ingridleiria Research Proposal and Grant
A proposal is not read the way a paper is read. It is read by someone with a scoring rubric in front of them and thirty-nine other applications behind it, who will spend between fifteen and forty minutes on yours, looking for specific things in a specific order. If a criterion they must score cannot be found, they score it low, because they cannot score what is not there and they will not go looking.
That is the failure, and it is why excellent research loses to adequate research. The proposal was written as an argument the applicant found compelling rather than as an answer sheet to the criteria the panel was given. A section the call required is absent, or is present under a heading the assessor did not expect. The feasibility criterion is answered with the word "feasible". The data the whole design rests on is described as "will be obtained", with no source, no status and no fallback, which is the single most common reason a funded project stalls and therefore the thing experienced panellists look for hardest.
The cost of the failure is not the rejection alone. It is a year, because most calls run annually, and the resubmission cannot begin until the next cycle.
When to use this, and when not to
Use it for any document whose purpose is to get research approved or funded: a doctoral thesis proposal, a departmental upgrade document, a national or international grant application, a fellowship, a scholarship research statement, a research plan attached to a job application, an internal seed fund, an ethics-adjacent project approval where the research design must be argued.
Use it also when a call has been received and it is not yet clear whether to apply, because the first two steps of the method answer that question at low cost.
Do not use it to build the underlying design. That is research-design, which produces the question, the hypotheses, the identification strategy, the equation and the exhibit list. This skill argues that design to a reader who has not seen the data, and it will expose gaps in it, but it is not a substitute for having done it.
Do not use it to defend a design against methodological attack; that is identification-defense, whose output supplies the threats-and-diagnostics paragraph here. Do not use it to prepare for the panel interview or the proposal hearing, which is thesis-defense-prep. Do not use it to write the paper that results; that is full-manuscript-build. Do not use it to choose a journal, which is journal-targeting.
Where the real question is whether the candidate has time to do the proposed work alongside everything else, that is thesis-advisor, and it should be answered before the proposal is written rather than after it is funded.
What you need before starting
The call, template or programme description, in full. Not a summary and not last year's version. Missing: fetch it from the funder's site; where no live lookup is available, ask for the PDF. Writing a proposal against a remembered call is how a required section goes missing.
The evaluation criteria and their weights, where published. Some calls publish the scoring sheet, and where they do it is the most valuable document in the process. Missing: infer the criteria from the required section headings and say that you have done so.
Eligibility, deadline and submission mechanics. Who may apply, whether an institutional sign-off is needed and how long it takes, whether the portal closes at a stated hour in a stated time zone. Missing: establish these first. Institutional sign-off routinely requires five to ten working days and is the most common cause of a proposal missing a deadline it was otherwise ready for.
The design document. Missing: build it with research-design before writing prose. A proposal written before the design exists produces a document that sounds like research and specifies nothing, which panels identify immediately.
The data, its access status, and what happens if access fails. Obtained, agreed, requested, or hoped for. Missing: this is not a gap to be papered over. Ask directly, and if the answer is that access has not been requested, that is the first thing the work plan must address and the proposal must say so.
The applicant's actual capability relative to what the design needs. Methods they have used, settings they know, data they have handled. Missing: ask, and write the qualifications section from the answer rather than from a CV. A panel is judging whether this person can do this project, not whether they are impressive.
The budget rules. What is eligible, what is capped, whether overheads are included, what documentation is required. Missing: fetch them; an ineligible line item can invalidate an otherwise fundable application.
Prior reviews, where this is a resubmission. Missing: ask. Resubmissions that do not visibly answer the previous panel's concerns are usually rejected for the same reasons.
The method
Read the call as a rubric and build the compliance checklist before writing anything. Every required section, every length limit, every attachment, every declaration, every formatting rule, every eligibility condition, with a tick box. This checklist is the spine of the whole exercise and it is what the final pass is run against. Proposals fail on missing sections more often than on weak science, and this step costs an hour.
Decide whether to apply, in writing, in three lines. Fit against the call's stated priorities, eligibility, and whether the work can be done in the period funded. Where the fit is poor, say so now. A proposal written to a call it does not fit consumes three weeks and scores in the bottom half regardless of quality.
