The self-grade
Both samples, marked against our own rubric, weak scores first
A self-grade is worth less than an independent one. It is published anyway, because a sample delivered without a score invites you to assume the writer thinks it is perfect, and we do not. Two dimensions are 3s. The reasons are specific and they are the most useful thing on this page.
Graded by the writer against the practice's own winning-proposal rubric, 31 July 2026, calibrated against four earlier graded samples. Sample 4 is the failing anchor.
Start with the weak ones
The two scores that are not 5s, and exactly why they are not
Both are the same dimension, funder fit, on both documents. Neither is a writing problem. Both are a limit of what a portfolio sample can honestly do without asserting figures about a real foundation that we cannot cite, and asserting those figures is the exact defect the rubric holds for.
Funder fit
The letter quotes a real published priority, "Maintain Healthy Bodies and Minds", and a real published cornerstone, and the geography and cause area match. That earns the fit. What it does not do is what the rubric's probe demands: check the ask amount against the published range, the deadline against the cycle, the program name spelling, the submission email.
We have not verified that $74,985 sits inside The Colorado Health Foundation's published range for this priority, and we have not confirmed that the Foundation accepts inquiry letters ahead of application. In client work those four checks happen before the first sentence is written, from the funder's guidelines and their Form 990 Schedule I. In a portfolio sample we would have to assert a range we cannot cite. So the dimension caps at 3 and stays there. It is capped by the sample's nature, not by the writing.
Funder fit, and it is worse here
Everything said about sample 5 applies. One thing more. A real full proposal is written to the funder's own application questions, in their order, in their fields, inside their character limits.
This proposal is written to a general foundation structure and weighted to a published federal rubric. It demonstrates how sections get built and budgeted. It is not a submission-ready document for this or any funder, and turning it into one would mean restructuring it around their form. That is real work and this sample has not done it.
Sample 5
The letter, dimension by dimension
Letters are scored on five of the seven dimensions. Bar is 20 of 25, and every dimension must be at or above 3.
1 Need, 5.
The county death rate is external, dated (provisional 2025, released April 2026) and attributed inside the sentence rather than in a footnote. The 41 unanswered pages, 18 later found, 23 never contacted, is the organization's own door as evidence, and it comes from a named artifact, the program manager's hand-kept shift log. There is a clock on it: the SOR subaward ends 30 September 2026. Deleting "Pueblo County" would not make this paragraph reusable, which is the rubric's own probe.
3 Program logic, 5.
Who is eligible, what they receive, how long for, delivered by whom. The causal claim is stated once and plainly: the 30-minute response works because willingness decays within hours. Cohort times unit cost reconciles against the ask.
5 Budget, 4, and not a 5.
The letter names every component of the ask and gives the total's provenance, which is a stronger arithmetic signal than most letters of inquiry carry. It does not give per-line unit costs. The standing rule is that every ask gets one sentence of unit logic, minimum. This has that sentence. It does not have the table. That is the difference between a 4 and a 5 in a two-page letter. It is a defensible compression rather than an omission, and it is still not a 5.
7 Voice, 5.
Read the third and fourth paragraphs aloud. The rhythm alternates. The image, the phone ringing at an empty desk, carries information rather than sentiment. No jargon, no triads, no negative parallelisms. Two sentences exceed 32 words and both are citation-carrying. The close gives the funder a way to say no and means it.
Sample 6
The proposal, dimension by dimension
All seven dimensions. Bar is 28 of 35, and every dimension must be at or above 3.
1 Need, 5.
One community condition, per ACF's scored restraint clause. The 388 emergency department presentations and the two-thirds overnight concentration come from the partner hospital under the MOU, which means the number has a document behind it. The last line of the third paragraph is the whole case: "The hours we do not staff are the hours the problem happens."
3 Program logic, 5.
The strongest single line in either document is in section 4: "This is not a new program. It is the existing program running during the hours it currently does not." That sentence is what the rubric means by a bounded slice of a proven whole. The volume projection reconciles: 214 across 2.4 peer FTE is 89 per FTE, the projection is 60, and the reason for projecting under the demonstrated rate is stated.
4 Outcomes, 5.
Every objective has an indicator, target, basis and source. Objective 2 targets 75% against a demonstrated 79% and says why. Objective 3 holds flat at the demonstrated 60%. Outputs and outcomes are separated, and section 5 says which outcome will not close until after the grant period ends rather than reporting a partial number as final. The evaluation section states that unknown 90-day status counts as not retained, which costs the applicant reported percentage points on purpose.
