What Your Form 990-PF Reveals About Your Foundation

Most foundation leaders think of Form 990-PF as a compliance artifact — something the accountant files, the board approves, and nobody reads. That was a fair assumption in 2010. It is no longer true.

Since the Taxpayer First Act took effect, every private foundation files that return electronically, and the IRS publishes the resulting data as structured XML that anyone can download in bulk. Your filing is no longer a PDF sitting in a drawer. It is a row in a national dataset, sitting alongside every other grantmaker in the country, joinable to every nonprofit you have ever funded.

That changes what your 990-PF is. It is not a tax form anymore. It is the single most detailed public description of your strategy that exists — more detailed, in most cases, than your own website. Grantseekers read it. Consultants bill against it. Journalists mine it. And increasingly, analytical systems read all 205,000 grantmakers at once and compute things about your foundation that you may never have computed about yourself.

This guide walks through what is actually legible in your filing, what can be derived from it once it is joined to everyone else's, and why the honest response is not alarm but attention.

Why your 990-PF is more public than you think

Private foundations have always faced stricter disclosure than other exempt organizations. Under IRS rules, a foundation must make its returns from the past three years available to anyone who asks, in person or in writing, along with its exemption application and IRS correspondence. Critically, private foundations do not get the donor-privacy carve-out that public charities enjoy: as the IRS puts it, "identities of contributors to a private foundation are not exempt from disclosure" (IRS, Public Disclosure Requirements for Private Foundations).

The practical change came from format, not law. The Taxpayer First Act, signed July 1, 2019, extended mandatory electronic filing to essentially all tax-exempt organizations and directed the creation of an open, searchable file of those returns (Venable LLP). For private foundations specifically, the IRS applied the requirement to tax years ending on or after July 31, 2020. The IRS had already begun releasing e-filed 990 data in machine-readable XML in 2016; today those files are published through the IRS's Form 990 series downloads.

That date is worth holding onto when reading any national analysis, including ours. Coverage of e-filed returns is effectively complete from tax years ending mid-2020 onward; before that, a shrinking share of foundations still filed on paper, so the earliest years in any filing-derived dataset are thinner than the recent ones. Trends that begin in 2017 should be read with that in mind.

The distinction matters enormously. A scanned PDF is public in theory. An XML file is public in practice — parseable, joinable, and comparable at national scale without a single phone call.

What is legible on the face of the return

Before any clever analysis, a reader gets a great deal simply by reading carefully.

Part I — Revenue and expenses. Your investment income, capital gains, contributions received, and the split between charitable disbursements and operating expenses. The ratio of administrative expense to grants paid is one of the first things an outside analyst computes, and one of the most frequently misread.

Part II — Balance sheet. Assets at book and fair market value, and the composition of your holdings.

Part VII-A — Governance and activities. Whether you have engaged in political activity, made loans to disqualified persons, or changed your governing documents.

Part VIII — Officers, directors, trustees. Names, titles, average hours devoted, and compensation. This is the single richest field for anyone mapping who governs American philanthropy.

Part XV — Grants paid and approved for future payment. The heart of the return. Recipient name, address, relationship to the foundation, purpose of the grant, and amount, line by line. Also your application procedures, submission deadlines, and any restrictions on giving.

That last section is the one most foundations underestimate. Part XV is a complete, itemized, dated record of every grant you made — and unlike a curated annual report, it does not omit the grants that did not work out.

What becomes visible when your filing meets everyone else's

Reading one 990-PF tells you about one foundation. Reading all of them at once tells you something categorically different, because grants are edges in a graph: every grant connects a funder to a recipient, and recipients connect back to their other funders.

