Product Manager Brainstorm Phase

ProductManager BrainstormPhase

BrainstormPhase converts evidence into ranked feature candidates. It should feel imaginative, but never detached from the research base.

Inputs

Read these if present:

• research.md
• evidence-log.md
• Existing roadmap, issues, docs, or user notes
• Product constraints from AGENTS.md, architecture docs, or runtime evidence

If the evidence base is thin, list gaps and run a small ResearchPhase top-up before scoring.

Candidate Generation

Generate candidates through four lenses:

1. Friction removal: what repeated pain did mock use reveal?
2. Visibility: what valuable state already exists but is hard to see, compare, or trust?
3. Control: where does the user need a safer, faster, or more reversible action?
4. Category expectation: what do adjacent products make users expect?

For each candidate, write the smallest useful version first. Add larger variants only after the small version is clear.

Scoring

Use a 1-5 scale unless the user gives a different rubric:

Recommended score:

(user_value * 2) + evidence_strength + differentiation - effort - risk

Candidate Matrix

Write candidate-matrix.md:

# BrainstormPhase Candidate Matrix

| Rank | Candidate | User Problem | Evidence | Value | Evidence | Diff | Effort | Risk | Score | Call |
|---|---|---|---|---:|---:|---:|---:|---:|---:|---|

## Top Recommendation

## Parked Ideas

## Rejected Ideas

## Assumptions to Validate

Quality Gate

Before moving to FeatureDocPhase:

• Rank at least five candidates unless the product surface is tiny.
• Include at least one "do not build" call.
• Name the top candidate and why it wins.
• Identify the riskiest assumption for the top candidate.

Out of Scope

• Gathering initial evidence: use product-manager-research-phase.
• Writing detailed feature docs: use product-manager-feature-doc-phase.
• Sending or summarizing the run: use product-manager-report-phase.