METHODOLOGY · TECHNICAL APPENDIX

Sources and scoring models.

The full catalogue of federal files we hold, what we take from each, and the scoring models we maintain. The methodology page explains how a question is answered; this appendix is the reference behind it.

Standard reports compare raw federal records against named groups and do not depend on these scores. Scores appear in a deliverable only when a commission asks for them, always labelled as AuditPoint analysis with their government inputs listed. Each report states the vintage of every source it uses.

Contents

• A · Source catalogue — care, freight, banking, credit unions, employers, contractors, retirement plans, higher education, other sectors
• B · Scoring models — banks, credit unions, carriers, employers, federal contractors, retirement plans, colleges
• C · Principles
• D · What this appendix leaves out

A · Source catalogue

Post-acute care

Commercial freight

Banking

Credit unions

Employers

Federal contractors

Retirement plans

Higher education

Other sectors, available on scoping

B · Scoring models

Every score below is an AuditPoint analytical output built from government inputs. None is an official designation of the FDIC, NCUA, FMCSA, OSHA, the Department of Labor, PBGC, the Department of Education or any other agency.

Bank Health Score

A 0–100 composite of five components: capital adequacy 25%, asset quality 25%, earnings 20%, liquidity 15%, regulatory standing 15%.

• Capital: Tier 1 leverage ratio, scored in bands from 15% and above down to below 4%.
• Asset quality: noncurrent loans (RC-N basis), net charge-offs and real-estate concentration against equity.
• Earnings: return on assets and net interest margin.
• Liquidity: loan-to-deposit ratio and uninsured deposit share.
• Regulatory standing: formal enforcement and CRA rating.
• Overrides: return on assets below −3% caps the score at 39; uninsured deposits above 75% cap it at 79.
• Tiers: STRONG 80–100, ADEQUATE 70–79, WATCH 40–69, STRESSED 0–39. Since September 2026, banks insured for under three years are shown as DE NOVO rather than scored, because early losses are normal for new banks.

Note on the June 2026 run. The June 2026 scores (December 2025 Call Reports) used a noncurrent-loan input that was misassigned for many banks. It was corrected before the July run. Re-scoring the June run with the corrected input, everything else unchanged, left the STRESSED tier exactly as it was: the same 15 banks.

Credit Union Health Score

A 0–100 composite on the same structure: capital 25% (net worth ratio against NCUA prompt corrective action thresholds: well capitalized at 7%, adequately at 6%), asset quality 25% (noncurrent loans and net charge-offs), earnings 20% (return on assets against a fixed reference of 0.94%), liquidity 15% (loan-to-share ratio against a reference of 70%), regulatory standing 15% (NCUA capital classification).

• Tiers: STRONG 80–100, ADEQUATE 65–79, WATCH 40–64, STRESSED 0–39.
• Net income: for large credit unions that report zero net income because of a mid-year transfer, it is rebuilt from the income statement.
• Data gap: NCUA does not publish institution-level enforcement orders in bulk, so enforcement is not scored. This is a documented gap, not a scored zero.

WeighStation Score (commercial carriers)

A 0–100 carrier risk score from four components: safety record, crash risk over 24 months, insurance stability (cancellation history) and authority integrity (revocation history). Tiers: GREEN, YELLOW, RED.

WorkRecord Score (employers)

A 0–100 employer compliance score: OSHA citation severity 35%, wage-hour violations 25%, OSHA penalty scale 20%, recency 20%. Tiers: CLEAN 80–100, MONITORED 60–79, CITED 40–59, CRITICAL 0–39.

Contractor Risk Score

A 0–100 vendor risk score for federal contractors: federal dependency 35% (obligation volume and trend, adjusted for large commercial firms), agency concentration 30% (Herfindahl index across awarding agencies), contract-type mix 20%, exclusion and integrity 15%.

• Integrity penalties: an active exclusion caps the score. FAPIIS records subtract points by type: terminations for default or cause 20–25, administrative agreements 20, nonresponsibility 15, defective pricing 10, indictments and convictions 30–35. Lookback windows run three to five years, with no limit for criminal findings.
• Tiers: STABLE 70–100, WATCH 50–69, EXPOSED 30–49, CRITICAL 0–29 for the population model; vendors with FAPIIS records use 65 / 45 / 25 so the integrity penalty is not counted twice.

PensionGuard (retirement plans)

Four models, each scored on its own statutory or fiduciary basis rather than one blended composite, on a STRONG / ADEQUATE / WATCH / STRESSED scale:

• Single-employer defined benefit: the ERISA §436 AFTAP bands: 80% and above, 60% to under 80%, and under 60%, where benefit accruals must freeze.
• Multiemployer defined benefit: PPA zone status, cross-checked against the PBGC Special Financial Assistance roster.
• Large-plan fiduciary: auditor's opinion and delinquent contributions, with limited-scope audit elections never scored as distress.
• Small-plan fiduciary: three tiers from the 5500-SF signals, since small plans have no auditor's opinion.

Limits: withdrawal liability is not reported at plan level and is not estimated; delinquent contributions are gross as filed.

CampusIndex (colleges and universities)

A closure-risk tier that reads solvency against enrollment trajectory.

• Solvency: where the Department of Education publishes its Financial Responsibility composite, that number is used unchanged. Where it publishes none, chiefly public institutions, a Composite Financial Index is derived from IPEDS audited financials on the four standard strength factors: primary reserve, viability, return on net assets and net operating revenue.
• Tiers: STRONG 2.0 and above, ADEQUATE 1.5–1.99, WATCH 1.0–1.49, STRESSED below 1.0, and CRITICAL where distressed solvency meets collapsing enrollment. The 1.5 and 1.0 lines are Department of Education thresholds; the 2.0 line is our own analytical split.
• Calibration: of institutions we scored that later closed, 71 of 90 in a clean holdout (78.9%) had been flagged WATCH, STRESSED or CRITICAL first.
• Limits: institutions often leave the IPEDS files the year they close, so the claim is "of institutions we scored that later closed," never "of all closures." Solvency comes from audited statements one to two years old, so this is an annual structural signal, not a real-time one.

C · Principles

• Primary sources only. Every input traces to a named federal file. No third-party aggregators, surveys or estimates.
• Derived figures are labelled. Any score, tier or composite is identified as AuditPoint analysis, with its government inputs listed.
• Vintage on every figure. Each report states the vintage of every source it uses.
• Field definitions from the source. Field meanings are read from each agency's own data dictionary or form instructions, not inferred from column names.
• No editorial verdicts. We report what the federal record shows. Conclusions stay with the reader.

D · What this appendix leaves out

We also maintain scoring models for care providers, hospices, nonprofits, food-safety establishments, vessels and railroads. They are not described here while they are under review, and none is used in a deliverable until that review is complete. The underlying federal records for those sectors remain available on scoping.