SCORING METHODOLOGY

How RANKIGI calculates governance scores.

No black boxes. No proprietary algorithms. The scoring methodology is fully transparent and independently verifiable.

Why We Publish This

A governance score is only credible if its methodology is open to scrutiny. We publish the exact weights, formulas, and edge case handling so organizations can understand, challenge, and independently verify every score RANKIGI produces.

WHO — Human Accountability (25% weight)

The highest-weighted dimension because human accountability is the foundational requirement of every major AI regulation.

Accepted human owner: +50 pts. Accepted approver: +30 pts. Accepted delegate: +20 pts. Maximum: 100.

WHAT — Declared Scope (20% weight)

Declared scope present: +40 pts. Valid active passport: +40 pts. All recent tool invocations within scope: +20 pts. Maximum: 100.

WHEN — Audit Trail Completeness (15% weight)

Events in last 7 days: +30 pts. Chain integrity 100%: +40 pts. No unexplained offline gaps >15 min: +30 pts. Maximum: 100.

WHERE — Environment and Identity (15% weight)

Framework declared: +30 pts. Environment declared: +30 pts. Ed25519 identity verified: +40 pts. Maximum: 100.

WHY — Intent Chain Coverage (15% weight)

Coverage ≥90%: 100 pts. 70-89%: 70 pts. 50-69%: 40 pts. 25-49%: 15 pts. <25% or not enabled: 0 pts.

HOW — Policy Compliance (10% weight)

Each installed and passing compliance framework: +20 pts. Maximum 5 frameworks × 20 = 100 pts.

Overall Score

Overall = (WHO × 0.25) + (WHAT × 0.20) + (WHEN × 0.15)
       + (WHERE × 0.15) + (WHY × 0.15) + (HOW × 0.10)

90-100: PRODUCTION READY — Certificate issued.
70-89:  GOVERNED — Certificate issued with recommendations.
50-69:  PARTIAL — No certificate. Remediation required.
0-49:   UNGOVERNED — No certificate. Significant gaps.

Certificate Issuance

Certificates issued at score ≥70. Expire annually. Revocable if: owner removed, chain integrity fails, critical unresolved violation, or org requests revocation.

Independent Verification

Every score is reproducible. Raw inputs available in dashboard. We welcome challenges: methodology@rankigi.com.