Loading…
Loading…
· Public charity
Wikimedia fdn supports wikipedia, its sister projects.
Every grant clustered by its grantee’s IRS cause code (NTEE), by year — across FY2018–2023.
Beneath the NTEE codes, this is what the grant descriptions actually say.
…and in their own words, year by year
Words distinctive to each year’s grant purposes (TF-IDF) · event-driven terms in cyan.
Your grants by size, and where they go.
By grant size · FY2023
Each grant placed by the hardship in its grantee’s ZIP, then read against the area’s typical level.
Dollar for dollar, your grants (FY18–23) land where the poverty rate runs at 11% — the area typically sits at 11%. 10% of your dollars reach neighborhoods with above-average need. Your grants skew toward lower-need ZIPs.
Matched: human-services grants are those whose grantee’s IRS cause code is Human Services. Placed by people below the poverty line in the grantee’s ZIP, averaged from the tracts in that ZIP; grantee ZIP can differ from where services are delivered.
US grants placed by each grantee’s ZIP, which can differ from where services are actually delivered. Need from public data — U.S. Census/ACS, CDC PLACES, Eviction Lab, USDA, United Way ALICE — shown as context about the area, never attributed to your giving.
The same grants, placed two ways — where each recipient sits, and where its stated purpose earmarks the money.
About 71% of WIKIMEDIA FOUNDATION INC’s grant dollars are earmarked, by their stated purpose, for a different county than the recipient’s own address — money that lands at a nonprofit in one place but is meant to do its work in another. Read by recipient address, 80% of the giving stays in CA; read by stated purpose it is 26% — less of the work is directed home than the recipients' locations suggest.
Recipient view: each grant at its grantee’s ZIP, mapped to a county. Directed view: each grant at the county its stated purpose names, falling back to the recipient’s county when the purpose names no place; US grants only. A county shows only if it carries the top 95% of that view’s dollars. Purposes are read from the foundation’s own 990 grant descriptions.
Which organizations you funded more than once, and which you funded a single time. Each grantee is matched to its own filings.
And the relationships you keep tend to grow: grantees you re-up on have seen median revenue of +48% since the first grant, against +16% for the ones you funded once.
21 repeat relationships — 13 still active in FY2023, 8 since wound down; 4 grantees were first funded in FY2023 (too recent to call).
How the two cohorts compare
Organizations
Total granted
Median revenue growth · since first grant
Still filing today
New vs renewed · share of each year
In FY2023, 94% of grant dollars renewed an existing relationship; $234k went to new ones.
Where new relationships form · theme of each grantee’s first grant
First-time = a grantee’s first year in your filing window; renewed = funded in an earlier year too. As a portfolio matures the renewed share naturally climbs — once funded, an org stays “renewable” — so the signal is the years that buck it (a new-grantee intake wave). The earliest year is left-censored: relationships that predate the data read as “new.” Theme is the grantee’s IRS cause.
Repeat = funded in two or more distinct years; growth and survival are read from each grantee’s own subsequent IRS filings.
The map has a center of gravity: the average direction from the rest of the sector toward the organizations the foundation funds — the shape of its giving. Scoring every nonprofit along that direction surfaces the ones that look most like the portfolio. These are the closest matchesthat aren’t grantees — a resemblance in what they say they do, not a recommendation.
Nyu is a private university with approximately 60,000 matriculating students in 20 schools and institutes.
The mission of mit is to advance knowledge and educate students in science, technology, and other areas of scholarship that will best serve the nation and the world in the 21st century.
Offer undergraduate and graduate degrees in liberal arts and professional programs.
To advance the scientific study of language and its applications.
Education and research
To advance scientific knowledge and to increase the number of scientifically informed individuals.
The National Gallery of Art serves the nation by welcoming all people to explore and experience art, creativity, and our shared humanity.
The washington university is a co-educational, nondenominational university with a long and distinguished history of teaching, research and (cont'd on schedule o)
Dedicated to the improvement and advancement of graduate education.
The American Academy of Sciences and Letters (AASL) promotes scholarship and honors outstanding achievement in the arts, sciences, and learned professions. It encourages the fruitful exchange of ideas within academia and society at large…
Enhance Columbia University's education mission by acquiring and publishing outstanding original works by scholars and others that contribute to an understanding of global human concerns. Columbia University Press became part of the…
For reference, the grantee most central to the portfolio’s shape is Code for Science & Society Inc and the most unlike its peers is Yale University. Resemblance is measured on each organization’s own IRS 990 mission text; it reflects how work is described, not its quality or impact.
