基于SOM神经网络聚类的用气客户全生命周期管理

Full life-cycle management of natural gas customers based on SOM(self-organizing maps) neural network clustering

为了提升市场竞争力,天然气销售企业必须从多个角度对用气客户进行全生命周期管理,以提升客户价值、增强客户忠诚度。决定用气客户全生命周期管理成效的关键就在于能否科学合理地对客户进行分类,而现有的分类方法则未能很好地体现客户价值的现状,不便于对客户进行有针对性的管理。为此,采用SOM神经网络聚类方法,针对天然气的产品特点,选取毛利额、用气时长、用气量增长率等3个指标,对中国西南地区某大型天然气生产企业的546家用气客户进行了实证分析。研究结果表明:(1)用气客户关系的全生命周期可划分为客户识别期、发展期、稳定期和衰退期4个阶段,进而有针对性地提出了识别期开发策略、发展期分级服务策略、稳定期价值提升策略和衰退期终止策略;(2)各个阶段具有不同的营销策略重点,因而能更好地识别和服务于重点及潜力客户,持续提升企业的市场竞争力。结论认为:所建立的方法能更有效、更准确地对用气客户群进行分类,科学合理地对用气客户进行全生命周期管理。

Natural gas sales companies must focus on the full life-cycle management of gas customers from many perspectives so as not only to improve the market competitiveness,but to enhance customer value and loyalty.The key to such solution lies in the scientific and reasonable classification of customers.However,the current classification methods fail to reflect the status of customer value,which is not conducive to targeted customer management.In view of this,SOM neural network clustering was adopted to establish a network learning algorithm,which was then verified by an empirical analysis of546customers belonging to a certain giant gas company in Southwest China.In this scenario,the three important indexes of grass profit margin,time and growth rate of gas consumption were selected to analyze the characteristics of various customer groups and their different values in four stages of customer full life-cycle,i.e.,Introduction,Growth,Maturity,and Decline.On this basis,strategies were put forward respectively at such four stages to acquire and reach customers,provide differential service,enhance customer value,and offset the churn.In particular,different and specific sales strategies should be adopted at each stage so as to identify and serve important and potential customers,thus improving a company’s market competitiveness.In conclusion,this method can classify natural gas customer groups effectively and accurately and manage the customer full lifecycle and lifetime values scientifically and reasonably.