--- name: funder_prospect_research description: "Research grant funders before writing anything: build the prospect list from peer recipients, Candid profiles, IRS Form 990-PF filings, and Grants.gov, run every prospect through five hard gates (eligibility, geography, exclusions, access route, deadline runway), score the survivors on a 100 point fit system where ask size is computed from the funder's own similar grants, and track deadlines and contacts on a grant calendar. Ships a tested scorer and calendar checker, a prospect sheet and calendar template, a guide to reading a 990-PF, and a worked example with four fictional prospects landing on pursue, test, park, and cut. Use when building or cleaning a grant prospect list, qualifying a funder before an LOI or proposal, sizing an ask, or deciding which funders to stop chasing." --- # Funder Prospect Research ## Purpose Decide which grant funders deserve your writing time, and prove it with the funder's own record before a single proposal paragraph gets written. Most wasted grant work is not bad writing. It is a good proposal sent to a funder that was never going to pay for that work, in that place, at that size, from that kind of applicant. This skill runs the research in a fixed order: define the work, find prospects from evidence, verify each one against primary sources, gate the ones that cannot fund you, score the ones that can, size the ask from the funder's paid grants, and put the survivors on a calendar with owners. One full example ships in `references/worked-example.md`, with its prospect sheet and calendar in `assets/`, so every number in it can be rechecked with the scorer. ## Workflow 1. **Write the research paragraph first.** One short paragraph each for subject, population served, geography at county level, support type and honest project cost, and your exact applicant type (for example 501(c)(3) nonprofit, school, government body, fiscally sponsored project). Every later step checks against this paragraph. The full method is in `references/sources-and-searches.md`. If the paragraph cannot be written plainly, the project is not ready to research. 2. **Build a wide prospect list from evidence, not directories alone.** Start with peer recipients: organisations like yours list their grant funders in annual reports and filings, and a funder already paying for similar work has passed the hardest test. Then search Candid grantmaker profiles by subject, geographic area served, population, and support type. Candid's own fit check is subject and geography, grants data, and key people. Then work the free routes: IRS Tax Exempt Organization Search, ProPublica Nonprofit Explorer, Grants.gov for federal opportunities, SAM.gov Assistance Listings for federal programs, state and community foundation programs, and corporate giving pages for employers in your service area. Aim wide here. The gates do the narrowing. 3. **Verify each prospect against a primary source.** For a private foundation, that is its Form 990-PF: every private foundation files one yearly and the form is public. Read it in the order taught in `references/reading-a-990pf.md`: Part XV line 2 (does it take unsolicited applications at all), Part XV supplementary information (who applications go to, deadlines, restrictions on geography, field, and institution type), Part XV line 3a (grants paid, by recipient, purpose, and amount), line 3b (grants approved for future payment), and the officers and trustees list (the honest access map). For a federal opportunity, read the notice of funding opportunity itself: Grants.gov states plainly that an ineligible applicant wastes the time and money spent applying. For any funder, the current guidelines page outranks older material on direction, and filings outrank marketing on amounts. Record the source and the check date for every fact in the sheet. 4. **Run the five hard gates.** Eligibility, geography, exclusion list, access route, deadline runway (at least 7 days for a letter of inquiry, at least 21 days for a full proposal, unless near final materials already exist). A failed gate cuts the prospect at any score, except runway, which parks a strong fit for the next cycle. The gate definitions are in `references/fit-scoring.md`. Record the failed gate and the source in the sheet so the dead end never gets researched twice. 5. **Pull the similar grants and size the ask before scoring.** From the funder's grants paid, list the grants most like yours: same subject, same geography, similar recipient. Use the last filing year, plus the two before it if the list is short. Run `python3 scripts/prospect_score.py stats --grants '10000;15000;20000;25000'` on the list. Set a first ask at or below the median similar grant. Candid's guidance is that grants data should shape the outreach, the proposal, and the donation size. An ask above the largest similar grant the funder has made is not ambitious, it is ungrounded, and the scorer gives it 0. 6. **Score the survivors and let the script compute the verdict.