01 / START HERE
A useful way to approach it.
Market research prompts work best when they name a customer, a decision and an evidence boundary. A request for a complete market report invites a long answer that can conceal weak assumptions. Use a sequence of bounded requests to frame the question, extract evidence, examine contrary explanations and decide what to validate next.
The prompts below are for desk research and analysis of material you supply or can access publicly. They do not create survey results, proprietary datasets or verified demand. Replace the bracketed inputs and ask for unknowns whenever the evidence cannot support a conclusion. Use a separate customer test for questions that public sources cannot answer.
02 / THE PROCESS
Go from question to next step.
- 01
Frame the customer question
Name a specific segment, geography, problem and buying situation. Ask which questions desk research could answer and which require interviews or a test. Avoid treating a broad trend as evidence that this customer will buy your offer.
- 02
Constrain the source set
Provide official statistics, published reports or accessible source excerpts. Specify the period and geography. If the assistant has no browsing tool, tell it to analyze only supplied material and return a source-request list for anything missing.
- 03
Extract before summarizing
Ask for source, date, claim, measurement definition, audience and caveat in separate columns. A number without its denominator or method is not ready to use. Keep estimates and directly observed measurements distinct.
- 04
Compare competing explanations
Request evidence that both supports and weakens the proposed opportunity. Several articles citing the same study are one evidence chain, not independent confirmation. Explain discrepancies before combining figures or trends.
- 05
Design a small validation step
Turn the largest remaining unknown into a practical next action with an owner and a decision rule. Examples include a consented interview or a limited offer test. Do not write a positive customer outcome before the test happens.
03 / YOUR STARTING POINT
A prompt you can make your own.
Replace the bracketed fields. Add only information you have permission to share. This is a reusable text prompt; it does not run a task from this page.
Use this sequence for [business question] about [customer segment] in [country] during [period]. 1. Framing: state the decision, alternatives and three answerable questions. Separate desk-research questions from those requiring primary validation. 2. Extraction: use only [public source URLs or pasted material]. Build claim | source | date | population | method | caveat. Do not invent statistics or citations. 3. Synthesis: explain agreement, disagreement and shared source origins. Separate supported facts, estimates and assumptions. 4. Challenge: identify the strongest evidence against [hypothesis] and explain missing information. 5. Next test: propose one bounded validation step with an owner, observable result and decision rule. Do not claim the test ran. Ask for inputs where the evidence is insufficient.
04 / SEE THE SHAPE OF IT
Example prompt sequence: a neighborhood meal-planning service
Fictional idea and authored research plan. No market size, customer demand or conversion results have been measured.
| Stage | Useful request | Boundary |
|---|---|---|
| Frame | Which households and occasions does this offer serve? | Define the segment before estimating anything |
| Extract | What does each supplied public source actually measure? | Do not substitute national totals for local buyers |
| Challenge | What alternatives already solve the same problem? | A common problem does not prove willingness to pay |
| Validate | What question should a consented customer interview address? | Interviews are a proposed next action, not completed evidence |
A useful result may be a narrower question and a list of missing evidence. That is more actionable than a confident market-size number built from unrelated assumptions.
05 / KEEP YOUR JUDGMENT
What to watch before you use it.
Combining incompatible figures
Different dates, populations and definitions can make figures incomparable. Preserve the source context and explain the mismatch.
Sources requested after the answer
Asking for citations at the end can encourage weak attribution. Make the evidence table part of the initial output contract.
Research confused with demand
Published discussion of a problem can guide a hypothesis. A buying test is still needed to establish whether your specific offer attracts customers.
06 / THE FINAL PASS
A little review goes a long way.
Your checks stay on this page. They do not approve or run a task.
07 / GOOD QUESTIONS
Before you begin.
Can these prompts produce a market size?
They can help structure a source-based estimate when suitable inputs exist. Any estimate needs explicit assumptions, comparable definitions and uncertainty; the prompts do not guarantee an exact market size.
Should I paste an entire confidential report?
Share only information you have permission to use and actually need for the question. A brief, non-confidential excerpt or public source may be enough.
What if the assistant cannot open a source?
Provide an authorized excerpt or ask it to record the source as unavailable. An inaccessible reference should not be presented as evidence that was inspected.