Skip to content

🧠 9) Final Project Conclusion – TelcoRetentionAI

After completing the end-to-end CRISP-DM pipeline, the TelcoRetentionAI project successfully met both its primary and secondary business objectives:


🎯 Primary Goal: Predictive Modeling for Churn

  • Among all tested models, the Gradient Boosting Classifier provided the best overall performance in terms of:
  • ROC AUC and F1-Score, ensuring high predictive power.
  • Stability across resampled datasets (SMOTE) and unseen test sets.
  • Interpretability through feature importance and SHAP explanations.

  • As such, Gradient Boosting is recommended as the primary model for churn prediction in production environments.

  • Logistic Regression remains a strong candidate when explainability and calibration are critical for business or compliance reasons.

  • Random Forest provides consistent performance and can be used in ensemble strategies to enhance model robustness and generalization.


📊 Secondary Goal: Business Insights and Recommendations

Visual and descriptive analytics enabled the discovery of actionable insights that support business decision-making:

  • Contract Type: Customers with month-to-month contracts showed the highest churn rate. Incentives for migrating to longer contracts are recommended.
  • Tech Support: Lack of tech support correlates with high churn. Upselling or bundling support services can improve retention.
  • Paperless Billing: Tech-savvy users on paperless billing tend to churn more. Loyalty perks and engagement tracking are key.
  • Senior Citizens: This group showed a proportionally high churn rate. Tailored support and simplified processes are advised.
  • Partner Status: Customers without a partner are more likely to churn. Campaigns should focus on personalization and independence.
  • CLTV Quartile: Customers with low lifetime value are the most volatile. High-CLTV churners should be prioritized for retention.

🧠 Strategic Impact

  • The project not only produced a high-performing AI model, but also delivered clear, explainable business intelligence.
  • Insights were visualized in dashboards and supported by data, guiding targeted retention strategies.
  • The combination of machine learning + human-readable analytics provides a solid foundation for data-driven decision making.

🚀 Next Steps

  • Deploy the Gradient Boosting model and monitor real-time churn predictions.
  • Integrate churn risk scores with CRM systems to trigger automated retention campaigns.
  • Refine feature engineering and retrain the model periodically with updated data.
  • Track ROI on retention campaigns to measure business impact.

TelcoRetentionAI represents a complete AI-powered solution for customer retention in the telecom industry, bridging the gap between data science, business strategy, and operational excellence.


⬅ Voltar para a Página Inicial