Methodology

Everything on this site comes from single audits filed with the Federal Audit Clearinghouse. This page explains what is counted, how the score is built, and where the data falls short. Data as of October 30, 2026.

Source and scope

• Source. The FAC's public data export: each audit's general information, its federal awards, its findings, and the findings' text. Every finding on the site links to its audit's FAC record.
• Years. Audit years FY2023 through FY2025. Repeat chains look further back, to FY2016.
• States. Colorado, Texas and Arizona, placed by the auditee's address (ZIP code, with the city as a tie-breaker), not by the state field on the form, which is sometimes wrong. A small number of organizations have a mismatch between the two; they are kept, under their address state.
• Organizations. An organization is its Unique Entity ID (UEI). If it filed more than one audit for a year, the one with the latest fiscal year end is used.
• Edition. This edition uses the FAC export as of October 30, 2026, the latest acceptance date in the data. The FAC updates its export weekly.

Findings

A finding is one reference number in one audit (for example, 2025-001). Its classifications are the ones the auditor marked, in the filing's own terms:

• Material weakness and significant deficiency: the two levels of internal-control finding. Material weakness is the more serious.
• Modified opinion: the finding relates to a qualified or adverse opinion, or a disclaimer, on a major program.
• Questioned costs: the auditor questioned costs charged to a federal award.
• Repeat finding: the auditor marked the finding as a repeat of a finding from an earlier audit.
• Other matters and other findings: noncompliance or other findings outside those classes.

A finding can appear on several award lines; a classification counts if any line carries it. The finding's text is shown as filed, except that the audit firm's name is replaced with "[the auditor]". We name organizations, never their auditors.

Repeat chains

A repeat finding lists the reference number of the earlier finding it repeats. We follow those references back through the organization's earlier audits (matched by UEI or EIN, up to three years apart) to measure how many audit years the problem has run. When the earliest finding we can reach is itself marked as a repeat but its own predecessor cannot be found, the chain is shown as "at least" that many years. FAC data has known gaps in these references (GAO-24-106173), so chains can be longer than shown, but not shorter.

Questioned costs

The FAC records whether a finding questions costs, not how much. Amounts are read from the finding's text: first by pattern matching, then, for ambiguous Colorado, Texas and Arizona findings, by a language model asked to quote the figure. In a hand check of 30 findings, 29 were right. Each amount carries its basis:

• Known: the auditor states the questioned amount.
• Likely (projected): the auditor projects or extrapolates an amount; shown, and scored at half weight.
• Known and likely, as one figure: the text gives a single figure that may combine both.
• Not determinable or no amount stated: the finding questions costs without a figure.

The score

Each organization's latest audit gets a score from 0 to 100, the sum of the eight parts below. Each part is a count or flag a reader can check against the audit. The caps add up to 100, so there is no scaling or hidden weighting. The score summarizes what the auditor reported. It predicts nothing, and it is not a judgment of the organization.

Questioned costs score the higher of two tiers, using the audit's total known amount, and only at $25,000 or more:

Likely (projected) amounts use the same tiers, halved and rounded up. A questioned-cost finding with no amount scores 3: an auditor who could not bound the problem has reported something more serious than a known small figure.

Prior-year lateness counts when the previous year's audit was submitted after its due date or was expected and never filed, and only when this year's audit has a finding. Going concern is a flag with modest weight. The caps flatten the largest audits: a state government with dozens of findings scores near the top every year, so the pages also show the raw counts.

Due dates and missing audits

A single audit is due nine months after the end of the fiscal year (for a June 30 year end, the following March 31). An organization is shown as having no audit on file only when it is at least 90 days past that deadline, and only when its previous audit reported federal spending above the threshold: $750,000 for fiscal years beginning before October 1, 2024, $1 million after. This is approximate. An organization whose spending fell below the threshold would not need an audit, and some audits are filed under another entity's number. Pandemic relief programs (Assistance Listings 21.019, 21.023, 21.026, 21.027, 21.029, 21.032, 59.072, 84.425 and 93.498) are winding down, a common reason for spending to fall; their share of the prior year's spending is shown where it matters. An audit submitted late is marked only when it was submitted 90 or more days after the due date.

Tribal audits

Tribes may choose not to make their reporting package public (2 CFR 200.512(b)(2)). For those audits the FAC publishes each finding's classification but not its text, so their pages show findings without narrative.

What we don't do

• Name or rate audit firms.
• Describe findings in words the filings don't use.
• Predict anything, or compare sectors.

Known limits

• FAC data contains duplicates, inconsistent entries and missing prior-finding references (GAO-24-106173). We clean what we can and say where a figure is a lower bound.
• Amounts read from text can be wrong; each finding links to the filing so a reader can check.
• Compliance-only findings score 0 unless they repeat, question costs or come with a modified opinion.