Research On Survival Analysis Model Of E-commerce Customer Churn And Retention Strategy

How to effectively reduce the customers'churn willingness is of great significance to e-commerce companies as the customer churn rate remains high.Although implementing mass advertising or optimizing products is easy to carry out,these retention measures are untargeted that will expose the companies to the risk of wasting resources,even dissatisfy the customers and expedite them to churn.However,with the era of big data coming,it is easier to collect the customers'behavioral characteristic data to model,which is beneficial for the e-commerce companies analyzing the customer churn events and designing the scientific retention strategies to prevent customers to churn,also provides a new path for e-commerce companies to understand the customers'consumption behaviors and respond tothechangesinthemarketdemandmoreeffectively.According to the stochastic characteristics of e-commerce customer churn events,a survival analysis model of e-commerce customer churn is constructed in this thesis based on 2037 pieces of real data.In addition,this thesis also proposes a mathematical model to solve the decision issues of customer retention,such as whom to target,target order,the cost of incentives,target size,etc.The mathematical model's operability and effectiveness are successfully verified using the sample data,it can help the e-commerce companies avoid committing a type I error?target a customer who does not need to be retained?and a type II error?not target a customer who needs to be retained?.Through research,the following conclusions have been obtained:?1?The method of survival analysis can effectively deal with the censored data,while simply deleting the censored data and directly treating the censored data as complete data are both inappropriate because of underestimating the e-commerce customer's survival rate.The survival distributions of two groups in a given consumption characteristic are statistically significant.The customers with frequent purchases,being the WeChat store's fans,locating in the province,having got benefits from the company's discount activities,having left messages when submiting the orders,having returned goods for rights protection and large monetary fluctuation in the amounts of consumption generally survive longer than who are not.?2?The three Cox regression models constructed in this thesis are identical in the recognition of the"risk factor"and"protective factor"that affect the risk of customer churn.The frequency of purchasing,customer's location,order messages,rights protection and the monetary fluctuation have good prognosis of customer churn.Morever,being the WeChat store's fans and having got benefits from the company's discount activities are good for lengthening the customers' survival time and reducing the risk of customer churrn,which fully shows that attracting customers to be the WeChat store's fans and giving customers discounts are still effective marketing strategies in the e-commerce environment.Especially,the coefficient of the covariate"average internal"is not constant,but a time-dependent function with the form of?₉?t?=-0.21+0.031og?t?.?3?The ultimate goal of the customer churn management is to maximize the profits from retaining.There are heterogeneity in the profits contributed from customers,each customer's profit depends not only on his propensity to churn,but also on his value,his probability of responding to the retention incentives,as well as the cost of incentives,the overall profit of the company also depends on the number of customers the company decides to target.Compared to the traditional method of ranking customers based on their propensity to churn,this thesis presents the profit increment as the ranking indicator,which can bring more benefits to the company without any additional retention costs.Therefore,the effectiveness of the proposed model in this thesis has been verified.