Once that graph exists, a set of behavioral facts about your foundation becomes computable — none of which appear anywhere on your own return:

• Whether you fund first or follow. Did money from you arrive before a grantee's other institutional funders, or after? Aggregated across a portfolio, this separates discoverers from followers.
• Where you are the only one. The count of grantees for whom you are the sole institutional funder. This is a measure of how load-bearing you are in your ecosystem — and of concentration risk for those organizations.
• Whether you held steady in downturns. Grant totals through 2020 and subsequent years, compared against the behavior of similarly sized funders.
• Who else backs your grantees. The set of funders whose portfolios overlap with yours, ranked by degree of overlap. Most foundations can name three or four. The filings typically reveal dozens.
• What happened next. Whether organizations you funded went on to raise from others — an association, never a claim of causation.
• How long you stay. Median relationship length with a grantee, and the share of grants going to organizations you had never funded before.

None of these require private information. They require the whole dataset and a willingness to compute across it. Plinth's public funder analysis does exactly that: 17,896,418 grants across 205,036 grantmakers, read from public IRS filings covering fiscal years 2017 through 2025.

The numbers that get computed about you

Here is a plain comparison of what a reader sees at each level of effort.

The gap between row one and row four is where most foundations' self-knowledge sits. It is common for a program officer to know their own portfolio intimately and have no idea that eleven other foundations fund the same six organizations — or that for nine grantees, no one else does.

Governance data is the most exposed section

Part VIII is worth a paragraph of its own, because foundations consistently underestimate it.

Officer and trustee names are public, and they are public for every filer. That means board membership is not one list but a national bipartite graph: people on one side, institutions on the other. Two foundations that share a trustee are connected whether or not either has ever noticed. Run across all filings, this produces the interlock map of American philanthropy — the paths by which an introduction can actually be made.

This is genuinely dual-use. It is the mechanism behind warm-introduction services, and it is also a governance fact your board should be aware of: conflicts, overlaps, and concentration in a small pool of trustees are all visible from the outside.

Three-quarters of independent foundations operate without paid staff, according to Urban Institute research on foundation expenses and compensation, which means for a large share of the sector, the trustee list is the organization.

What your filing does not say — and why that matters

An honest account has to include the limits, because the most common failure in foundation analysis is over-reading the data.

Filings lag. A 990-PF describes a fiscal year that typically ended 12 to 24 months before the data becomes available. Anything computed from filings is a portrait of the recent past, not the present. Any analysis worth trusting dates every figure to its fiscal year.

Grant purpose fields are inconsistent. Some foundations write a sentence; some write "general support"; some write nothing meaningful. Classification of purpose is inference, and should be labeled as such.

Recipient matching is imperfect. Names in Part XV are free text. "Boys and Girls Club" could be any of hundreds of affiliates. Good entity resolution handles most of this; nothing handles all of it.

Payout is not generosity. The qualifying-distribution calculation includes reasonable administrative expenses, not only grants — so a payout percentage is not a clean measure of how much reached charities. We cover this in detail in The 5% Payout Rule Explained.

Correlation is not causation. If grantees you funded later raised more money, that is an association. It may reflect your validation, or it may reflect that you pick organizations already on an upward trajectory. Both are interesting. Only one is a claim you can defend.

The fields that most often go wrong

Because the filing is now read at scale, small data-entry choices have outsized downstream effects. A handful of fields account for most of the misrepresentation foundations later complain about.

Recipient names entered inconsistently. If the same grantee appears as "Riverside Youth Ctr", "Riverside Youth Center Inc.", and "Riverside Youth Center" across three years, automated matching may treat them as three organizations — and your multi-year relationship reads as three one-off grants. Entering the grantee's legal name consistently, and ideally its EIN where your system supports it, is the single highest-leverage accuracy fix available.

Blank or boilerplate purpose fields. "General support" repeated 60 times tells a reader nothing, so they will infer your strategy from recipient names instead. Their inference will be cruder than the truth.

Address of the wrong entity. Some foundations enter the grantee's fiscal sponsor address, or a national headquarters rather than the local affiliate actually doing the work. Geographic analysis then places your giving in the wrong place — a common source of "we don't fund there" complaints about published maps.

