Skip to content

Project Goals

Primary Goal:

Develop a binary classification model to predict whether a customer will cancel their service in the next billing cycle.

Secondary Goal:

Generate business insights through descriptive and inferential analytics, including dashboards that highlight:

  • Impact of features such as contract type, tech support, billing profile, etc.;
  • Demographic and geographic patterns among high-churn customers;
  • Actionable recommendations for targeted customer segments.

🎯 Primary Goal Analysis – Churn Prediction Model

The primary goal of this project was to develop a binary classification model capable of predicting whether a customer will cancel their service in the next billing cycle. The model selected for this task was the Gradient Boosting Classifier, trained using resampled and standardized data.

Model Performance Summary

Metric Churn = No (0) Churn = Yes (1)
Precision 0.87 0.54
Recall 0.79 0.67
F1-Score 0.83 0.60

Overall Accuracy: 0.76 ROC AUC Score: 0.8204

Key Insights

  • The model achieved a ROC AUC of 0.8204, indicating good overall discrimination between churners and non-churners.
  • Recall for churners (0.67) shows the model can correctly identify 67% of customers who actually churned — a critical metric for proactive retention strategies.
  • The precision for churn (0.54) indicates that just over half of predicted churners are true churners, reflecting some trade-off between catching more churners and accepting false positives.
  • The F1-score of 0.60 for churn indicates a balanced compromise between precision and recall in this imbalanced classification scenario.

6.4) Secondary Goal:

Generate business insights through descriptive and inferential analytics, including dashboards that highlight:

  • Impact of features such as contract type, tech support, billing profile, etc.;
  • Demographic and geographic patterns among high-churn customers;
  • Actionable recommendations for targeted customer segments.

6.4.1) Churn by Contract Type

Churn by Contract Type

The bar plot clearly shows the distribution of customer churn (Churn = 1) across different types of contracts (0 = Month-to-month, 1 = One year, 2 = Two year).

🔍 Key Insights:

  • Month-to-month contracts (0) have the highest churn rate. The churn count is nearly as high as the retention count.
  • One-year contracts (1) show a significantly lower churn count compared to month-to-month customers.
  • Two-year contracts (2) exhibit the lowest churn rate among all types.

📌 Graph Interpretation

  • Customers with short-term contracts are more likely to churn, which aligns with the hypothesis that flexibility leads to higher risk of attrition.
  • Longer commitment contracts (one or two years) serve as a natural retention mechanism.

💡 Actionable Recommendations:

  • Offer incentives or discounts to month-to-month customers to encourage migration to annual plans.
  • Use contract length as a key predictor in churn models to detect high-risk segments.
  • Launch loyalty programs for short-term users to boost retention and customer commitment.

6.4.2) Churn by Contract Type

Churn by Contract Type

🔍 Insight: Technical Support Has a Strong Impact on Churn

  • Customers with no Tech Support (0) have the highest churn rate, com uma grande proporção de clientes que saíram (Churn = 1).
  • Customers who opted for Tech Support (1 or 2) têm baixas taxas de churn, indicando maior retenção.
  • A diferença visual entre os grupos mostra que ter suporte técnico é um forte fator de retenção.

📌 Graph Interpretation

  • Providing Tech Support services is strongly correlated with customer retention.
  • Customers with no technical assistance are more prone to churn, possibly due to unresolved technical issues or low engagement.

💡 Actionable Recommendations:

  • Proactively offer Tech Support to customers at risk.
  • Include free or discounted Tech Support trials during the early tenure period.
  • Target upselling of Tech Support to customers in churn-prone segments.

6.4.3) Paperless Billing

Paperless Billing

📌 Graph Interpretation

  • Customers who use paperless billing (PaperlessBilling = 1) show a significantly higher number of cancellations (Churn = 1) compared to those who don’t.
  • Customers with traditional billing (PaperlessBilling = 0) have a lower churn rate.

🔍 Key Insights:

  • Paperless Billing users tend to churn more frequently.
  • Possible explanation: These customers might be more tech-savvy, price-sensitive, or more exposed to competitor promotions.
  • Their behavior may also suggest lower brand engagement, as the communication model is entirely digital.

💡 Actionable Recommendations:

  1. Launch retention campaigns for paperless billing users with a high probability of churn.
  2. Monitor digital engagement — are these users opening emails, checking their online account, etc.
  3. Offer exclusive perks or loyalty programs to keep this group engaged and connected to the brand.

