Predicting Customer Churn for a Telecom Provider

How a mid-size telecom used churn prediction to act before customers left

Scroll

Challenge

A mid-size telecom was losing subscribers to competitors. By the time retention teams saw the signal—complaints, downgrades, or port-out requests—it was often too late to save the relationship. Leadership wanted a predictive churn model so they could intervene earlier with targeted offers and outreach.

We were asked to build a churn model using usage, billing, support, and optional survey data; score customers by churn risk; and recommend how to use the scores in retention campaigns and capacity planning.

Approach

We built a churn prediction model using historical behavior and outcomes. We used usage, tenure, billing, support contacts, and—where available—NPS or satisfaction. We trained and validated the model, produced risk scores and segments, and outlined how to feed scores into retention workflows and measure lift.

Solution

We delivered a churn score and playbook.

We consolidated 18 months of customer-level data, defined a churn outcome window, and built a model that ranked customers by likelihood to churn. We validated lift over a holdout period and segmented the base into risk tiers. We delivered score logic, a refresh cadence, and a short playbook for retention: who to contact first, which levers to use, and how to track save rates. Timeline: 8 weeks.

Take the next step

See how Intellimark can help you predict churn and act before customers leave.

Outcome

The telecom could prioritize retention by risk:

Metrics / Results

−22% voluntary churn in the highest-risk segment after 6 months

Churn score refreshed monthly and used by retention and care

3 risk tiers with clear action rules

Save-rate tracking in place for campaigns

Fin.

Churn is expensive when you find out after the fact. They could see it coming.

Explore more

Churn Modeling · Account Health

Experience to Impact · Account Health in Healthcare