How we calculate funder giving metrics

Data sources

Private foundations file IRS Form 990-PF every year. The IRS publishes the electronically filed returns (e-files) as machine-readable XML. From each return we read:

• The itemized grant list ("Grants and contributions paid during the year"): each recipient's name, city, state, status, purpose and amount.
• Financial totals: total grants paid, the distributable amount and qualifying distributions.

We keep up to five filing years per foundation. A "year" is the tax year of the return, which for many foundations is a fiscal year rather than a calendar year. Paper filings and returns the IRS has not yet released are not included, so the latest year shown can lag the current year by one to two years.

First-time grantees ("open to new grantees")

For the latest year on file we list every organization the foundation paid. A grantee is first-time if its name does not appear in any earlier year we hold for that foundation (for example 2020–2023 when the latest year is 2024). The share is first-time grantees divided by all distinct grantees that year; several grants to the same organization count once.

We only publish the figure when:

• the latest year has at least 10 distinct organizational grantees;
• there are at least 2 earlier years to compare against;
• each earlier year's list has at least 25% as many grantees as the latest year's. A short or truncated earlier list ("see attached") would make returning grantees look new, so we say nothing instead;
• no more than 50% of the latest year's grants went to individuals. Scholarship and hardship funds give to different people every year by design; "openness" does not describe them.

Grants to individuals are always left out of the grantee count. We treat a row as a grant to a person when the filer marks the recipient's status as individual ("I", "IND", "Individual"), when the name reads "College ITF Student" (in trust for), or when the status is blank or "N/A", the purpose is person-shaped (scholarship, tuition, hardship, funeral, rent, utility or heating help), and the name has no organization word in it.

Because our history window is at most five years, "first-time" means "not funded in the years we can see". An organization funded in 2015 and again in 2024 counts as new.

The state rankings ("Most open to new grantees in …") are stricter still: a foundation is ranked only with at least 20 itemized grants in each of its last three years and at least 20 distinct grantees in the latest, so a small funder with a handful of new names cannot top the list. Ties are broken by the larger number of grantees.

First vs. repeat grant size

For each grantee in the latest year we add up everything the foundation paid it that year, then average (and take the median of) those totals separately for first-time and returning grantees. A first-time average well below the repeat average usually means the foundation starts new relationships with smaller "trial" grants.

Repeat rate

The share of the latest year's grantees that the foundation also paid in the immediately preceding year. It is published under the same minimums as the first-time share, applied to that one preceding year. A high repeat rate means most of the budget renews existing grantees.

Giving trend

We compare total grants paid in the latest year with the total three years earlier, using the return's own total-grants line when both years have it and the sum of the itemized list otherwise. Growth of 20% or more is growing, a drop of 20% or more is shrinking, and anything in between is stable. Figures are not adjusted for inflation. One large multi-year pledge paid in a single year can swing the comparison.

Payout vs. the 5% rule

Private foundations must distribute roughly 5% of the average value of their investment assets each year. The 990-PF states that required minimum as the distributable amount, and reports what the foundation actually spent on charitable purposes as qualifying distributions (grants plus reasonable administrative and direct charitable costs). We show qualifying distributions divided by the distributable amount for the latest year that has both.

• Older e-files often report qualifying distributions as 0. We treat 0 as missing rather than "paid nothing" and fall back to the next most recent year.
• Below 100% in one year is not a violation: a foundation has until the end of the following year to meet the minimum, and can apply excess distributions carried over from the five prior years.
• Ratios above 10× (usually a spend-down year) are shown as "well above" rather than as a number.

Name matching and its limits

990-PF grant lists carry no recipient EIN, so we match grantees across years by name. Before matching we normalize each name: drop a trailing " - 2018 - GENERAL SUPPORT"-style suffix and any "DBA …" alias, lowercase it, turn "&" into "and", remove punctuation, and drop a leading or trailing "The" and trailing corporate words (Inc, Inc., Incorporated, Corp, LLC, Ltd…). Placeholder recipients like "Various" or "See attached" are ignored.

What this can get wrong:

• Overstating newness. A grantee that changed its name, was spelled differently, or was listed under a parent ("University Foundation" one year, "University" the next) looks new. This is the main source of error, so treat the first-time share as an upper bound.
• Understating newness. Two different organizations with the same normalized name (two "Boys & Girls Club" chapters listed without a city) look like one returning grantee.
• Filing gaps. A foundation that skipped or paper-filed a year has that year missing from its history.

Update cadence

Profile figures are computed from the grant lists whenever a profile page is regenerated, which happens within a few days of a new 990-PF being loaded. The state rankings ("Most open to new grantees in …") are recomputed in a batch after each 990 data load, so they can briefly trail the profile pages.

Spotted a figure that looks wrong? Email hello@grantable.co with the funder's name and we will check it against the filing.