Caracterización del abandono de clientes mediante el uso de algoritmos de aprendizaje automático para empresas de firma electrónica en Ecuador

Abstract

Currently, the use of electronic signatures for various procedures is gaining significant prominence. As a result, there is an increasing number of companies dedicated to this purpose, intensifying competition and prompting the adoption of customer loyalty or retention strategies. It is crucial to consider customer scores in areas such as service quality and recommendations, as well as the department handling their requests and the number of support interactions. By analyzing these factors, it is possible to implement attractive promotions tailored to customer needs. The more precisely these promotions are aimed at the target audience, the more likely customers will choose our product instead of opting for the competition. Unlike traditional processes, which involve contacting all customers without offering discounts or promotions to those at risk of leaving, this work incorporates training a machine learning model to predict customer churn. It also identifies a specific segment of customers with the highest risk of churn and provides an interactive dashboard to monitor customer service operations. As a result, we developed an early detection system for customers with a high likelihood of churn, focusing on the top 10% segment, which yields a 63% economic benefit compared to a random sample of customers. Keywords: electronic signature, customers, early detection, machine learning, churn, promotions.