Executive Summary
Key churn metrics and top findings
The overall churn rate is 26.6%. The strongest predictor of churn is Tenure — customers in high-risk segments show markedly higher churn probability than the average. 1106 customers are predicted to have a churn probability above the 50% threshold and represent the top priority for retention outreach.
Churner vs. Retained Customer Profile
Statistical comparison of churned vs retained customers across all predictor variables
| Variable | Mean Churned | Mean Retained | P Value |
|---|---|---|---|
| Tenure | 17.91 | 37.88 | 0 |
| Monthly charges | 74.47 | 61.14 | 0 |
| Total charges | 1530 | 2560 | 0 |
| Contract | 1 | 1 | 0 |
| Internet service | 0.94 | 0.72 | 0 |
| Online security | 0.23 | 0.61 | 0 |
| Tech support | 0.23 | 0.61 | 0 |
| Payment method | 1 | 1 | 0 |
| Paperless billing | 0.75 | 0.53 | 0 |
| Senior citizen | 0.26 | 0.13 | 0 |
| Partner | 0.35 | 0.53 | 0 |
| Dependents | 0.16 | 0.34 | 0 |
| Gender | 1 | 1 | 0.3742 |
The table ranks each predictor by statistical significance (p-value) when comparing churned to retained customers. Variables at the top show the largest and most reliable differences between the two groups. Numeric variables show group means; categorical variables show the proportion of customers with the 'Yes' or non-reference level.
Churn Rate by Customer Segment
Observed churn rates by contract type, internet service, and payment method
Each bar shows the observed churn rate within a segment defined by contract type, internet service, or payment method. The highest-risk segment is Payment method: Electronic check with a churn rate of 46%. Segments with fewer than 5 customers are excluded to avoid misleading rates.
Tenure Distribution by Churn Status
Tenure in months compared between churned and retained customer groups
Churned customers have a median tenure of 10 months compared to 38 months for retained customers. Newer customers (short tenure) churn at a disproportionately high rate, suggesting that early engagement and onboarding are critical retention levers.
Logistic Regression: Churn Odds Ratios
Exponentiated logistic regression coefficients with 95% confidence intervals
Each bar represents an odds ratio from logistic regression — values above 1 increase churn odds, values below 1 are protective. Whiskers show the 95% confidence interval; predictors whose interval excludes 1 are statistically significant. The strongest risk factor is Internet service: Fiber optic (OR = 1.719).
Random Forest Variable Importance
MeanDecreaseGini importance scores from a random forest model
MeanDecreaseGini measures how much each variable reduces classification impurity across all trees in the random forest. Unlike logistic regression odds ratios, this captures non-linear effects and variable interactions. The most important predictor by this measure is Total charges.
Predicted Churn Probability Distribution
Distribution of per-customer predicted churn probabilities from logistic regression
The histogram shows the distribution of predicted churn probabilities for all customers. Probabilities near 0 indicate low churn risk; near 1 indicates high churn risk. 22.1% of customers are predicted above the 50% threshold and merit proactive retention action.
High-Risk Customer List
Top 25 customers ranked by predicted churn probability
| Customer ID | Tenure | Monthly Charges | Contract | Predicted Churn Prob |
|---|---|---|---|---|
| 7216-EWTRS | 1 | 100.8 | Month-to-month | 0.815 |
| 8775-LHDJH | 1 | 90.6 | Month-to-month | 0.811 |
| 5192-EBGOV | 1 | 85.7 | Month-to-month | 0.809 |
| 1069-XAIEM | 1 | 85.05 | Month-to-month | 0.809 |
| 7024-OHCCK | 2 | 93.85 | Month-to-month | 0.807 |
| 4695-WJZUE | 2 | 88.55 | Month-to-month | 0.806 |
| 5277-ZLOOR | 2 | 85.55 | Month-to-month | 0.805 |
| 2545-EBUPK | 2 | 84.05 | Month-to-month | 0.804 |
| 0723-DRCLG | 1 | 74.45 | Month-to-month | 0.804 |
| 1977-STDKI | 1 | 73 | Month-to-month | 0.803 |
| 4713-LZDRV | 2 | 79.6 | Month-to-month | 0.803 |
| 0021-IKXGC | 1 | 72.1 | Month-to-month | 0.803 |
| 1374-DMZUI | 4 | 94.3 | Month-to-month | 0.802 |
| 3878-AVSOQ | 1 | 71.25 | Month-to-month | 0.802 |
| 9488-HGMJH | 1 | 71.15 | Month-to-month | 0.802 |
| 8051-HJRLT | 1 | 70.55 | Month-to-month | 0.802 |
| 9605-WGJVW | 1 | 70.2 | Month-to-month | 0.802 |
| 1846-XWOQN | 1 | 70.15 | Month-to-month | 0.802 |
| 4871-JTKJF | 1 | 69.65 | Month-to-month | 0.802 |
| 8375-DKEBR | 1 | 69.6 | Month-to-month | 0.802 |
| 2636-ALXXZ | 1 | 69.6 | Month-to-month | 0.802 |
| 7206-GZCDC | 1 | 69.25 | Month-to-month | 0.801 |
| 8098-LLAZX | 4 | 95.45 | Month-to-month | 0.801 |
| 1400-MMYXY | 3 | 105.9 | Month-to-month | 0.801 |
| 3068-OMWZA | 1 | 88.8 | Month-to-month | 0.8 |
The 25 customers with the highest predicted churn probability are ranked here. Each row shows the customer ID, their tenure, monthly charges, contract type, and predicted probability of churning. Use this list to prioritize personalized retention outreach — customers with short tenure and month-to-month contracts combined with high charges are at greatest risk.