Kuwala Analytics logo

Telecom Customer Churn and Retention Analysis

Kuwala Analytics · demonstration on the public IBM Telco sample dataset
Data refreshed at Oct 13, 2026 10:40 PM
No customers match this combination of filters. Try removing one.
Total Customers
Churn Rate
Retention Rate
stayed last month
Avg Monthly Charges
per customer
Avg Tenure
months with company
Monthly Revenue at Risk
Customer Lifetime Value
avg monthly × avg tenure

Churn Rate by Contract

Share of customers who left, by contract type

Churn Rate by Tenure Band

How churn falls the longer a customer stays

Monthly Charges vs Churn

Customers (bars, left axis) and churn rate (line, right axis)

Service Adoption vs Churn

Churn rate with and without each add-on, internet customers only

Payment Method Breakdown

Customers who stayed and who churned, by payment method

Churn Rate by Demographics

Share of each group who left

Top At-Risk Customers

Current customers ranked by model risk score (0 to 100). Click a column to sort.

Why this list matters: in a back-test, the riskiest 10% of customers contained 29% of all leavers and the riskiest 20% contained 51%. Random calling would find 27%; this list finds 67% to 76%.

Churn Prediction Model

Gradient Boosting, tested on 1,409 customers the model had not seen (figures do not change with filters)
78%Recall
52%Precision
0.62F1 score
0.84ROC-AUC

Top Churn Drivers

Relative importance (strongest driver = 100)

Risk Score Distribution

Customers by risk score, split by what actually happened

In plain language: of every 100 customers who really left, the model flagged 78 in advance (recall). Of every 100 customers it flagged, 52 really left (precision). Classes were balanced during training so the model does not ignore the smaller group of leavers. A simpler logistic regression scored almost the same, so the result does not depend on one technique.

Recommended Retention Actions

Based on the full customer base (not affected by filters)

1. Win the first six months

52.9% of customers in their first 6 months leave, against 9.5% after 4 years. Add a welcome call, a 30 day check-in and a 90 day review for every new customer.

2. Move month-to-month customers onto contracts

Month-to-month churn is 42.7%, versus 11.3% on one year and 2.8% on two year contracts. Offer a discount or free add-on for committing to 12 months.

3. Bundle online security and tech support

Internet customers with online security churn at 14.6% versus 41.8% without; tech support shows 15.2% versus 41.6%. Trial these free for 3 months with customers who have no add-ons.

4. Shift electronic check payers to automatic payment

Electronic check payers churn at 45.3%, about three times the 15.2% to 16.7% seen on automatic payments. Give a small bill credit for switching.

5. Review the fibre optic experience and price

Fibre customers churn at 41.9% and account for $114,300 of the $139,131 lost each month. Survey leavers and check price and service quality against competitors.

6. Call the highest-risk customers every week

The top 20% of customers by risk score contained 51% of all leavers in a back-test. If a retention offer kept 30% of them, that is about $269,664 a year (the 30% save rate is an assumption, not a measured result).