AI Skills Marketplace / Customer Success

Health-Score Weight Calibrator

The Health-Score Weight Calibrator is a free Claude Skill that statistically fits customer-health-score signal weights against real renewal/churn outcomes using L2-regularized logistic regression, instead of hand-guessing weights in a config UI.

↗ View Source ⬇ Download Full Repo (.zip)
MIT
License
2 functions
Decision branches
L2-regularized fit
Escalation rules
Yes
Multi-tool support
Aug 14, 2026
Last updated
14/14 passing
Tests
README.md View raw ↗

What this is (and isn’t)

A calibration function, not a hosted service or a CS platform. It runs entirely in your own environment against a dataset you supply.

Files

FilePurpose
calibrator.jscalibrateWeights() and scoreAccount() — zero dependencies
test.js14-test suite covering convergence, sign-correctness, edge cases, and determinism
SKILL.mdFull skill definition, loadable in Claude Code/Desktop

Why this is a real gap

Even Gainsight’s own AI Scorecards present health-score weighting as a config UI, not a statistical fit against real outcomes. This skill closes that gap with an actual fitted logistic-regression model, reporting accuracy and log-loss so you can see how well the weights actually explain your own churn history.

FAQ

Is this skill free to use?

Yes. Free to download and run yourself. MV3 charges $175/hr only if you want it wired into your real CS platform and signals piped in automatically.

What data do I need?

At least 4 historical account rows with numeric signals and a churned/renewed outcome, with both outcomes represented in your data.

Does my account data get sent to MV3?

No. The calibration runs entirely in your own environment. Nothing leaves your machine.

Which CS platforms does this work with?

Any platform you can export signals from, including Gainsight, Vitally, Planhat, HubSpot, or a homegrown system. It is platform-agnostic by design.

How is this different from a CS platform's built-in health score?

Most CS platforms (including Gainsight's own AI Scorecards) treat weighting as a manual config UI. This skill fits the weights statistically against your own historical outcomes instead.

JR
Jordan Reeves
ABM & Outbound Pipeline, MV3 Marketing