Knowledge card L3 · Workflows informational

Scope the Question Before You Run Deep Research

What it is

A discipline for Deep Research: define the exact question, the output shape, and the boundaries before you kick off a run that will take many minutes and read dozens of sources.

scope before you runDefine scopeSet boundariesRun researchReview sources
A tight scope up front is what makes a long research run pay off.

Why it works

Deep Research is expensive in time and hard to steer once running. A vague brief ('research electric cars') produces a sprawling report you then have to re-scope by hand. A precise brief ('compare total 5-year cost of ownership for three named EVs for a UK commuter') returns something decision-ready on the first pass.

When to use it

Questions that genuinely need breadth and citations: market comparisons, literature scans, vendor evaluations, due diligence. Anything where you'd otherwise open twenty tabs.

When not to use it

Quick factual lookups or anything a normal chat answers in seconds. Deep Research on a simple question just makes you wait longer for the same answer.

Prompt

Deep Research brief:
- Question: <the single decision you're trying to make>
- Output: <table / ranked list / memo>, <length>
- Must cover: <the 3–5 dimensions that matter>
- Boundaries: <geography, time range, sources to prefer or avoid>
- Audience: <who reads this and what they'll do with it>

Example

Instead of 'research CRM tools', a precise brief — 'compare HubSpot, Pipedrive, and Attio for a 5-person B2B team on price, API quality, and migration effort; output a table plus a one-line recommendation' — returns a report you can act on without re-scoping.

Advanced version

Ask for the source list and confidence up front: add 'flag any claim you couldn't corroborate across two independent sources.' That turns the report from a wall of assertions into something you can trust selectively.

Common mistakes

  • Firing off a broad topic and getting a long, unfocused survey you have to redo.
  • Not specifying the output shape, so you get prose when you needed a comparison table.
  • Using Deep Research for questions a normal chat answers instantly — paying minutes for seconds of value.

Related