Skip to content
promptbeginnerFree

Research brief builder

Turn a vague topic into a scoped research plan with ranked questions, a source plan and a stop rule before anyone starts searching.

You'll need: A decision the research is meant to inform

Fill in your details

Values are inserted into every prompt below. Nothing you type leaves your browser.

5 field(s) left — unfilled variables stay as {{placeholders}}.

  1. 1Fill in the brief inputs and run the prompt

    Be specific about the decision. 'Learn about heat pumps' produces a reading list; 'Decide whether to offer heat-pump installs in Q2' produces a research plan. Paste what you already believe even if you are unsure; the prompt turns those beliefs into assumptions to test.

    You are a research lead writing a brief before any searching starts. Turn the topic below into a scoped research plan that a colleague could carry out without asking you questions.
    
    Topic: {{topic}}
    Decision this research must inform: {{decision}}
    Who will read the findings: {{audience}}
    What we already know or believe: {{known}}
    Constraints (deadline, budget, geography, sources to avoid): {{constraints}}
    
    Write the brief with these sections:
    1. Decision statement: one sentence in the form "We need to decide X by Y; this research reduces uncertainty about Z."
    2. Key questions: 5 to 8 questions, ordered by how much the answer would change the decision. Label each Fact (verifiable), Estimate (needs a proxy or model) or Judgment (needs expert opinion).
    3. Assumptions to test: restate each belief from "what we already know" and say what evidence would disprove it.
    4. Source plan: for each question, the 2 or 3 most credible source types (for example regulator filings, peer-reviewed reviews, vendor documentation, practitioner interviews) and one source type to treat with caution, with the reason.
    5. Out of scope: what we will deliberately not research.
    6. Deliverable: format, length and how confidence should be labelled for this audience.
    7. Stop rule: the condition under which we have enough to decide.
    
    Rules:
    - Do not answer the research questions yourself.
    - Do not name specific reports, studies or statistics unless you are certain they exist; describe source types instead.
    - If the decision is ambiguous, ask up to three clarifying questions first and wait for answers.
  2. 2Pressure-test the plan

    Ask the model to attack its own brief. Remove any question that would not change the decision, then share the brief with whoever will do the research.

    Review the brief you just wrote as a sceptical decision-maker. For each key question, state in one line what we would do differently if the answer came back "yes" versus "no". Flag any question where the answer would not change the decision and recommend cutting it. Then list the single biggest risk that this plan misses.

Example input

Topic: offering heat-pump installation to existing boiler customers. Decision: launch a heat-pump service in Q2 or not. Audience: co-founders and ops manager. Known: demand rising, installers scarce, grants exist. Constraints: two weeks, no paid reports, UK only.

Example output

1. Decision statement We need to decide whether to launch a heat-pump install service by 31 January; this research reduces uncertainty about demand among our existing customers, installer capacity and margin after grants. 2. Key questions Q1 (Estimate) What share of our ~4,000 boiler customers would consider a heat pump in the next 12 months? Q2 (Fact) What certification do we need to install and to let customers claim the current grant? Q3 (Estimate) What is a realistic gross margin per install after grant timing and labour? ... 7. Stop rule Stop when Q1-Q3 each have at least two independent sources and the margin range is narrow enough that the decision is the same at both ends.

Troubleshooting

  • The model starts answering the questions: repeat the rule 'Do not answer the research questions yourself' and ask it to rewrite only sections 2-7.
  • Questions are generic ('What is the market size?'): make the decision field more specific, including a date and the option you are leaning towards.
  • It cites a named report you cannot find: treat it as a hallucination and ask for the source type instead.

Known limitations

  • The brief is only as current as the model's knowledge; it plans research but does not verify facts.
  • Source-type suggestions are generic and need a subject-matter check in regulated fields (health, finance, law).

More from ChatGPT / Claude / Gemini