Análise de sobrevivência em marketing, considerando o modelo exponencial com fragilidade compartilhada, na predição de churn

Abstract

Services companies like telephony, health plans, internet and others, have redirect their marketing strategies in the customers retention. By maintaining a business relationship with their consumers, companies guarantee capital collection, reducing spending on finding new users. The study of the occurrence of churn is important for customer retention because, by knowing the profile of these consumers, companies can identify others who may breach the contract. This work aims to predict the probability of occurrence of churn based on the relationship time between customer and company, considering services characteristics and the profile of the consumer, using survival analisys and regression models. The exponential model was used for lifetimes, considering customers divided into different groups, or categories. A frailty term was considered to describe the common risk shared by individuals within the group. With the shared frailty model, it was possible to estimate the probability of churn, at a given time, for the group as a whole, as well as for each of the individuals in the group. The shared frailty model was applied to a database of a phone company, whose analyzes were carried out with two clusters: type of subscription and education level of the customers. With the frailty model estimated for the clusters, it was possible to satisfactorily rank the risk of the groups.