Optimizing churn through predictive interventions for customers with high propensity to cancel
Beneficios y resultados
Antecedentes
Liberty Latin America operates in over 20 countries across Latin America and the Caribbean, offering a wide range of communications and entertainment services under brands like Flow, Liberty, Más Móvil and BTC. The company provides "triple-play" and "quad-play" services, including digital video, broadband internet, telephony, and mobile services. Additionally, Liberty Latin America is noted for its extensive sub-sea and terrestrial fiber optic cable network connectivity services.
Desafíos
Issue Identification: Due to Liberty's business model of offering prepaid telephone services with no contracts, no credit checks, and no activation fees, subscriber churn was inevitable. High subscriber churn resulted from a lack of insights into customer behaviors and cancellation patterns, compounded by the absence of predictive capabilities in identifying potential churn towards the end of the service period; Issue Impact: Ineffective churn mitigation strategies resulted in substantial revenue losses and low profitability for Liberty
Solución
NowVertical's Role: NowVertical analyzed historical transaction and usage data from both active and churned customers. A machine learning model identified distinct behavioral patterns among churned customers, which were then applied to current subscribers to predict their likelihood of canceling at the service period's end. This approach enabled early identification of at-risk customers, providing a window for targeted interventions to prevent cancellations.
Implementación
Data Collection and Preparation: Gathered historical transaction and usage data from both active and churned customers. Cleaned and preprocessed the data to ensure quality and consistency for analysis; Data Modeling: Developed comprehensive data models to structure and organize the collected data, facilitating easy access and analysis; Machine Learning Development: Utilized data science techniques, including Random Forest and LightGBM methodologies, to identify distinct behavioral patterns among churned customers and apply these patterns to current subscriber data; MLOps Integration: Integrated MLOps practices to streamline the deployment, monitoring, and maintenance of the machine learning model; Prediction and Identification: Implemented the machine learning model to analyze current subscribers' data in real-time and identify at-risk customers early; Targeted Interventions: Provided actionable insights to CRM and marketing teams through high-performance dashboards, enabling targeted interventions to retain at-risk customers
.svg)
.svg)
