How to Write a Policy Research Commissioning Brief
Define a policy research question, decision, scope and deliverables before commissioning work, using a practical briefing worksheet.
A policy research commissioning brief is a short written statement that fixes four things before work begins: the decision the research is meant to inform, the question the research must answer, the boundary of what is and is not included, and the tests by which the finished work will be accepted. It is written to constrain the process and the standard of evidence, not to steer the conclusion. A useful brief can be read by a decision-maker who will never see the underlying research, by a researcher who has not worked in the commissioning organisation before, and by whoever signs off the invoice, and each of them should come away with the same understanding of what is being bought.
Begin with the decision, not the topic
A subject label such as "school readiness", "skills shortages" or "regulatory burden" does not identify the decision the research must inform. A researcher receiving that can produce something defensible that answers no one's question.
State the decision, the decision owner and the point at which the answer is needed. A single sentence is enough: which body will decide what, by when, and what would change if the evidence pointed one way rather than another. If nothing would change, the research is probably documentation rather than evidence, and should be commissioned as such.
Then name the users. Distinguish the person who takes the decision from those who will act on the findings — a board, a programme director, a funder, a group of practitioners. Their prior knowledge matters: a brief that assumes familiarity with programme data will produce a report that a new board member cannot use.
Set the boundary: scope, exclusions and evidence access
For this worksheet, describe scope through subject boundaries, exclusions and evidence access.
Subject boundary: which populations, places, time periods and policy instruments are in scope. Naming an out-of-scope comparator or time period prevents a literature review that is broad but useless.
Exclusions: say explicitly what the research should not do. Common exclusions are primary data collection, individual casework, legal advice, cost modelling and recommendations on budget allocation. Exclusions protect both sides: they stop a supplier from pricing work the commissioner did not want, and they stop the commissioner from expecting it.
Evidence access: list the material the researcher will be given, such as internal records, monitoring data, named contacts and published statistics, and its condition. State whether internal data is complete, how it is held, what cannot be shared, and which permissions the commissioner will obtain. Where access is uncertain, say so rather than implying availability.
Where the work involves external researchers, the UK Government Office for Science's Academic Engagement Toolkit advises that written arrangements clarify purpose, outputs, time, roles, attribution and information handling, and that payment and expenses be discussed early. That guidance is written for UK government engagement with academics and does not set terms for any particular network; its principle, that roles and information handling are settled in writing before work starts, applies to any commissioning relationship. Confidentiality and data protection obligations should be proportionate to the material involved and verified against the commissioner's own policies.
Define the deliverable without fixing the conclusion
The brief should specify format, length range, structure and audience for the output, and should stop short of specifying what the output should say. A brief that asks for "evidence supporting option B" will receive it, and the commissioner will have learned nothing they did not already believe.
Useful specification covers:
- the deliverable or deliverables, and whether a presentation, evidence table or data appendix is included;
- the structure expected — for example, question, method, findings, limitations, implications;
- how sources are to be attributed, so that each material claim can be traced;
- how uncertainty and disagreement in the evidence are to be reported;
- what happens to underlying data and notes at the end of the project.
Method should be discussed, not dictated. State the standard the method must meet: fit to the question, proportionate to the timetable, transparent about sampling and analysis. Then invite the supplier to propose an approach. Prescribing a method in advance removes the supplier's ability to tell you that a different design would answer the question better.
Acceptance checks that test rigour, not agreement
The hardest part of a brief is defining acceptance without making approval conditional on a congenial answer. The practical solution is to write acceptance checks that are procedural and can be assessed before the findings are read.
Checks that work include: every material claim attributed to a named source; counter-evidence and competing interpretations addressed rather than omitted; the limits of the evidence stated in the same document as the findings; the boundary between what the evidence shows and what the commissioner might do about it kept visible; and any departure from the agreed question or method explained. A report that fails these tests fails regardless of its conclusion. Assess acceptance against the agreed requirements, including the stated limits, rather than a preferred conclusion. Put that principle in the brief before work begins.
Add a practical acceptance step: a short review of a draft structure or interim note, at which scope problems can be corrected before the final document exists. This is a check on fit, not on findings.
