated the deficiencies of the model.

Key words: beta-geometric/NBD model, customer value, B2C website, loss warning