6.4.4) Churn by Senior Citizenship

Churn by Senior Citizenship

📌 Graph Interpretation

The bar chart shows the distribution of churn among senior citizens (SeniorCitizen = 1) and non-senior customers (SeniorCitizen = 0), with the churn status represented by hue.

  • Non-Senior Customers (0):
  • Represent the majority of the customer base.

  • Show a significant number of churns (Churn = 1), although more customers stay.

  • Senior Citizens (1):

  • Comprise a smaller segment of the base.
  • Display a proportionally higher churn rate than non-seniors.
  • The bar heights for churned vs non-churned are more balanced for this group.

🔍 Key Insights:

  • Churn Risk by Age Group:
  • Senior citizens are more likely to churn relative to their representation in the dataset.

  • This demographic might be more sensitive to service quality, usability, or pricing.

  • Marketing & Retention Implication:

  • Senior customers should be considered a priority target for personalized retention campaigns.
  • Simplified billing, improved support, and loyalty incentives could reduce churn risk in this group.

💡 Actionable Recommendations:

  • Tailor customer experience for older clients.

  • Offer additional onboarding or support for new senior customers.

  • Visualization & Communication:

  • This dashboard-style visualization supports data-driven storytelling for business stakeholders.

6.4.5) Churn by Partner status

Churn by Partner status

📌 Graph Interpretation

The visualization shows the relationship between customer churn and whether the customer has a partner (1) or not (0).

🔍 Key Insights:

  • Customers without a partner (Partner = 0) have a higher churn rate than those with a partner.
  • Among customers with a partner (Partner = 1), the number of churned customers is significantly lower, despite a similar total population size.
  • This indicates that partnership status may contribute to customer stability, possibly due to shared financial responsibilities or stronger service value perception within households.

💡 Actionable Recommendations:

  • Create custom retention offers for customers without a partner, who show higher churn risk.
  • Use Partner status in churn prediction models to improve accuracy.
  • Design targeted messaging that emphasizes independence, convenience, and personalized value for single customers.

6.4.6) Churn by Churn by Customer Lifetime Value Quartile

Churn by Churn by Customer Lifetime Value Quartile

📌 Graph Interpretation

  • The CLTV metric, represented by TotalCharges, is a strong inverse indicator of churn.
  • Lower-revenue customers are more volatile and more likely to churn.
  • Retention strategies should prioritize high CLTV customers at risk of churning (combining churn probability × CLTV).

🔍 Key Insights:

  • Customers in the "Low CLTV" quartile show the highest churn rate compared to other quartiles.
  • As CLTV increases, the number of churners decreases significantly.
  • The "Very High CLTV" group has the lowest churn count, indicating better customer retention among high-value clients.

💡 Suggested Actions

  • Implement personalized offers for customers with high CLTV and moderate churn probability.
  • Use predictive models to monitor changes in CLTV and churn behavior across segments.
  • Design loyalty programs to incentivize longevity and increase total customer value.

💡 Summary of Actionable Business Insights

📦 Contract Type

  • Offer discounts or upgrades to month-to-month customers to encourage migration to longer contracts.
  • Use contract duration as a key churn predictor.
  • Develop loyalty programs for short-term clients.

🛠️ Tech Support

  • Proactively offer Tech Support to at-risk customers.
  • Provide free or discounted trials during early months.
  • Target upselling Tech Support in high-churn segments.

💻 Paperless Billing

  • Launch retention campaigns for paperless billing users with high churn risk.
  • Monitor digital engagement behavior (email opens, logins).
  • Provide exclusive digital perks or rewards for this tech-savvy group.

👵 Senior Citizenship

  • Offer tailored onboarding and support for senior customers.
  • Simplify interfaces and billing for older demographics.
  • Personalize retention messaging for this sensitive user group.

💞 Partner Status

  • Create custom retention offers for customers without a partner.
  • Use Partner status in predictive models to improve segmentation.
  • Develop messaging that emphasizes autonomy and value for singles.

💰 Customer Lifetime Value (CLTV)

  • Prioritize high CLTV customers with rising churn risk.
  • Monitor CLTV × churn probability to guide retention investments.
  • Implement VIP programs or loyalty incentives to boost lifetime value.

⬅ Voltar para a Página Inicial