AI Skills Marketplace / Customer Success

Churn Risk Cohort Survival Analyzer

The Churn Risk Cohort Survival Analyzer is a free Claude Skill that computes real customer retention curves using the Kaplan-Meier product-limit estimator, correctly accounting for customers who have not churned yet instead of dropping them or wrongly counting them as permanently retained, and compares retention across cohorts.

↗ View Source ⬇ Download Full Repo (.zip)
MIT
License
3 functions
Decision branches
censoring-aware survival math + honest null-median reporting
Escalation rules
Yes
Multi-tool support
Aug 17, 2026
Last updated
14/14 passing
Tests
README.md View raw ↗

What this is (and isn’t)

A real survival-analysis engine, not a health-score model. A health score predicts an individual customer’s risk today. This answers a different, cohort-level question: what fraction of customers like this one are still active after N days, computed the statistically correct way.

Files

FilePurpose
survival.jscomputeKaplanMeierCurve(), computeMedianSurvivalTime(), compareCohortSurvival() — zero dependencies
test.js14-test suite, hand-verified against the classic Kaplan-Meier worked example
SKILL.mdFull skill definition, loadable in Claude Code/Desktop

The two real guardrails

Correct handling of censored customers: someone who has not churned yet is still at risk through their current tenure but is never wrongly counted as an event, avoiding both the understated-retention mistake of dropping active customers and the overstated-retention mistake of counting them as permanently safe. And honest median reporting: returns null, not an extrapolated guess, when the observed data never shows survival dropping to 50 percent.

FAQ

Is this skill free to use?

Yes. Free to download and run yourself. MV3 charges $175/hr only for implementation help wiring this into your real subscription or billing platform's actual tenure and churn data.

How is this different from a health-score model?

A health score predicts an individual customer's risk today. This computes a cohort-level retention curve, the real statistical fraction of customers still active at each point in time, using the same method (Kaplan-Meier) used in clinical and reliability survival analysis.

What happens to customers who have not churned yet?

They are correctly treated as censored observations: counted as at risk through their current tenure, but never counted as a churn event. This is the core problem naive retention math gets wrong.

What if my data never shows 50 percent survival?

computeMedianSurvivalTime() returns null rather than extrapolating a guess. Many real retention curves never reach that point within the observation window.

Does my customer data get sent to MV3?

No. All calculations run entirely in your own environment.

RB
Ryan Brooks
Technical SEO, MV3 Marketing