** Fill one row per prospect in `assets/prospect-sheet.csv`: gate answers, the six hand scored fields (mission 0 to 25, population 0 to 15, geography strength 0 to 10, grant history 0 to 20, support type 0 to 10, access 0 to 5), the similar grants list, and the ask. Ask fit (0 to 15) is computed from the median and the 75th percentile of the similar grants, never scored by hand. Run `python3 scripts/prospect_score.py score --input assets/example-prospects.csv` with your file in place of the example. Bands: 75 and up pursue, 55 to 74 test with a letter of inquiry, 35 to 54 park, under 35 cut. Every score needs evidence in the notes. A score from memory is a guess wearing a number. 7. **Work the verdicts, cheapest test first.** For test band prospects, the next step is a letter of inquiry where the funder takes one: a one to two page summary covering mission, amount requested, program details, needs, and outcomes. That is the format funders themselves describe for an LOI first step. For pursue band prospects, go straight to the funder's stated process. For park band and runway parked prospects, record what has to change and the date to look again. For cuts, record the reason in one line and stop spending time. 8. **Put every live prospect on the grant calendar.** Copy `assets/grant-calendar.csv` and enter each stage as its own row: LOI, full proposal, report, call. Include owner, status, and where it gets submitted. Check it with `python3 scripts/prospect_score.py calendar --input assets/grant-calendar.csv`, which sorts by due date, flags overdue and due within 30 days items, and fails if any funder's LOI lands after its full proposal. Recheck every date against the funder's current guidelines before acting on it. Calendar dates copied months ago are leads, not facts. 9. **Record contacts properly and keep the file clean.** For each live prospect, name the contact and role from a public source: funder staff page, Candid profile, filing officer list, or the opportunity notice. One prospect, one row, one owner. Rescore when guidelines change, a new filing year lands, your program changes, or a declined LOI comes back with feedback. Review parked prospects quarterly and delete nothing: a closed row with a reason is next year's saved week. ## Output Contract - The prospect sheet as a CSV in the format of `assets/prospect-sheet.csv`, one row per prospect, with gate answers, scores, similar grants, ask, source check date, owner, status, and next step filled from evidence. - Scorer output on the final sheet showing a verdict per prospect and RESULT PASS, with every gate failure and oversize ask printed with its reason. - The grant calendar as a CSV in the format of `assets/grant-calendar.csv`, checked with the calendar command, sorted by due date, with no LOI after its full proposal. - For each pursue or test prospect: the sized ask with the similar grant median behind it, the named contact and role with its source, and the next step with a date. - For each cut or parked prospect: one line naming the gate or the score band and what would have to change, so the decision holds up next cycle. - A short research note listing the sources searched, the search terms used, and the filing years relied on, so the whole search can be repeated or audited. ## Operating Rules - Gates before points, evidence before scores, the funder's record before the funder's marketing. Stated interests open a conversation. Paid grants close the argument about fit. - Never reshape the work to chase a funder. Describe the project as it is. A stretched description that wins a grant for work you do not do costs the money, the relationship, and the organisation's name. - The ask comes from the funder's similar grants, not from the size of your need and not from the largest grant on its list. First ask at or below the median similar grant, unless an invited process or a prior grant record supports more, and that support is written in the sheet. - Prospect research ethics are fixed rules, not style. Collect information lawfully, record it factually and only where it is relevant to the work, attribute it to its source, store it in the organisation's own files with access limited to the people working the prospect, and never carry a prospect file to another organisation. Those are the Association of Fundraising Professionals standards for prospect research. - Business contact details from public funder sources belong in the sheet. Personal details about individuals do not. Funder staff research needs a name, a role, and a work contact route, nothing more. - No percentage based pay and no finder fees for prospect research or grant writing help, staff or consultant. Fixed fee or salary. Percentage pay rewards inflated asks and oversize applications, which is the failure this skill exists to prevent. - Dates and deadlines get rechecked against the funder's current material before anything is sent. A filing tells you what a funder did. Only current guidelines tell you what it wants now and when it wants it. - If the honest result is a short list, ship a short list. Three pursued prospects with sized asks and calendar dates beat thirty hopeful names. The worked example starts with four researched prospects and ends with two live actions, one park, and one cut, and that is a good week of prospect research.