Enhancing the methods of customer behavior analysis to increase customer satisfaction. Case study: Syriatel Telecom Company

Abstract

The telecommunication industry is in the strongest competition ever, as this sector gets disrupted by new arising competitors with high technical infrastructure as 5G networks. However, the current customer satisfaction measures are based on subjective questionnaires without utilizing the vast amount of objective network KPIs and telecom systems data into account. This work presents a model that tackles this lack of research and provides a high impact solution to survive in the tough competition of the telecom industry. The paper addresses two fundamental questions: 1) To what extent satisfied/dissatisfied customers can be classified based on telecom systems data that was produced during users’ interactions? 2) Can satisfaction indicators be derived from telecom systems data? this study discusses a machine learning problem, and compare 7 classifiers and analyze data for 10,000 real users from the Syrian telecom company Syriatel. 120 extracted features were drawn from the most significant available sources: billing, network, customer service system, and customer demography data. The best result for customer satisfaction classification was 87%, achieved with XGBOOST classifier. Furthermore, the paper identifies the most 9 potential indicators for satisfaction. Our goal was to classify customer satisfaction/dissatisfaction based on the objective data that is generated from every service interaction on the network or customer care centers.