Best for
- Features that start building before anyone agrees what they are
- The 'wait, I thought we were building X' meeting in week three
- Giving builders (human or AI) a target instead of a vibe
What you give it
- The feature idea and its context: the evidence, the requests, the strategy it serves
What you get back
- The brief: problem (evidenced), audience (specific), proposed approach (one paragraph, not a spec), success metrics (measurable), non-goals (explicit)
- The disagreements surfaced now: the brief's draft circulated forces the 'actually I assumed…' conversations into week zero
- The decision hooks: open questions named with owners, so the brief unblocks building instead of becoming another document
How it works
- Forces the problem before the solution: what is broken, for whom, evidenced by what — a feature without an evidenced problem is a solution shopping for one.
- Keeps the approach at brief altitude: one paragraph of direction, not a spec — the brief aligns intent; design and spec come after alignment.
- Makes success measurable and dated: the metric, the threshold, the when — 'users love it' is not a success criterion.
- Writes the non-goals loudly: what this deliberately is not — the section that prevents the scope creep and the week-three surprise.
Example
You: Write the brief for 'add commenting to reports' before the team starts.
Result: The brief's drafting surfaced the week-three fight early: the problem evidence (14 support requests, 2 churned accounts citing it) pointed at async review between colleagues — but half the team assumed real-time collaboration (a different, much bigger feature). Decided on paper: async commenting, threaded, email-notified; real-time explicitly in non-goals with its reasoning. Success defined measurably (comments used weekly by 20% of multi-user accounts within a quarter; the report-sharing support theme shrinking), two open questions assigned (permissions model — product; notification batching — design), and the team built the right smaller thing in three weeks instead of the wrong bigger thing in ten.
Limits — please read
- A brief aligns; it does not design — the spec and the wireframes follow it, not replace it.
- Evidence thinner than conviction gets said: 'hypothesis, validated by…' beats confident fiction.
- One page is the discipline; a brief that needs five pages is usually two features wearing one name — it says so.