Approved-versus-paid confusion. Part XV distinguishes grants paid during the year from those approved for future payment. Conflating them inflates or deflates apparent annual giving and distorts trend analysis.

Missing relationship disclosures. The relationship field exists to disclose connections between the foundation and a recipient. Omitting a genuine relationship is a compliance issue; over-declaring creates confusion.

None of these require extra work at filing time. They require the underlying record to be captured properly at the point of decision, which is a systems question rather than an accounting one.

What foundations should actually do about this

The productive response is neither anxiety nor indifference. It is to read your own filing the way an outsider would, before someone else does.

1. Pull your last three 990-PFs and read Part XV end to end. Not the summary — the line items. Most leaders find at least one surprise.
2. Compute your own concentration. What share of your dollars went to your top five grantees? Your top twenty? Boards routinely guess wrong by a wide margin.
3. Check your purpose fields. If your grant descriptions are blank or generic, you are letting third parties infer your strategy from recipient names alone. Writing one clear sentence per grant is the cheapest reputation work available to you.
4. Look at your sole-funder list. Identify the grantees for whom you are the only institutional money. That is a strategic fact and, for them, a risk.
5. Search for your own foundation on public datasets and see what is being said. Plinth's funder pages are free to search with no sign-up.

Foundations that treat the filing as a communications surface rather than a compliance chore end up better represented in every downstream system that reads it — because the data those systems read is the data you wrote.

Where software fits

If your grant records live in spreadsheets and email, producing an accurate Part XV is an annual archaeology project, and the purpose fields suffer accordingly. If they live in a system that captures purpose, geography, and outcome at the point of decision, the filing becomes a report rather than a reconstruction.

Tools like Plinth take this further by holding applications, decisions, payments, and grantee reporting in one place, so that the record behind your 990-PF is a byproduct of running the program rather than a separate exercise. Portfolio insights surfaces the concentration and pattern questions above against your live data, and impact reporting turns the same underlying records into board and public reporting. The same discipline that makes your filing accurate makes your annual report faster.

For foundations that want to see the outside view first, the public dataset is free to search, and the free tier includes full access to the underlying grants API.

Frequently asked questions

Is my foundation's Form 990-PF really public?

Yes. Private foundations must make the last three years of returns available to anyone who requests them, and the IRS publishes e-filed returns as machine-readable XML. Unlike public charities, private foundations cannot withhold contributor identities.

Can people see individual grants we made?

Yes. Part XV of Form 990-PF itemizes grants paid and approved for future payment, including recipient name and address, grant purpose, and amount. There is no threshold below which grants become invisible.

Are trustee names and compensation public?

Yes. Part VIII lists officers, directors, and trustees with titles, average hours, and compensation. Because this is public for every filer, shared board members across foundations are visible in aggregate.

How out of date is the public data?

Typically 12 to 24 months behind the fiscal year described, because of filing deadlines, extensions, and IRS publication schedules. Any credible analysis dates figures to their fiscal year rather than presenting them as current.

Does a low payout percentage make us look bad?

Not on its own, and reputable analysis does not treat it that way. Payout is calculated against average investment asset value and includes qualifying administrative expenses, so a single year's figure reflects asset valuation and timing as much as intent. Multi-year patterns are more meaningful than any one year.

Can we correct something that is wrong in a public dataset?

Usually yes. Reputable datasets cite the underlying filing for every figure, which makes errors traceable to either the source return or the processing. If the filing itself is wrong, an amended return is the fix.

What is the single highest-value change we can make?

Write a specific, one-sentence purpose for every grant. It costs almost nothing, and it determines how accurately every downstream reader — human or machine — describes what your foundation does.

Recommended next pages

• How to Benchmark Your Foundation Against Its Peers — building a defensible comparison set
• The 5% Payout Rule Explained — what qualifying distributions actually include
• Foundation Portfolio Analysis — seeing what your grantmaking adds up to
• Co-Funder Networks — who else funds your grantees
• AI for Funders — where automation helps across the grant lifecycle

Last updated: August 2026