Why a single fit score misleads
Most eligibility screens fail the same way: they average. A funder scores 9 out of 10 on mission alignment, geography, and grant size, so the team starts drafting, and three weeks later someone notices the guidelines say “invited proposals only.” The strong average hid a hard disqualifier.
The fix is structural, not motivational. Split the screen into two stages with different math.
Disqualifiers are binary and fail closed. A gate is a condition the funder controls and you cannot negotiate: tax status requirements, funding geography, whether unsolicited applications are accepted at all. Gates take exactly two values, pass or fail, and an unanswered gate counts as a fail until you find evidence. That is what fail closed means: missing information stops the screen instead of sliding through as a hopeful “probably fine.”
Fit factors are scored and weighted. Once every gate passes, the question changes from “may we apply” to “is this worth our hours compared to the other funders on the list.” That is a ranking problem, so scores and weights are the right tool there, and only there.
Two failure patterns make the separation worth enforcing. First, sunk research time converts weak prospects into pursuits: after six hours of reading a foundation’s filings, “no” feels like waste, so the screen bends. Second, teams stretch terminology, renaming an existing tutoring program “workforce readiness” because the funder uses those words. A gate row that asks “does a program you already run fit without renaming it” catches the stretch before it reaches a proposal draft.
Stage one: six gates that fail closed
| Gate | The question | Where the evidence lives |
|---|---|---|
| Legal status | Does the funder require a tax status you hold today, or accept fiscal sponsorship you already have in writing? | Funder guidelines, your IRS determination letter |
| Geography of benefit | Do the people your program serves live inside the funding geography? | Guidelines, program enrollment records |
| Program area | Does a program you already run fit the stated program areas without relabeling? | Guidelines, recent grants list |
| Applications accepted | Does the funder accept unsolicited proposals or letters of inquiry right now? | Funder site, 990-PF Part XV |
| Request size | Is your planned ask inside the funder’s observed grant range? | Guidelines, 990-PF grants list |
| Timing and capacity | Can you meet the next deadline with staff hours you actually have? | Funder calendar, your tracker |
Two of these deserve a note. Geography of benefit is about where outcomes happen, not where your office sits: a nonprofit headquartered in one county serving families in a neighboring one should screen against the second county. And the timing gate is a real gate, not a pep talk. If the deadline is in twelve days and the person who owns proposals has four free hours, the honest answer is fail for this cycle, and the funder goes back into research for the next one.
When a gate fails, record which one and why in your grant tracking spreadsheet before moving on. A dated “failed gate 4, preselected grantees only, per 2024 filing” saves the next person from repeating the same six hours of research next year.
Stage two: fit factors worth scoring
Score each factor 0, 1, or 2, multiply by its weight, and sum. The matrix uses eight factors with weights of 1 to 3; the maximum is 32.
| Factor | Weight | What a 2 looks like |
|---|---|---|
| Priority alignment | 3 | Explicit match in the funder’s own current words |
| Past grantee similarity | 3 | Multiple grantees of your size and type in the last two filings |
| Grant size match | 2 | Your ask sits within the middle half of recent grants |
| Renewal behavior | 2 | Repeat grantees appear across consecutive filings |
| Relationship or warm path | 2 | A board member or partner has a real, current connection |
| Burden versus award | 2 | Short application for a meaningful award |
| Payout and asset trend | 1 | Total grants paid stable or rising across filings |
| Reporting expectations | 1 | Requirements you already meet for another funder |
Our working thresholds: pursue at 22 or above, hold and revisit between 16 and 21, drop below 16. Set your own if you like, but set them before scoring a specific funder, because thresholds chosen after the fact always confirm the decision someone already wanted.
A zero on relationship is normal and is not a problem to fix by inventing one. The factor exists because a genuine warm path changes how you open a conversation, not because cold applications are doomed.
Grant eligibility matrixAll six fail-closed gates and all eight weighted fit factors, each with its evidence question, source, and recording rule. Open it in any spreadsheet tool and complete one copy per funder.
