Study On Formation And Measurement Of Uncontract Customer Asset Risk

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.