Deep dive — methodology
How the A / B / C determination is derived
Every $499 Snapshot lands on exactly one overall determination. Not a color code, not a health score — an A/B/C determination tied to ATO + RMF + DoD AI Ethical Principles attestation stance. Here is exactly how it is derived.
The three categories
| Category | Meaning | Attestation stance |
|---|
| A |
No high-severity drift events. No group-differential signal. Baseline vs recent distributions match within threshold. |
Sufficient for attestation. Signed evidence available on file for ATO renewal + OMB M-24-10 attestation. |
| B |
Medium-severity drift or single-event high-severity without group-differential signal. Aggregate metrics stable. |
Conditional. Named remediation + re-audit within defined window supports attestation. Contracting Officer notification recommended. |
| C |
Multiple high-severity drift events + significant asset-group differential. |
Insufficient. Program office should not represent the AI as ATO-defensible, RMF-compliant, or DoD Ethical Principles adherent (Reliable / Traceable / Governable) until named remediation completes. |
The evidence chain
- Sensor thresholds active during the audit:
- Mean-shift threshold: 1.5 standard deviations from baseline
- Variance-ratio threshold: 1.5x baseline
- KL-divergence threshold: 0.2
- Asset-group differential threshold: 12% shift across groups
- Event classification:
- Any single-threshold cross → medium-severity event
- KL-divergence > 1.0 OR group-differential signal → high-severity event
- Category assignment:
- 0 high-severity + 0 group-differential → A
- 1-2 high-severity, no group-differential signal → B
- ≥ 3 high-severity or any group-differential signal fired → C
Why this sample lands on C
12 high-severity drift events across the 90-day audit period. No cross-group differential signal was fired (F3 collapse was very large but was not accompanied by an offsetting shift in another group — the 12% cross-group threshold requires a matched pair). The high-severity event count alone is sufficient for C. The F3 red-lane collapse from 34% to 8% is the operational picture behind the numeric determination.
Why the A/B/C structure matters for federal AI
Continuous-scale scores (0-100 dashboard, red/yellow/green tri-lights, "health-score" percentages) push program offices toward "just above the threshold" reporting behavior. A/B/C determinations force a single stance question:
Would you sign the OMB M-24-10 AI Use Case Inventory attestation as-is today? Would you defend the ATO package as-is? A/B/C answers the question that GAO / DoD IG / Congressional Armed Services Committee / your Contracting Officer actually asks — instead of the question the sustainment vendor's dashboard prefers.
How does this help me?
Knowing your category BEFORE the ATO renewal / OMB attestation / GAO audit letter / accident-investigation-board request lands is the whole value of the $499 discipline.
Read: The value of knowing your category first →
$499 Snapshot. 3 business days.
Get your program office's A / B / C determination on your actual AI surface — before someone else's question forces the answer.
Buy $499