Product Brief

A brief exists to get people aligned before the work starts — what we're building and why (PRD), or how we propose to build it (RFC). Its job is to surface disagreement early while it's cheap, give everyone one source of truth, and produce a reviewed decision engineering can execute against.

A brief is not a novel. A two-page PRD with clear metrics beats a forty-page doc nobody reads. A RFC with honest alternatives beats a design that pretends the chosen path is the only path.

Define phase chain: [[idea-shaping]] (is it worth it?) → product-brief (align stakeholders) → [[spec-first]] (testable engineering spec) → [[work-planning]] (tasks). Record settled decisions in [[decision-docs]].

When to Use

Product brief / PRD — the what and why:

• Feature or initiative needs sign-off across product, engineering, design, support, legal
• Multiple stakeholders own success metrics or scope boundaries
• Estimation or staffing needs a shared scope document
• Executive or customer commitment requires written alignment

Technical RFC — the how:

• Significant architecture, platform, or migration choice before coding
• Cross-team technical contract (API, data model, infra) needs review
• Trade-offs affect performance, security, cost, or operability for years
• Alternatives exist and reviewers must trust the pick

Both — large initiatives: PRD for user outcomes + RFC for technical approach (can be one doc with clear sections).

Skip when [[idea-shaping]] hasn't confirmed pursue — don't PRD a idea still fuzzy. Skip RFC for small changes with one obvious approach inside one team — go to [[spec-first]].

Process

1. Choose the artifact — PRD, RFC, or both

Name the author, reviewers, and decision date at the top — accountability matters.

2. Write the product brief / PRD

Ground in evidence from [[idea-shaping]] when available — don't restart from a solution slogan.

Template:

# [Feature name] — Product Brief
Status: Draft | In review | Approved
Author: … | Reviewers: … | Target release: …

## Problem & audience
Who has the pain? What do they do today? Evidence (data, quotes, support volume) — not assertion.

## Why now
Strategic fit, market, regulation, competitive pressure, technical enabler.

## Goals & success metrics
Primary metric + target (e.g. publish rate +10% in 90 days).
Guardrail metrics (don't harm conversion, latency, support volume).

## Scope — in
Must-have behaviors for this release (bullet list).

## Scope — out
Explicitly deferred (prevents scope creep in review).

## User stories / requirements (prioritized)
Must / should / could — each testable:
- As a seller, I see a suggested price range when listing, so I can publish confidently.

## Dependencies
Teams, APIs, legal, design, data, third parties.

## Risks & mitigations
| Risk | Mitigation |
|------|------------|

## Open questions
| Question | Owner | Due |
|----------|-------|-----|

## Non-goals
What we are deliberately not solving.

Requirements must be testable — "fast checkout" fails; "guest checkout completes in < 3 steps" passes (hand detailed criteria to [[spec-first]]).

Success metrics need baseline + target + measurement method ([[observability]] hooks planned).

3. Write the technical RFC

For how we'll build after (or alongside) product alignment:

# RFC: [Title]
Status: Draft | In review | Accepted | Superseded
Author: … | Reviewers: …

## Context & problem
Current system, constraints, scale, why change now. Link PRD if user-driven.

## Goals & non-goals
Technical outcomes (latency, reliability, cost) — not duplicate entire PRD.

## Proposed design
Architecture diagram (simple), components, data flow, key interfaces ([[interface-design]]).
Migration impact if any ([[migration-path]]).

## Alternatives considered
| Option | Pros | Cons | Why not |
|--------|------|------|---------|
| A — chosen | … | … | |
| B | … | … | rejected because … |
| C — do nothing | … | … | |

## Trade-offs & impact
Performance, security ([[hardening]]), operability, cost, team ownership, timeline.

## Rollout plan
Phases, flags, backward compatibility ([[launch-readiness]], [[incremental-delivery]]).

## Open questions
| Question | Owner | Due |

## Appendix
Diagrams, spike results, benchmarks ([[perf-budget]]).

Alternatives section is mandatory — reviewers trust RFCs that show roads not taken ([[decision-docs]]).