A worksheet for drafting the brief
| Field | What to settle | Failure to avoid |
|---|---|---|
| Decision | The decision, its owner, the date it is taken, and what evidence would change | A topic instead of a decision |
| Question | The primary question and any necessary secondary questions, phrased so that evidence could answer them | Questions that presuppose the answer |
| Audience and users | Who decides, who acts, what they already know | Assuming shared background knowledge |
| Exclusions | What is out of scope: populations, places, methods, advice | Silent exclusions discovered late |
| Evidence access | Material supplied, its condition, permissions, restrictions | Implying access that is not arranged |
| Output | Deliverables, structure, attribution, treatment of data at close | Format specified, standard unstated |
| Acceptance checks | Traceability, counter-evidence, stated limits, separation of findings from recommendations, response to review | Approval contingent on a preferred conclusion |
| Unresolved assumptions | What the brief assumes and has not verified, and who will check it | Assumptions hidden inside the scope |
Fictional example: one filled worksheet
The following is a fictional illustration, not a description of any client or project. A small regional nonprofit is deciding whether to extend a school-based mentoring programme into a second district.
| Field | Fictional entry |
|---|---|
| Decision | The board decides in March whether to approve a second-district pilot for the 2027–28 school year; a finding that comparable districts could not sustain staffing would weigh against extension. |
| Question | What does available evidence indicate about the conditions under which school-based mentoring programmes of this design have been sustained in districts with similar staffing and transport constraints, and what does that evidence not establish? |
| Audience and users | Board (decision), programme director (delivery), funder liaison (reporting). None has read the published evaluations. |
| Exclusions | No primary data collection involving pupils; no assessment of individual staff; no recommendation on budget allocation; no legal advice on contracting. |
| Evidence access | Three years of anonymised internal programme records; published evaluations identified jointly; two named local authority data contacts, subject to their agreement. No pupil-level records. |
| Output | Report of no more than 15 pages plus a 2-page summary; evidence table giving source, finding and limitation per row; short methods note. |
| Acceptance checks | Each material claim attributed; counter-evidence and non-transferable conditions addressed; limits stated alongside findings; options presented with consequences but no single recommendation; departures from the brief explained. |
| Unresolved assumptions | That comparable districts publish usable data; that internal records are complete for three years. The programme director will verify both before the brief is issued. |
The example is deliberately unglamorous. Its value lies in the exclusions and the acceptance checks, which do the work that scope statements alone cannot.
Trade-offs and limits
A brief long enough to remove all ambiguity becomes a design document that forecloses the supplier's judgement. A brief short enough to invite proposals leaves room for misunderstanding about exclusions and access. The worksheet above sits between the two: detailed on boundary and acceptance, thin on method.
Openness about the question trades against accountability for the answer. The more precisely the question is phrased, the easier it is to accept or reject the work, and the easier it is to encode an assumption. Reading the question aloud and asking "could evidence contradict this?" is a cheap test.
Specifying evidence access trades against realism. Promising internal data that turns out to be incomplete damages the timetable and the relationship; naming the assumption and its owner is more honest than asserting availability.
Finally, a brief cannot guarantee quality. It can make the standard explicit, make exclusions visible and make acceptance independent of the conclusion, but the judgement of whether a piece of research is any good still rests with the commissioner. Budget time for that reading, and identify in advance who will do it.
Questions that recur
Should the brief name a preferred method?
Name the standard the method must meet and any methods that are unavailable for practical or ethical reasons. Leave the design to the supplier, and ask them to justify it against the question.
How long should a brief be?
Long enough to cover the eight fields above without repetition. Length is a poor proxy for clarity; check whether the reader can identify the decision, the exclusions and the evidence standard from the brief.
What if the evidence turns out to be too thin to answer the question?
Say in the brief that a documented "insufficient evidence" finding is an acceptable outcome, provided the search and its limits are described. Without that permission, suppliers are pushed towards overstating what they found.
Does a brief replace a contract?
No. It sets the intellectual and evidential standard; commercial terms, payment, liability and the handling of data belong in the agreement, which should be consistent with the brief.
A practical next step is to draft the eight fields on one page and give it to the person who owns the decision, asking them to mark anything they would not be able to defend to their board. Revise once, then issue. Researchers considering project-based policy work, and commissioners checking how an expert network describes its own scope, can review Policy Expert Network's about page for the company's operating information.