Desarrollo de un Modelo de Predicción de la Deserción de Clientes de una Empresa de Medicina Prepagada utilizando Análisis de Supervivencia

Abstract

Companies today place more importance on customer retention. Attracting a new consumer can cost up to five times what it costs to retain an existing customer. Economically it is much better to retain customers who can attract new customers and, in addition, it generates much more value to the company, not only the economic benefit. The present work is focused on predicting, through the Cox Model, the desertion of clients of a Prepaid Medicine company.

This model is used through the R software, mainly using the “survival” package and secondarily the “survminer” package. The “survival” package allows not only the development of the model itself, but also the validation of the assumptions of the Cox model and the evaluation of the effectiveness of the model. The “survminer” package mainly allows us to visualize the results (survival curves, censorship diagrams, among others) in a clearer way and also contains other functions available to plot fitted curves for the Cox model and visually examine the assumptions of the same.

Some of the results obtained coincide with studies carried out in other countries on customer desertion, not exactly in the same sector of prepaid medicine. It will also be seen that the difficulties in predicting the behavior of customers are the same in one industry and another, keeping of course the particularities of each one. The models obtained were validated with a test base and each one had its own particular findings.