For each theme you fund, this compares how the sector’s money is shifting with how your own giving is shifting.
Sector change and giving change are each shown relative to their own range, growing (right) or shrinking (left); both normalized, so inflation isn’t mistaken for growth.
Your grantees are a median of 13 years old; the field is 14. You back the younger end — and your money leans older still.
The field is 24% startups (under 5 years old) — 9% of your grantees by number, and just 2% of your money.
The orgs you fund almost never close — 0.0% lost their exemption, against 12% of the field you don’t fund.
Age = years since IRS exemption (a founding proxy). “Closed” = auto-revocation for 3 years of non-filing — a floor on closure, not proof, and bigger established orgs lapse least. Association, not causation.
Each one resolved to its own IRS returns and tracked year by year — your grant beside their revenue.
Where your money sits — by cause, then by grantee
Each org by its size and your share of it — top-left is where you’re load-bearing
A decade per grantee — revenue shaded, your grants as bars, all rows on one timeline.
The grantmakers whose grantees overlap with yours far more than size alone predicts. Each orbits closer the stronger the alignment; the arcs between them show where they also fund each other. Here it reads as a tightly interlocked camp — most of these funders back each other's grantees too.
Open a dossier: The Ford Foundation · Mozilla Foundation · The Andrew W Mellon Foundation · Gordon E and Betty I Moore Foundation · John D and Catherine T Macarthur · Silicon Valley Community Foundation · New Venture Fund · Vanguard Charitable Endowment Program · National Philanthropic Trust · Fidelity Investments Charitable Gift Fund · Donor Advised Charitable Giving Inc · American Online Giving Foundation Inc
Affinity is a Gamma-Poisson posterior co-funding rate, re-centered on the typical rate, so thin evidence shrinks toward no signal. A research starting point: overlap is association, not proof of shared intent.
Every dot is one organisation WIKIMEDIA FOUNDATION INC funds. Left–right is the share of its income from government; up–down is the share from you. Bigger dots raise more. Filter to federal, a single department, or state — and drag the year to watch it move.
Government income is each org’s traced federal awards (USASpending — grants and contracts) plus state payments (open checkbooks) as a share of its total revenue (IRS Form 990); the self-reported government-grant line (990 line 1e) is carried for cross-check. An association, not a claim that your grant caused the public funding. Coverage is precision-first — a floor, not a census; state records exist for 9 states, so a grantee outside them shows no state figure (dimmed) rather than a false zero. Federal award amounts are obligations, which can span years, so a single-year share is indicative. 990 filings lag 12–24 months.
Through Plinth
This dossier was generated cold from public filings. In Plinth it’s a working system — every applicant assessed and every grant monitored. Here’s a taste, run on one of your grantees: Wiki Society of Washington DC.
Agentic due diligence · confidence × risk
Limited financial data in public filings.
4 years of Form 990 filings, still active.
US 501(c)(3); EIN 452106571 on file with current IRS Form 990 filings.
Board composition & governance documents — verified live in Plinth from the applicant.
OFAC / UN sanctions screening — run live in Plinth at assessment.
Staffing, M&E and activity alignment — assessed live in Plinth from submitted proposals.
Live · Plinth’s real DD engine
Run the actual assessment on Wiki Society of Washington DC, cold from public data.
Generated live from public IRS filings + open web sources by Plinth’s agentic engine. Shown as an illustration of the product, not a formal assessment.
Post-award monitoring · continuous checks
What you see here is static and public. In Plinth it’s operational — the full six-pillar framework on every applicant (with their own documents + live registry and sanctions checks), monitoring dashboards on every award, custom board reports, and eligibility routing for intake.
Preview generated from this grantee’s public IRS filings and independent reporting. Pillars marked “live in Plinth” require the applicant’s submitted documents. Shown as illustration, not a formal assessment.
Warm introductions · Powered by PlinthPro
Find your warmest path to Wikimedia Foundation Inc through people who sit on both boards. Search for your organization and Plinth traces the introduction across shared trustees and officers.
Every link is a documented governance overlap in public IRS 990 filings — not a personal network — and each hop carries its own confidence tier. Low-confidence or distant paths are held back rather than guessed.