Map every criterion to the section that will answer it, before drafting. Where the call's template does not have an obvious home for a criterion, decide where it goes and make sure the assessor can find it, using the call's own vocabulary once so the match is visible. This mapping is the difference between a proposal that scores what it deserves and one that does not.
Write the design sections first, from the design document: question, hypotheses, identification, methods, data. These are the load-bearing content and everything else is built around them. Writing the motivation first produces a motivation that the design then fails to match.
Argue feasibility with evidence at every point where it is claimed, using the rules below. This is the criterion that separates funded from unfunded applications among proposals of similar quality, and it is the one applicants most consistently treat as a formality.
Build the work plan backwards from the end date, with milestones that are observable events rather than activities. "Analysis complete" is not a milestone; "cleaned analysis file delivered, with the codebook" is. Allow realistically for data acquisition and cleaning, which take longer than analysis in the overwhelming majority of empirical projects and are budgeted at about half their true duration in the overwhelming majority of proposals.
Build the budget from the work plan, line by line, each tied to an activity, with unit costs and a reason. No round numbers without a basis. A budget that does not reconcile with the timeline is the fastest way to lose a panel's confidence in everything else, because it is the one section they can check arithmetically in two minutes.
Write the risk section honestly, with mitigations. Three to five risks, each with what would be observed first and what would be done. Panels are not deterred by acknowledged risk; they are deterred by a proposal that appears not to have noticed any.
Write the summary last, when the rest exists and can be summarised rather than promised.
Run the compliance checklist and the length check as a separate final pass, on the assembled document, against the call. Every box ticked, every limit measured rather than estimated. Where a section is over length, cut from motivation and literature, in that order, and never from design or feasibility.
Have it read by someone outside the specialism, to the criteria. The most useful reader is one who resembles the panel: competent, not expert in your subfield, working from the rubric. Where no such reader is available, read it yourself against the rubric a day later, scoring each criterion out of the published weight, and fix whatever scores poorly.
The proposal structure
Adapt the headings to the call's template; keep the content and the order within each section.
- Summary or abstract, to the length required: the question, the setting, the design, the expected contribution, and why this applicant. Written last, read first, and often the only part every panel member reads.
- Motivation and significance: the problem in the world, with numbers and sources, and why answering it matters to the funder's stated goals. Map explicitly to the call's priorities where they exist. Two to three paragraphs; panels know the problem exists.
- State of knowledge and the gap: the closest work, with citations verified to the
literature-verificationstandard, framed as what prior designs could not establish. Short. A long literature section in a proposal reads as an inability to prioritise. - Research questions and hypotheses: stated formally, each hypothesis falsifiable, each tied to the exhibit that would test it.
- Design and methods: the source of variation in words, then the estimating approach, then the identifying assumption, then the threats and the planned diagnostics, then the power calculation or minimum detectable effect where feasible. A reviewer should be able to predict the resulting paper's main table from this section. If they cannot, the section is not specific enough.
- Data and access: sources, years, units, sample size, the access route and its current status, the expected wait, and the fallback if access fails. Treat this as a load-bearing section, not an administrative one.
- Work plan and timeline: phases with milestones and dates, in a table or a simple bar chart. Where a chart is used, keep it monochrome, with phases separated by fill texture and dash pattern rather than by colour and the legend outside the plot area, since proposals are frequently printed in greyscale.
- Expected outputs and dissemination: papers with target venues, data or code to be released, briefings, presentations. Match these to what the funder actually rewards; a funder whose mission is policy influence scores a policy brief above a fourth journal article.
- Budget and justification, where required.
- Ethics, data management and risk: approval status and route, data protection arrangements for any microdata, a data management plan covering storage, retention and sharing, and the top three risks with mitigations.
- Applicant or team qualifications: two paragraphs on the specific capabilities this design needs, evidenced. Not a CV summary.
- References: verified, in the required style, and inside the word limit if references count toward it, which is a rule applicants routinely miss.
Feasibility, which is what actually gets scored
Feasibility is claimed everywhere and evidenced almost nowhere. Panels have learned this and read the feasibility signals rather than the feasibility claims. The signals are:
Data status stated precisely. "Held under agreement number X, expires 2029" scores. "Publicly available" scores if it is true and the URL is given. "Will be requested from the ministry" scores only if the request route, the typical wait, and the fallback are all named. "Access is anticipated" scores zero and taints the rest of the section.