4. Circulate for review — structured feedback

Before "Approved":

1. Share early draft — async comment window (48–72h minimum for cross-team)
2. List reviewers explicitly — product, eng lead, security if needed, support, legal for sensitive
3. Review meeting only if async left blocking disagreements — with pre-read required
4. Capture decisions in doc — don't rely on meeting memory
5. Resolve open questions — answer, defer with owner/date, or descope

Feedback types:

Approval means: stakeholders accept scope, metrics, and approach — not "everyone loves it."

5. Gate implementation — no build on open blockers

Do not start [[work-planning]] / [[incremental-delivery]] while:

• Open questions block core flow or success metric
• Security/legal review outstanding for regulated features
• RFC alternatives unresolved — "we'll pick in sprint 1"

Allowed parallel work: spikes time-boxed to answer open questions ([[idea-shaping]] validation), design exploration — not production feature code on contested scope.

6. Hand off to engineering artifacts

Spec-first is narrower than PRD — engineering contract for one team sprint, not stakeholder alignment.

Update brief status to Approved with date; link spec/ADR in header.

7. Keep the brief alive — or supersede

Briefs are living during Define/Plan, frozen at Approved for that release slice.

When scope or approach changes materially:

• Update brief with changelog section, or
• Supersede with new brief/RFC referencing old ([[decision-docs]] immutability pattern)

Post-ship retrospective: did metrics move? Feed [[idea-shaping]] for v2.

8. Scenario playbooks

New user-facing feature

Idea-shaping → PRD draft → design review → eng review → approved → spec-first → build.

Platform migration (no UX)

RFC with alternatives → spike for risky unknown → approved → migration-path plan → incremental delivery.

Executive priority drop

Shape problem in 1-pager → PRD scope minimal v1 → explicit out-of-scope → fast review with named approver.

Cross-team API initiative

RFC interfaces first → partner teams comment → contract frozen → PRDs per consumer if needed.

Customer-driven one-off

Evidence it's representative segment → PRD or explicit "custom exception" with sunset → decision-doc.

PRD without eng review

Risk: unbuildable scope — always include eng reviewer for feasibility and dependencies.

RFC without product context

Risk: solves technical elegance not user outcome — link PRD or problem statement.

Brief after code started

Retrospective doc only — mark as retrospective; don't call it alignment.

Common Rationalizations

• "Everyone already agrees." — Writing it down discovers they don't.
• "A doc slows us down." — Less than building wrong thing or reversing in sprint 4.
• "We'll align as we go." — Mid-build disagreement is the most expensive kind.
• "Skip alternatives in RFC." — Reviewers can't trust the choice without them.
• "PRD is PM only." — Eng, design, support input prevents surprise scope.
• "Open questions we'll solve in implementation." — Blockers stay blockers; name owners now.
• "Success metric is obvious." — Obvious metrics get measured differently by each team.
• "RFC is too technical for a brief." — That's why RFC exists separately from PRD.

Red Flags

• Cross-cutting work with no shared written plan
• PRD lists solutions but vague problem or metric
• RFC single option, no alternatives or trade-offs
• Open questions open when coding starts
• Brief written after ship to rationalize
• Success metrics without baseline or measurement plan
• No named approver — "the team" approved
• Scope out section empty — creep guaranteed
• RFC ignores migration, security, or rollback
• PRD requirements not testable ("intuitive", "modern")
• Stakeholder never read doc — approval theater
• Brief and spec contradict — no single source of truth

Verification

• Right artifact chosen — PRD, RFC, or both; idea shaped if needed ([[idea-shaping]])
• Problem, audience, evidence — not solution-only narrative
• Goals with measurable success + guardrails; scope in/out explicit
• Requirements prioritized and testable (PRD)
• Design, alternatives, trade-offs, rollout documented (RFC)
• Dependencies, risks, open questions with owners and dates
• Structured review completed; approval recorded with date
• No implementation on unresolved blockers
• Handoff to [[spec-first]] / [[work-planning]] / [[decision-docs]] clear
• Brief status updated; supersede path if approach changes