Case Study
Customer Analytics
Customer Churn & Retention
Which customers are at risk of leaving, and which intervention is worth making?
Download PDF
A complementary visual resource for a deeper view of the case.
The business challenge
How can customers at risk of leaving be identified early enough, and how should the business decide whom, when and how to contact?
Indiscriminate retention can be expensive. The challenge combines an operational churn definition, calibrated prediction, value and incremental intervention effect.
Key business questions
- How should churn be defined in this business?
- Which signals anticipate it with useful lead time?
- Which customers are eligible for an intervention, and which one?
Data & evidence
- Activity, usage, purchase or renewal history.
- Service interactions and issues.
- Exposure to retention actions.
- Costs, margin and expected value.
How I would approach it
- Define churn and horizon around the business cycle.
- Create temporal windows without future information.
- Compare with a simple baseline and evaluate calibration.
- Combine risk, value and eligibility.
What you would get
- Operational churn definition and monitoring framework.
- Calibrated risk scoring and priority profiles.
- Intervention and exclusion criteria.
- Incremental retention measurement design.
From analysis to action
Illustrative example: A high churn probability does not mean an offer will prevent departure. Risk prediction and causal effect are different questions.
- Design journeys by risk, value and likely reason.
- Prioritise product or service improvements.
- Create learning through experiments.
How success should be measured
- Calibration and performance by segment.
- Incremental retention versus control.
- Cost per retained customer and net value.
Other approaches to consider
Classification, survival and competing risks answer different questions. Uplift requires suitable experimental data.