【作者】 王海伟；

【导师】 王雅林；

【作者基本信息】 哈尔滨工业大学 ， 技术经济及管理， 2007， 博士

【摘要】 客户是企业生存与发展的基础,在竞争日益激烈的环境中,获取并保持优质的客户资源成为企业关注的焦点,也因此成为营销理论研究者关注的热点。企业对客户资源的重视导致了将资产的概念延伸到客户的说法,从本世纪初开始,客户资产开始受到学者们前所未有的关注,客户资产价值的衡量、客户资产价值构成以及客户资产战略经营成为这一领域的研究热点。然而和金融资产相似,客户所带来的现金流具有不确定性,也是一种高风险的资产。企业希望能够跟客户保持一种长期的、稳定的关系,带来源源不断的现金流,但是客户并不总是保持购买规模与频率的一贯性,因此现金流有很大的波动性,它取决于企业与客户之间的关系水平,同时也受到客户内外部因素的影响,只有对其风险形成机理进行掌握,并使其风险值得到有效的度量,才能对风险进行有效控制,否则客户资产将是无意义的资产。而目前的研究中鲜有对客户资产风险进行研究的文献,理论研究的不足及现实的需要成为本研究的主要动力。本文针对国内客户关系管理领域概念混淆的现状,对相关概念进行了梳理,在此基础上分析客户的资产特性及其风险特性,提出客户资产风险的概念。并依据此概念对客户资产风险相关理论进行剖析,从宏观环境层次、行业竞争层次、企业内部层次、客户自身层次分析非契约型客户资产风险成因。继而以“问题出现——退出评估——沟通行为——关系挽救——退出行为”为客户风险产生的因果主线,利用“情感”概念体现客户对这些“行为”的决策,建立客户资产风险形成过程模型,并提出不同生命周期阶段下客户对“行为”的“情感”反应,继而从关系衰退角度构建客户风险阶段动态变化过程模型。接着针对这一过程提出客户资产风险驱动要素及其调节因素,构建客户资产风险理论模型,采用结构方程建模对客户资产风险理论模型进行验证,利用分组线性结构方程模型的方法对相关影响因素的调节效应进行分析,发现负面归因、转移成本、竞争者吸引力、多样性寻求心理、生命周期阶段、主观规范等因素对客户资产风险驱动要素的因果关系有调节作用。最后在实证研究的基础上对客户资产风险理论模型进行修正,系统阐述客户资产风险形成机理。然后,本文将客户资产风险分为流失性风险、衰退性风险、波动性风险三个层次,提出这三个层次风险的度量方法,并首次采用双变量多层贝叶斯模型对客户购买间隔和购买金额联合分布进行拟合,模型假设客户购买行为和购买金额服从双变量对数分布,利用“正态—Wishart”共轭先验分布族对客户购买间隔和购买金额后验近似标准分布进行推导,借助梅托普利斯-海斯丁算法求解无法推导出后验近似标准分布的参数,从而对购买间隔及购买金额进行预测。基于此拟合分布,可求出客户资产波动性风险、衰退性风险、及其流失性风险值。最后本文通过对一高分子企业进行实证研究,发现双变量多层贝叶斯模型在所有预测检验上都优于传统预测方法,特别是对客户购买金额的预测体现出更加出色的预测效果,因此该模型能够较好拟合客户购买行为。从模型优化角度来看,本文设计了7个竞争模型,实证发现随着模拟次数的增加,预测能力显著增强;变量的增减对于模型影响不大,这说明模型预测准确性并不依赖于先验信息统计变量的选择,对于目前我国企业数据库不健全情况下,这一特点使该模型相比数据挖掘模型适用性更强。更多还原

【Abstract】 Customers are the foundation of an enterprise’s development. Under the increasingly severe competitive circumstances, it has become the key point of an enterprise to get and sustain high quality customer resources, and also the point of researchers on marketing theory. Paying attention on customer resources had led to the saying that is the stretch of asset’definition to customers. From the early beginning of this century, scholars had paid great attention to the customers’asset, customer asset’weight, structure and management has become the hotspot of this frontier. But according to the similarity of financial asset, the cash flow that customers had brought is uncertain, and it is also a kind of high-risk asset. Company want to keep the long and stable relationship with customers to bring the high cash flows, but customers do not always keep the persistent of scale and frequency, so cash flows is fluctuated, it depends on the level of relationships of company and customers, it is influenced by all kinds of external factors as well, so only to acquire its mechanism of risk, and to make its value to be well measured, we could control the risk, or customer asset are meaningless.This dissertation deals with the related definitions, analyze the identity of customer asset and its risks, bring forward the definition of customer asset risks based on the realistic of the mixture of definitions of domestic customer relationship management. According to this definition, we research on the related theory of customer asset risks. From the macro-environment, industry competition, company internals and customers themselves, analyze the cause of uncontract customer asset, and then from fail incident, exit attitude, communicate deed, saving relations to merchandise deeds. All of these become the main line of the cause of customer risks, using affections to demonstrate the reflection of customers to these incident and deeds, in the end to structure the dynamic customer relationships’process from the ankle of the relations decline. According to this process, this dissertation brings out the factors of customer asset risks’driving and adjustments; structure the model the customer asset theory. Appealing to structural equation modeling to validate the direct and indirect relations of this new model, using grouping linear structure model to analyze the related influential factors, to find negative attribution, transferring costs, attraction of competition, variety-seeking, Life cycle phase and subject normal, these factors could adjust the consequence relation of customer asset risks driving factors. Finally based on the experimental research, revise the model of customer asset risks theory; systematically expatiate the mechanism of customer asset risks. This dissertation divides the asset risks into churn risk declining risk and fluctuant risk. Bringing out the measurement for these three methods, and appeal to the model of double variable multi-layer Bayes to approximate the customer interpurchase time and purchase amount,. Model is based on the hypothesis that customer purchase behavior and purchase amount obey the log-normal distribution and use“Normal-Wishart”distribution to deduce the posterior distributions of customer’s purchasing interval and purchasing amount. By using Metropolis-Hastings algorithm to deduce the approximative parameters in posterior standard normal distribution, this dissertation forecasts the customer’s interpurchase time and purchase amount. Based on the approximated distribution, we can solve the fluctuant risk and declining risk as well as churn risk values. In the end, this dissertation makes an empirical research on an enterprise of macromolecule industry. We find Bivariate Hierarchical Bayesian has more advantages over the traditional forecasting methods, especially presents excellent forecasting result on customer purchase amount. So we can say the model can approximate the customer purchase behavior well. From the aspect of model’s optimization, this dissertation designs 6 competitive models and finds the forecasting performance becomes more excellent with the simulation iterations increasing. Besides, we find the number of the variables has no distinct impact on the model which indicates the forecasting accuracy has no direct relations with the statistical variables selection of the prior information. At present, there are no perfect database systems in China’s enterprises. Under this circumstance, the model shows more applicable than the model of data digging.更多还原