Framing the question: half the result

Deep research amplifies your question. A weak question gives you amplified waffle. So let's build a strong one.

The research-request formula

DECISION: whether to open a café in district X of city Y in 2026.
CONTEXT: budget up to 40k €, no food-service experience, but I've
run retail; I'm looking at a 45 m² unit.
SUB-QUESTIONS:
1. How many cafés are in the district, and what's the footfall?
2. Typical economics of a café this size: revenue, rent,
   time to break even?
3. Main reasons cafés close in year one?
4. What do owners themselves say (forums, interviews)?
SOURCES: prioritise recent data, local sources and owner
experience; don't lean on franchise marketing articles.
FORMAT: summary → sub-question by sub-question with links →
risk table → what to check in person.

Why each block earns its place

• Decision — it focuses: the model keeps what changes your choice instead of "everything about coffee".
• Context — it cuts the irrelevant (advice written for chains doesn't apply to you).
• Sub-questions — your plan instead of a random one: the model looks for what you actually need.
• Source requirements — they filter out marketing junk before it reaches you.
• Format — the report arrives ready to work with.

A trick: have the model improve your question

Before you launch: "Here's my research request. Which sub-questions am I missing? What should I pin down to make the result more useful?" — a minute of back-and-forth saves you a whole rerun.

Practice

Build a request with the formula for a real decision of yours (a purchase, a service, a job move). Save it — we'll run and check it in the lessons ahead.