Methods the applicant has used before. Where the design requires a method the applicant has not used, say so and say how that will be addressed: a collaborator, a course, a preliminary implementation already completed. Panels do not require that everything be familiar; they require that the gap be seen.
A timeline with slack. A plan in which every phase begins the day the previous one ends is a plan that has never survived contact with a data provider. Panels read zero slack as inexperience.
A preliminary result, where one exists. Even a descriptive figure on a subsample is worth more than any amount of confident prose, because it proves the data exists and the applicant can handle it.
A fallback for the main design. One paragraph: if the identifying variation turns out not to support the main design, what is the second-best defensible analysis, and is it still worth funding? A proposal with a fallback is a proposal by someone who has done this before.
Budget discipline
Every line ties to an activity in the work plan and carries a unit cost and a reason. Personnel effort in months, at a stated rate. Data acquisition costs with the provider's published price where one exists. Travel with the number of trips, the destinations and the purpose, tied to dissemination. Computing, transcription, translation, participant payments, open access charges, each with its basis.
Two rules that catch most problems. First, no round numbers without a basis: a data cost of 5,000 is read as a guess, and 4,780 with a source is read as a quote. Second, the budget and the timeline must reconcile: if the work plan has a research assistant for nine months, the budget has nine months of that assistant, and a panel member will check.
Where the call caps a category, respect the cap in the budget rather than in a footnote asking for an exception.
Fellowship and scholarship variants
Research statements for fellowships and doctoral applications are shorter, typically two to five pages, and the register is personal. The content that survives is the question, why it matters, why this applicant is positioned to answer it, and what the fellowship specifically makes possible that would not otherwise happen. The budget detail drops out unless asked. Programme-specific criteria, such as fit with a named host, contribution to a field the programme exists to support, or a mobility requirement, are answered explicitly and by name.
What does not change: the design must still be specific enough that a reader can see the study, and feasibility is still evidenced rather than asserted. The commonest failure in fellowship statements is a beautiful motivation attached to a project that could be any project.
Worked example
Situation. A postdoctoral researcher, Elena Marchetti, was preparing an application to a national research council's early-career scheme. The call funded three years, capped at 240,000, with a published rubric: significance 25 percent, design and methodology 30 percent, feasibility 25 percent, applicant 20 percent. Six pages of case for support, references excluded from the limit. Institutional sign-off required ten working days. Six weeks to the deadline. Her project used linked employer-employee records to study how firm-level training investment responds to minimum wage changes.
Task. A compliant, competitive application, submitted with the sign-off margin intact.
Action. The compliance checklist ran to 23 items and immediately produced two findings. References were excluded from the six pages, which meant about 400 words that had been mentally allocated to the limit were free. And the scheme required a data management plan as a separate two-page attachment that Elena had not known about, which was four hours of work she had not planned.
The criterion map showed the weighting was unusual: feasibility at 25 percent was equal to significance, and design plus feasibility was 55 percent of the score. The first draft's allocation was almost exactly inverted: two and a half pages on significance and the literature, two pages on design, and half a page on data and timeline. Rebalancing to roughly one page of significance, two and a half of design, one and a half of data, access and timeline, and one of qualifications was the single largest improvement made.
The data access section was the problem. The linked records were held by a national statistical institute, access required an accredited-researcher application, and Elena had described this in the first draft as "access will be sought at the start of the project". Under the feasibility rules that is a zero. Three things were done. She started the accreditation application immediately, so that by submission the status could be stated as "submitted, decision expected within 16 weeks, reference number given". She wrote the wait time into the work plan as an explicit sixteen-week phase during which the theoretical framework and the design of the survey component would be completed. And she wrote a fallback naming a public firm-level survey that could support a reduced version of the analysis, with a sentence on what would be lost.
