Skip to content

6) Project objectives addressed

🎯 6.1) Objetive 1. Detect key drivers of churn

"Detect key drivers of churn"

✔️ Implemented Using Feature Importance - The project used the GradientBoostingClassifier, a tree-based ensemble method that provides a reliable feature importance metric. - A bar chart of the Top 10 most important features was generated. - This helps identify attributes like Contract, TechSupport, and TotalCharges that most influence churn. - Enables marketing and CRM teams to understand why customers are leaving.

Detect key drivers of churn

🔍 Business Insight (Detect key drivers of churn)

Top Predictors Identified (Gradient Boosting Feature Importance)

  1. Contract – By far the most influential factor. ▸ Customers with flexible (month-to-month) contracts are much more likely to churn than those on long-term agreements.

  2. MonthlyCharges – Highly predictive of churn. ▸ Customers paying higher monthly fees are more prone to leave, possibly due to cost sensitivity or unmet expectations.

  3. OnlineSecurity – Strong negative correlation with churn. ▸ Customers without online security are more likely to churn, suggesting this service may increase customer stickiness.

  4. Tenure – Loyalty indicator. ▸ Customers with shorter tenure are more likely to churn. Long-standing customers are less at risk.

  5. TechSupport, OnlineBackup – Meaningful contributors. ▸ Add-on services such as tech support and backup correlate with lower churn rates, likely due to enhanced service engagement.

  6. InternetService, TotalCharges – Relevant, but less dominant. ▸ Fiber optic users may churn more, and lower total charges often reflect newer, high-risk customers.

  7. Dependents, Partner – Moderate influence. ▸ Customers with no dependents or partner tend to churn more, possibly due to lower stability or loyalty.

📈 Business Insights

  • Contractual flexibility is the most powerful driver of churn. Month-to-month users should be prioritized for retention incentives or upgrades to longer-term plans.
  • Value-added services like OnlineSecurity and TechSupport show a protective effect. Bundling them in promotional packages could reduce churn.
  • High-bill customers are more volatile. Consider delivering personalized experiences or loyalty rewards to increase perceived value.
  • Demographic features (partner/dependent status) can help create behavioral customer personas for targeted retention campaigns.

🎯 6.2) Objetive 2. Estimate individual churn probabilities (scoring)

"Estimate individual churn probabilities (scoring)"

✔️ Implemented Using Model Predictions - The model outputs individual churn probabilities with predict_proba(). - Probabilities are stored in a DataFrame and sorted. - This allows teams to identify high-risk customers and act proactively. - The top 10 churn-prone customers are highlighted.

Churn Scoring

🔍 Key Insights – Churn Scoring (Objective 2: Estimate churn probability - scoring)

  1. High accuracy in identifying churn-prone customers Among the top 10 customers with the highest predicted churn probability, 9 actually churned (Actual_Churn = 1). This shows the model is highly effective at capturing relevant patterns.

  2. Very high probabilities (above 0.92) Most listed customers have probabilities above 0.93, indicating that the model produces highly reliable risk scores for prioritization.

  3. Direct applications for personalized retention Identifying high-risk customers enables proactive actions by Marketing and CRM teams, such as:

  4. Offering personalized discounts;
  5. Launching re-engagement campaigns;

  6. Prioritizing high-risk customers for support.

  7. Segmentation and automation of actions The churn scores can be combined with CLTV or demographic profiles to segment campaigns, increase ROI, and reduce churn strategically.

  8. Potential for integration into dashboards and CRM systems The scores can be exported or visualized in dashboards for real-time churn risk monitoring and decision-making.

🎯 6.3) Objetive 3. Prioritize retention actions based on churn risk and CLTV

"Prioritize retention actions based on churn risk and Customer Lifetime Value (CLTV)"

✔️ Implemented Using Priority Score - A new score was calculated: Retention_Priority_Score = Churn_Probability × TotalCharges - TotalCharges acts as a proxy for CLTV. - Customers are segmented by quartiles of priority. - Enables focused retention strategies on high-value, high-risk customers.

Retention Prioritization Based on CLTV

🔍 Key Insights (Objective 3: Retention Prioritization Based on CLTV - Customer Lifetime Value- (TotalCharges)):

  • The Retention Priority Score effectively ranks customers based on both the likelihood of churn and their financial value to the company.
  • The top 10 customers with the highest retention priority were identified, combining high churn risk with high CLTV.
  • These high-value at-risk customers represent critical targets for personalized retention strategies.
  • This approach allows marketing and CRM teams to:
  • Allocate resources efficiently to customers who matter most.
  • Design proactive campaigns aimed at preserving long-term revenue.
  • Segment the base by quartiles of retention score for scalable action plans.

Quartile Segmentation

Quartile Segmentation Insights

Customers were divided into quartiles based on their Retention Priority Score:

Priority Quartile Avg. Churn Probability Avg. CLTV
Very High 0.482956 4018.60
High 0.365823 2484.32
Medium 0.316204 1774.69
Low 0.218525 559.90

🔍 Key Insights

  • Very High Priority Segment has the highest churn probability (48.3%) and the highest average CLTV (~$4,000), indicating these are high-value customers at high risk of churn.
  • High and Medium segments show moderate churn probability with substantial revenue potential, suitable for cost-effective retention actions.
  • Low Priority Segment includes customers with lower churn risk and low revenue contribution, suggesting less urgent action is needed.

⬅ Voltar para a Página Inicial