Methodology

Filing-detail sections come from a supported Form 990, Form 990-EZ, or Form 990-PF e-file — the actual XML document the organization filed with the IRS. The long form supplies program descriptions, functional expenses, compensation, governance, balance-sheet composition, and Schedule I grants. Form 990-EZ supplies a narrower set: programs, directly reported revenue and expense lines, balance-sheet amounts, and Part IV leadership compensation. Form 990-PF supplies its books-column financial lines, direct charitable activities, foundation holdings, Part VIII compensation, and separate Part XV grants paid and approved for future payment. We do not edit or estimate the underlying figures.

For headline financials and trends, we use one provider for the entire displayed series. Comparable annual IRS e-files are selected only when at least two consecutive periods of one return type—all Form 990, all Form 990-EZ, or all Form 990-PF—are available, the newest period matches the filing-detail record, and no newer relevant filing makes that history stale or unsupported. A transition between return types ends the IRS series. Otherwise the whole financial series comes from the ProPublica Nonprofit Explorer summary data. We never combine providers within one series or fill a missing IRS value from the summary. ProPublica also supplies the organization metadata and available filing links. It powers search when the optional local index is disabled and is the wholesale search fallback when that index is enabled. The frozen IRS Exempt Organizations Business Master File index may then supply search. For an exact EIN that ProPublica cannot identify, it may also supply registry identity. Its asset, income, and revenue summary fields are not report financials and are never used here.

Fiscal-year labels use the year in which the tax period ends. IRS archive years describe publication cohorts, not tax years. Eligible IRS-backed reports show at most the latest five contiguous comparable annual periods. ProPublica fallbacks show at most the latest five distinct available summary fiscal years and may contain gaps. Reports can use only the IRS periods present in the active corpus.

Structured IRS XML remains the authoritative source for every filing-detail figure. When the selected rich filing has an IRS Object ID, the report links to ProPublica's complete filing viewer and keeps any available PDF as a separate source link. We do not blend PDF text or OCR into XML-derived figures. Paper-only filings continue to use the summary fallback; broader PDF/OCR extraction requires its own separately validated data release.

• Surplus / deficit — total revenue minus total expenses for each year.
• Programs & spending — the mission and the largest program services described in Part III, in the organization's own words, with the spending each reported; for grantmakers, the Schedule I grantmaking activity instead. Placeholder descriptions ("See Schedule O") are left to the verbatim detail rather than shown as programs. Form 990-EZ reports its own direct program and expense lines but not the long form's functional-expense split. Form 990-PF reports direct charitable activities and grants paid; grants approved for future payment stay in a separate table and are not counted as current-year grants paid.
• Operating reserve — net assets expressed as months of expenses at the selected period's annual rate. This is a rough comparison, not a measure of cash available for operations: where the filing reports it, we also show what the balance is made of (cash and savings, investments, property) and how much carries donor restrictions. Form 990-PF instead keeps book values and fair-market value distinct and presents them as foundation holdings; it does not translate those assets into public-charity operating-reserve months.
• Debt ratio — total liabilities divided by total assets: of everything the organization owns, the share that is owed to others.
• Revenue mix — each revenue category (donations & grants, program services, investments, other) as a share of total revenue, tracked across years. For Form 990-EZ, the report also shows the form's directly reported revenue lines without recasting them as long-form contribution subcategories.
• Spending mix — the Part IX split of total expenses into program services, management & general, and fundraising, plus the largest itemized expense lines (shown within Programs & spending). The split is shown as context; it is not scored against a watchdog threshold, and it does not measure program quality. This functional split is long-form-only.
• Who runs it, and what they're paid — the actual people listed in Part VII (officers, directors, key and highest-paid staff) and their reported compensation, including the highest-paid leader, read against the size of the organization. Form 990-EZ uses its Part IV officer, director, trustee, and employee rows; Form 990-PF uses its Part VIII rows. Related-organization compensation is not implied for either shorter detail tier.
• How it's governed — Part VI board size and the share that is independent, accountability policies (conflict-of-interest, whistleblower, document retention), whether the financial statements were independently audited with board oversight, whether the filing reports a significant diversion of assets, and staff and volunteer counts. Form 990-EZ has no Part VI equivalent, so this section is omitted rather than inferred from other EZ questions.
• Grants made — the Schedule I grants to other organizations, totaled and ranked, so you can see where the money is passed through. An aggregate grants-paid line on Form 990-EZ is shown only as an expense; it is never turned into recipient rows. Form 990-PF Part XV rows describe that foundation's own grants and never feed the reverse matched-grant index. To keep very large filings usable, a report displays the largest 200 paid rows and the largest 200 future-approved rows while always labeling the filing's complete reported row count and aggregate amount.
• Revenue growth — the trend in total revenue across the available years.

Comparisons use one group throughout the report. Organizations must have the same supported return type, IRS subsection and three-character NTEE classification, positive annual expenses between half and twice this organization's, and filing periods ending within 12 months. The target and duplicate filings are excluded. Only exact annual IRS XML filings with matching BMF tax periods qualify. Report identity and selected filing must also agree with the comparison record.

We start with addresses listed in the same state. When at least 10 organizations qualify, we use that state; otherwise we widen to the same Census division. We select up to 25, prioritizing the same state, then proportional closeness in annual expenses, with EIN as a deterministic tie-breaker. Territories remain within the same territory. There is no national fallback. Regional comparisons state how many members are in the target state. Listed addresses may differ from service areas; shared classifications do not establish identical programs or outcomes. Category labels use the original core codes in the IRS Form 1023 instructions.

An average requires at least 10 selected peers, with at least 8 reporting the measure and 80 percent reporting coverage. Each organization has equal weight: percentage averages are averages of each organization's percentages, not shares of pooled dollars. The target must report the same measure. Missing values remain missing, and reported zeros remain zero. Form-specific limits apply: 990-EZ and 990-PF do not supply the long-form spending breakdown, and 990-PF revenue is not forced into the public-charity revenue mix.

Up to five selected members link to their Readable 990 reports, with filing names, locations, annual expenses and fiscal years. Those reports may select a different filing period. Verified links may remain when fewer than 10 peers qualify, but no average is shown. Unavailable comparisons are explained without guessing an exclusion reason. BMF financial fields do not enter any comparison. These figures provide context, not ranks, scores, grades, recommendations, or measures of program quality. An average can still be influenced by unusually large or small values within the group.

The written readout at the top of each report is an AI-assisted summary generated from the computed figures above — not from the raw document, and not from any outside source. It is meant to orient a reader, not to certify or rate an organization. Always verify important decisions against the original filing.

A Form 990 is a tax filing, not a full audit. It can lag a year or more behind real events, may contain filer errors, and omits a lot of context (program outcomes, the liquidity of reserves, related entities). Very small nonprofits file a short 990-N “postcard” with no financial detail. Form 990-EZ filers report a narrower four-section detail tier when their e-file is present in the active corpus; Form 990-PF has its own four-section foundation-detail tier when a supported PF e-file is present. The active corpus can lag parser support, so filing-detail sections cover only organizations whose supported IRS e-file is in that corpus. Financial trends fall back wholesale to ProPublica summaries when the available IRS periods do not meet the comparability and freshness rules. Treat Readable 990 as a starting point for questions, not a final verdict.