The wrong turn. The first work plan was built forward from the start date and it fitted, comfortably, with three months to spare at the end. It was rebuilt backwards from the end date and the picture changed: the forward plan had assumed data access on day one, and once the sixteen-week accreditation wait was inserted, the analysis phase overran the funding period by seven weeks. The fix was not to compress the analysis, which would have been the visible temptation. It was to move the survey component from year two to the accreditation window in year one, where it used time that was otherwise idle, and to cut a planned third case comparison entirely. That cut cost one paragraph of the significance section and saved eleven weeks.
The budget was rebuilt from the revised plan. The original had a research assistant at 18 months chosen because it looked reasonable; the revised plan showed the assistant was needed for 11 months, concentrated in the data construction phase, which freed 31,000 that went to a second year of survey fieldwork the design actually needed. Every line was given a basis, including the data access fee, which was quoted from the institute's published schedule.
The risk section named three risks: accreditation refused, minimum wage reform postponed, and attrition in the survey component. Each with an early signal and a response.
Result. Submitted with six working days of sign-off margin. Scored in the funded range, with the panel's written feedback singling out the data access section and the fallback as the reason the feasibility score was high. The accreditation decision arrived before the funding start date, which meant the sixteen-week contingency was not needed and the project began ahead of its own plan.
Elena's assessment afterwards was that the rebalancing against the rubric weights was worth more than any improvement to the science, and that starting the accreditation application six weeks before submission was worth more than both.
A second scenario, where it goes differently
A doctoral thesis proposal for a departmental committee, twelve pages, no budget, no external funder, assessed by three faculty members who will supervise or examine the candidate later.
The rubric is different and it is largely unwritten. The committee is assessing whether the project is doable in the programme's timeframe by this student, whether the student understands what they are proposing, and whether it is a thesis rather than a paper. Significance matters less than at a funder, because the department already believes the field is worth studying.
Three things change. The scope question dominates: a proposal that is too large is the most common reason for a resubmission at this stage, and it should be pre-empted by the proposal itself naming what is in and what is explicitly out of scope. The fallback becomes central rather than reassuring, because the committee's real fear is a student in year four with a design that did not work. And the timeline is assessed against the programme's milestones rather than a funding period, which means it must name the upgrade, the fieldwork window, and the submission date.
The budget section disappears and is replaced by a resources section: what data, what software, what access, what training, what supervision the project requires, and whether the department can supply it. That last clause is the one students omit and committees care about.
The register also changes. A funder wants confidence; a thesis committee wants to see the candidate's judgement, including about what they do not yet know. A proposal that concedes two genuine uncertainties and says how they will be resolved reads better here than one that concedes none.
Output
The compliance checklist
| Requirement | Source in the call | Where answered | Done |
The criterion map
| Criterion | Weight | Section answering it | Evidence used |
The proposal, in the call's required format and within its limits.
The work plan
| Phase | Months | Activities | Milestone, as an observable deliverable | Depends on | Slack |
The budget
| Line | Activity it serves | Unit | Quantity | Unit cost | Total | Basis |
The risk register
| Risk | Early signal | Response | Effect on the plan if it occurs |
A one-paragraph plain summary for a non-specialist reader, which many calls require and which is also the fastest test of whether the project can be explained at all.
Failure modes
Writing before reading the call as a rubric. Recognise it when the draft has sections the call did not ask for and lacks one it did. Build the checklist first.
Effort allocated by the applicant's interest rather than by the rubric weights. Recognise it by measuring the pages against the weights. If feasibility is 25 percent and occupies half a page, rebalance.
Feasibility asserted. Recognise it in the phrases "will be obtained", "is anticipated", "should be straightforward". Replace each with a status, a route, a wait time and a fallback.
A timeline with no slack. Recognise it when every phase starts the day the previous one ends. Build backwards from the end date and insert the waits.
A budget that does not reconcile with the plan. Recognise it by checking every personnel month in the budget against the plan. Panels do this.
Round numbers. Recognise them on sight. Give each a basis or replace it with a quoted figure.
A literature section that reviews rather than argues. Recognise it when it summarises papers in sequence. It should name what prior designs could not establish and stop.
No fallback. Recognise it when the proposal has exactly one route to a result. Add the paragraph; it is the cheapest score available.
Length limits estimated rather than measured. Recognise it when nobody has counted. Measure against the call's own counting rule, including whether references and figures count.
A resubmission that does not answer the previous panel. Recognise it when the prior review is not in front of the writer. Address each previous concern explicitly, and say where.
