Early Detection of Customer Churn in Telecoms
The Retention Director wants to identify which current customers are highly likely to cancel their contract next month, so as to trigger loyalty calls and personalised offers before they leave.
Where the money comes from
What to do now
- 1Calculate the monthly margin per customer (not just ARPU).
- 2Run the model on the top 15% by risk with a 10% control group.
- 3Measure incremental churn avoided and the cost of discounts.
For your committee
One page with the decision, the impact, the risks, the steps and the KPI to track.
The print dialog opens: choose "Save as PDF".
Watch out: more volume is not always more profit
Retaining at any price: if the discount is greater than the customer's margin, every retention loses money. Prioritise by customer margin x risk, not just by risk, and do not retain customers with negative margin.
KPI for the committee
Retained margin net of discounts for every euro invested in retention.
Detail for your data team
Proposed technique: Logistic Regression
Why: The goal is to classify a binary event (Leaves = 1 / Stays = 0) and obtain a calibrated probability. Logistic Regression makes it possible to rank customers from highest to lowest risk and to quantify how much each complaint multiplies the probability of churn.
Cost of getting it wrong: False negatives mean losing recurring revenue (LTV); false positives mean giving away retention discounts to customers who were not planning to leave.
Data needed: CRM and Billing: Customer tenure, contracted tariff, current discounts and payment punctuality.; Network and Usage: Trend in voice minutes and GB consumed over the last 90 days.; Call Center / Support: Number of calls for technical incidents or billing complaints.; Churn History: Labelled record of customers who cancelled (1) vs. stayed (0) in the last 12 months.