CSV templateHow to read a 990-PF for real signals
Private foundations file Form 990-PF annually, and it is the closest thing to ground truth about what a foundation actually funds, as opposed to what its website says. The IRS describes the form’s purpose and contents on its About Form 990-PF page, and Candid’s guide to finding 990s and 990-PFs lists the places to retrieve filings free, including Candid’s own search tools and the IRS Tax Exempt Organization Search.
Four places in the filing feed the matrix directly:
- Part XV, the application information section. Foundations state here whether they accept applications and, if so, in what form and by when. The phrase “contributes only to preselected charitable organizations” is a hard fail on gate 4. This section alone settles more screens than any other source.
- The grants paid schedule. The full list of recipients and amounts. From it, pull the median grant (gate 5 and the size-match factor), look for organizations that resemble yours (grantee similarity), and check whether the same names repeat year over year (renewal behavior).
- Total grants paid, across two or three filings. A foundation whose grantmaking dropped by half may be winding down or reorienting; a rising line supports a better payout score.
- Officers and trustees. The names your board reads to find a genuine connection, which is how the relationship factor gets its evidence.
One caution absorbed from hard experience: a 990-PF is a snapshot that is usually one to two years old by the time you read it. Filings tell you patterns; only the funder’s current guidelines tell you rules. When the two disagree, the guidelines win, and a phone call or email to the funder beats both.
Worked example
One fictional nonprofit screened against two fictional funders
Cedar Bend Youth Alliance is a fictional nonprofit with a $420,000 budget and an after-school reading program serving 120 children in one county. It plans a $15,000 ask. Both funders below are fictional; the point is the mechanics.
Funder A, the fictional Harwood Family Foundation. Gates 1 through 3 pass: no unusual status requirements, same county, and children’s literacy is a named priority. The grants list even shows a median grant near $18,000. But Part XV of the most recent filing states the foundation contributes only to preselected organizations. Gate 4 fails, the screen ends, and no fit score is calculated. On a single-score screen, Harwood’s excellent alignment would have produced a high number and a wasted proposal.
Funder B, the fictional Bright Plains Fund. All six gates pass, so stage two runs:
| Factor | Score | Weight | Points |
|---|---|---|---|
| Priority alignment | 2 | 3 | 6 |
| Past grantee similarity | 1 | 3 | 3 |
| Grant size match | 2 | 2 | 4 |
| Renewal behavior | 2 | 2 | 4 |
| Relationship or warm path | 0 | 2 | 0 |
| Burden versus award | 2 | 2 | 4 |
| Payout and asset trend | 1 | 1 | 1 |
| Reporting expectations | 2 | 1 | 2 |
Total: 24 of 32, above the pursue threshold of 22. Cedar Bend logs the score, the two weakest factors, and the evidence links, then moves Bright Plains to qualified status in its tracker. The zero on relationship stays a zero; nobody manufactures a connection.
What each result means in practice
A completed matrix produces one of three outcomes, and each has a next action rather than a feeling.
Pursue means the funder enters your pipeline with a deadline and an owner. For a foundation that asks for a letter first, the next artifact is usually a letter of inquiry, not a full proposal.
Hold means the fit is real but not competitive against your other options this cycle. Record the score and the recheck date. Holds are where next year’s best prospects come from, because the research is already done.
Drop means a failed gate or a low score, written down with its reason. A documented no is an asset; an undocumented no is a question someone will re-research in eighteen months.
If this is your organization’s first time screening funders at all, run the matrix on three candidates before writing anything, and read our guide to your first grant proposal for how the screen fits into a 30-day path. And keep the boundary honest: a matrix disciplines where your hours go, but it measures fit, not outcomes, and it cannot guarantee that any funder, however well matched, will invite, fund, or renew you. What it can do is stop you from spending forty hours on applications that were ineligible before the first sentence was written.