5 Budget, 5. Verified by hand.
44,720 + 10,956 + 4,661 + 7,400 + 6,300 + 948 = 74,985. Fringe checks: 44,720 × 0.245 = 10,956.40. Supervision checks: 62,400 × 1.245 = 77,688, × 0.06 = 4,661.28. Indirect checks: 74,985 × 0.148 = 11,097.78. Cost per participant checks: 74,985 ÷ 60 = 1,249.75. Every line answers "why this amount and not half of it." Marginal cost of $1,250 is reconciled against all-in cost of $2,855 with the difference explained rather than presented as efficiency. The request declines indirect and names what it declines. Our own graded sample 2 was caught with an unexplained gap between 20 × $1,650 and a $40,000 ask. This budget exists partly to not repeat that.
6 Sustainability, 5.
Percentages, not gestures: 46 / 22 / 19 / 9 / 4. The Medicaid recovery estimate is specific, sourced to FY2025 billing actuals, and states what it does not cover, $31,176. The county contract renewal is dated. The prior-transition evidence is concrete: two multi-year federal awards ended, 2014 and 2019. And it refuses the standard lie: "We are not offering you a plan where the program becomes self-funding, because these are recurring costs and recurring costs need recurring money."
7 Voice, 4, and here is where it is weakest.
Sections 2, 4, 8 and 9 read the way they should. Section 3 is a personnel list with bolded lead-ins and reads like an org chart in prose. Section 11 is a bare enumeration. Section 6, the logic model, is a table and cannot help being one, but the sentence introducing it is flatter than the rest of the document. The bolded inline lead-ins throughout section 4, "Who is eligible.", "What happens.", "For how long.", are a defensible proposal convention, since reviewers score sections independently and headings help them find things, but it is also the exact pattern a humanizer flags as machine-shaped. It stays for scoring reasons and it costs a point on voice. That trade is deliberate and it should be visible.
Red flags
Eight things that stop a document from shipping, checked one by one
Any one of these is an automatic hold in client work. Both samples were run against all eight. One is documented rather than clear, and it is the declared premise of the exercise rather than a defect hiding in it.
| # | Flag | Status |
|---|---|---|
| R1 | Fabrication or unverifiable claims | Documented, not a hold. Every internal figure is invented, because the organization is invented, and the banner on both pages says so before anything else. The county statistics are constructed to demonstrate sourcing form. In client work this would be an automatic hold until every one traced to a live citation. Here it is the declared premise of the exercise. |
| R2 | Guarantee or contingency language | None. Both documents state what the program is designed to do at rates it has already produced. Sample 6 section 9 volunteers a condition under which the applicant would stop. |
| R3 | Copy-paste residue | None. Every proper noun checked. One funder, one applicant, one program, one partner, one evaluator, across both documents. |
| R4 | Generic boilerplate doing a section's job | None. Delete "Pueblo" from any section and it stops making sense. |
| R5 | Personally identifying information | None. No beneficiary or donor is named or described identifiably. |
| R6 | Overclaimed track record | None, and this is the flag the portfolio exists to respect. Nomar Grants claims no award, no client and no relationship with any named organization, in the header of both documents. The applicant's own record is stated as reviewed-standard, a clean opinion and a Single Audit with no findings, rather than "a proven track record of excellence." That upgrade is exactly what sprang sample 4. |
| R7 | Limit violations | Not applicable. No live funder's limits are being obeyed or broken, because no live application is being made. Sample 5 holds to its stated 2 pages. |
| R8 | Ask arithmetic broken | None. Verified by hand above and reconciled to the dollar across both documents. |
With the funder-fit cap on both disclosed to any prospect who reads them, which is what this page is.
Why publish this
A score you can only see if the writer chose to show it
A sample delivered without a score invites the reader to assume we think it is perfect. We do not.
Nobody audits this. There is no requirement to publish a rubric, no requirement to publish the marks, and certainly no requirement to lead with the two lowest ones. The reason to do it anyway is that a grant writer's whole job is deciding which claims about your organization are defensible, and the cheapest way to show that judgment is to point it at our own work first.
The two 3s are not a writing failure. They are the exact boundary where honest work stops and invention would have to start. That boundary is the product.