grant-proposal-craft

Grant Proposal Craft

How a proposal actually gets scored, and how to write to the scorer. Grounded in a 2026 workshop for Indian researchers, but the mechanics are funder-neutral — the same discipline applies to NIH, NSF, ERC, DST, or a foundation.

A proposal is a sales document and a methodological argument at the same time. The reviewer's job is to form an opinion fast; every section should lower the cost of that opinion in your favor.

Hard rules

1. No fabricated citations. Every cited precedent must resolve to a verifiable record (DOI, PMID, arXiv ID, ISBN, URL with retrieval date). A gap in the literature that is a claim, not a verified absence, must be phrased as "as of this review" and the search logged.
2. No claim without provenance. Every quantitative claim — a budget line, a projected n, a timeline date, an acceptance figure — must point to a table, the OIM/circular, or a dataset the agent can show. "Trust me" is not provenance.
3. No silent failure. Every tool or lookup used to build a section (a literature search, an agency OIM read, a budget calculation) must report what happened on non-zero. A "we found no precedent" that came from a failed query is a false gap; say so.
4. Title last. The title is written after the body, when the point is settled. It must be clear, informative, and interesting without heavy jargon.
5. Concrete beats plausible. Every objective, method, and expected outcome is specific enough that a skeptic can test it. Vague aims are the most common reason a strong idea reads weak.

When to use

• You are drafting the body of a research grant proposal and need the structure a reviewer actually scores (any funder).
• You are critiquing a draft and want a dimension-by-dimension pass (novelty / usefulness / methodology / who) with the specific gaps named.
• You need the lay-summary / abstract / title discipline, or a pitfall and budget realism check before submission.

The four scoring dimensions

Most structured reviews (and every workshop rubric) score four independent axes. Write to all four; a proposal that nails novelty but ignores who loses to one that is merely competent across the board.

1. Novelty — is there something new? A funder wants to fund a gap, not a rerun. Frame the gap concretely, and if the closest prior work is nearly identical, do not apply — reframe or pick another scheme.
2. Usefulness / relevance — what does it buy the field or society? Stated plainly, in terms the reviewer's discipline values.
3. Methodology — is the plan doable? Methods are specific, and each anticipated failure has a named contingency (a "pitfall plan").
4. Who (PI fit) — is this the person to do it? A track record + a genuine reason this PI, not the next equivalent one, is the right owner.

The why/how/who spine: why now (the gap + its value), how (methods + pitfalls), who (the PI's fit). A good proposal answers all three before the reviewer asks.

Writing the sections

• Title — last. Clear + informative + interesting, no jargon. Test: can a smart non-specialist tell what the project is from the title?
• Abstract / synopsis — the movie trailer, not the credits. It must make the point land in the first two minutes: gap → what you'll do → why it matters. Lead with the point, not the background.
• Lay summary — "the parent test": write it so a parent with no training in the field understands the problem and why it's worth doing. Zero jargon, no acronyms unexplained.
• Project description — objectives first, then rationale, primary data, methods, expected outcome.
  • Objectives: SMART — specific, measurable, achievable, relevant, time-boxed. Each maps to a deliverable.
  • Rationale: the concrete gap (sourced), not a survey of the field.
  • Methods: the specific approach, and — the signature of a strong proposal — the anticipated pitfalls and the named remedy for each.
• Anticipated pitfalls — write them yourself, before the reviewer does. "If X fails, we fall back to Y because Z." A proactively managed risk reads as confidence, not as doubt.
• Deliverables & timeline — a Gantt chart or table aligning each objective to a date and a concrete output (a dataset, a model, a paper, a pilot). Objectives without dated deliverables are not fundable.
• Budget — split non-recurring vs recurring; stay inside the agency's OIM / circular caps on personnel, overhead, and contingency. Realism is scored — an inflated or infeasible budget flags the whole proposal.
• Supporting documents — letters of support, an AI/plagiarism declaration where required, resource list, CV, and a publication list. Assemble early; they gate the submission.

Specificity discipline

The single most common mistake, per the workshop: a vague aim + vague method + vague outcome makes an excellent-sounding project look unfundable. For every claim ask "specific enough to test?" and answer in one clause.

Propose a better approach, not an "alternative." Positioning your work as "we also did X" is a claim of mere addition; positioning it as "the existing method breaks in case C, and here is why our approach is the right fix" is a claim of necessity. Funders fund necessity.

Pitfalls

• Writing the title first and dragging the whole proposal into jargon.
• A background-lead abstract that buries the point — the reviewer's first two minutes are the whole first impression.
• Survey-style rationale ("X is important, Y is important...") with no concrete gap.
• No pitfall plan — the reviewer finds the risks for you and docks the methodology score.
• Deliverables without dates, or objectives that no deliverable maps to.
• A budget that ignores the OIM caps, or that is not internally feasible.
• "We also did X" positioning instead of the better-approach framing.

Validation

• Title appears last in your edit history and passes the "smart non-specialist" test.
• Abstract's first two minutes carry the point; lay summary passes the "parent test."
• Every objective is SMART and maps to a dated deliverable in the Gantt.
• Every anticipated failure has a named contingency.
• Budget sits inside the agency's OIM/circular caps, split recurring vs non-recurring, with each line traceable.

References

• Workshop source: local transcript grant_talk.txt (IISc grant manager, 2026) — four dimensions, why/how/who, title-last, abstract-as-trailer, lay summary, SMART objectives, pitfall planning, Gantt deliverables, budget realism, "better approach not alternative," ISTIP/IndiaBioscience discovery.
• SMART objectives: https://en.wikipedia.org/wiki/SMART_criteria
• Gantt timeline discipline: standard project-management practice.

Cross-references

• indian-grant-ecosystem — pick the scheme and read the landscape before drafting.
• grant-review — the reviewer's view of these same four dimensions.
• specific-aims — when the target is a US NIH-style specific-aims page.
• nih-r01 — the full R01 mechanics (foreign-scheme target).

(For the final provenance + citation pass before submission, run the meta/integrity-audit skill. Content skills in this repo do not cross-reference meta/ slugs as bullets, so it is referenced in prose here rather than as a - meta/... entry.)

Changelog

• 1.0.0 (2026-06-18) — Initial release. Funder-neutral proposal craft: the four scoring dimensions (novelty / usefulness / methodology / who), the why/how/who spine, title-last, abstract-as-trailer, lay-summary "parent test," SMART objectives, anticipated-pitfall planning, Gantt deliverables, realistic budgeting, and the "better approach, not alternative" framing, distilled from the 2026 workshop.