Institutional sign-off discovered late. Recognise it when the internal deadline was never established. Find it in the first hour of work.
Edge cases
The call does not fit and the deadline is close. Say so and recommend against applying, with the reason. A poor-fit application costs three weeks and provides no useful feedback for the next attempt. The exception is where the funder gives written feedback that will improve the next application, which is worth something and should be stated as the reason for proceeding.
No access to the live call document. Work from what is available, mark every assumed requirement clearly in the checklist, and put a blocking note at the top of the deliverable that the call must be checked before submission. Do not silently guess a word limit.
The data does not yet exist and must be collected. The proposal's centre of gravity moves to the instrument, the sampling frame, the recruitment route, the expected response rate with a basis, and the power calculation under that response rate. Collection risk becomes the first risk in the register. survey-and-instrument-design supplies the instrument content.
A resubmission after rejection. Get the panel feedback, map every point to a change, and where the call permits a response letter or a cover note, use the response-to-reviewers discipline. Where it does not, make the changes visible in the text itself, because the same panel may read it again.
A collaborative application with several institutions. Establish the lead, the budget split and the sign-off chain in the first week, not the last. Multi-institution sign-off routinely takes three times as long as single-institution sign-off and is the commonest cause of a missed deadline in this category.
The applicant is not eligible. Check this before anything else and say so immediately if it fails. Eligibility is binary and no amount of quality compensates.
Ethics approval will not arrive before the funding start. Say so, state the submission date and the committee's typical turnaround, and design the first phase so it does not require approval. Panels accept this routinely when it is planned and penalise it when it is discovered.
Open access or data sharing mandates. Budget the charges and write the data sharing commitment so that it is one the project can actually keep, which usually requires knowing what replication-package will contain before promising it.
Quality bar
- Every required section is present, in the call's format, and within the measured length limit.
- Every published evaluation criterion is answered in a named section, and the pages are allocated roughly in proportion to the weights.
- Data access status is stated precisely, with the route, the expected wait, and a fallback if it fails.
- The design section is specific enough that a reviewer could predict the main table of the resulting paper.
- The work plan's milestones are observable deliverables, and the plan contains visible slack.
- Every budget line ties to an activity in the work plan and carries a unit cost and a basis.
- Three to five risks are named with an early signal and a response.
- Every citation is verified, and the reference style matches the call.
Adapting this to your context
The worked example is a national research council early-career scheme in economics with a published four-criterion rubric. Funders differ more than fields do.
- The rubric. That split belongs to one council. NIH scores significance, investigators, innovation, approach and environment, with approach carrying most of the weight. NSF has two criteria and broader impacts is a real one. ERC weights the principal investigator as heavily as the project. Rebalance the pages against yours.
- Power, not minimum detectable effect. The design section mentions power in passing, an economics habit. Health, psychology and education panels expect a formal calculation with the assumed effect size and its source, alpha, attrition, and for clustered designs the intraclass correlation and design effect.
- What counts as preliminary data. Here, a descriptive figure on a subsample. In lab and clinical work, pilot data with an observed effect size; in qualitative work, negotiated access and two or three pilot interviews.
- Registration commitments. Where your field expects prospective registration or a reporting guideline, name both in the outputs section. Panels in health and psychology read their absence as a weakness.
- What not to change. Build the compliance checklist from the call before writing a word, and evidence feasibility with a status, a route, a wait time and a fallback rather than an adjective.
Related skills
research-design produces the question, hypotheses, identification strategy and exhibit list that this proposal argues for; build it first. identification-defense supplies the threats and diagnostics paragraph and the fallback design. literature-verification verifies the citations, and a fabricated reference in a funding application is a far more serious matter than in a draft. survey-and-instrument-design supplies the instrument where data must be collected. research-ethics-and-data-protection supplies the ethics and data management sections. preregistration-and-analysis-plan overlaps heavily with the design section and should be written consistently with it. replication-package defines what a data sharing commitment can honestly promise. thesis-advisor decides whether the applicant has the time this proposal assumes. thesis-defense-prep prepares the proposal hearing or the panel interview. response-to-reviewers supplies the discipline for a resubmission after panel feedback.