The BANT Lead Qualification Schema snippet ships with MV3’s own reasonable default weights. Those are a real starting point, not a claim that they fit every business. This skill computes weights from your own historical closed-deal data instead, using point-biserial correlation, a standard statistical technique, not a black box.
How it works
const { calibrateWeights } = require('./calibrate.js');
const weights = calibrateWeights(myClosedDeals);
// { weights: { budget_confirmed: 5.2, authority_level: 3.1, ... },
// correlations: {...}, sample_size: 84 }
Sample size matters
Under 30 historical deals, the result includes an explicit warning field. Correlations from small samples are noisy: treat the output as directional, not final, until you have more closed deals to calibrate against.
What this is (and isn’t)
Documentation + a calibration algorithm, not a hosted service. You run this against your own exported CRM data: nothing leaves your environment.
Files
| File | Purpose |
|---|
SKILL.md | The skill definition |
calibrate.js | Zero-dependency point-biserial correlation engine |
test.js | Synthetic dataset with a known ground truth, verifies the engine recovers it |
Security & validation
Verified 2026-08-12: npm audit 0 vulnerabilities (zero dependencies), node test.js 8/8 passing, including a planted-ground-truth test, a small-sample-warning test, and a constant-value edge case that must return 0 correlation, not NaN.
Support
Open an issue on GitHub Issues. Want it run against your real CRM export? Book a scoping call ($175/hr).
License
MIT
FAQ
Is this skill free to use?
Yes. Free to download and run yourself. MV3 charges $175/hr only if you want it run against your real CRM export for you.
What data do I need?
Closed deals with budget_confirmed, authority_level, need_severity, timeline_days, and whether each deal closed_won. Export this from your CRM.
Does my deal data get sent to MV3?
No. The calibration engine runs entirely in your own environment with zero outbound network calls.
How is this different from the BANT snippet?
The snippet uses MV3's fixed default weights. This skill computes custom weights from your own win/loss history instead.
Where do I get support if something breaks?
Open a GitHub Issue for bugs or questions. For paid help running this against your real CRM, book a scoping call.
