Product grooming

Product Grooming

Sprint planning fails when stories are vague. Grooming is the ceremony that makes backlog items dev-ready — user story, testable acceptance criteria, rough estimate, dependencies, and a Ready / Not ready verdict — before anyone commits capacity.

This skill is the process behind the product-grooming agent. It synthesizes team voices; it does not replace product-manager, architects, or QA.

Pairs with [[spec-first]] for AC quality, [[work-planning]] for estimates and dependencies, [[idea-shaping]] for still-vague epics, [[product-brief]] when stakeholder alignment is missing, and [[interface-design]] when API/schema boundaries need early agreement.

When to Use

• Backlog refinement session before the next sprint
• Epic too large — needs splitting into sprint-sized stories
• Stories lack acceptance criteria or estimates
• Same items keep reappearing ungroomed sprint after sprint
• Cross-team alignment needed (product, eng, arch, QA)

Skip when items are already sprint-ready with AC, estimates, and no blocking questions.

Process

1. Set the session frame

• Goal: Which sprint is this grooming for? (usually next)
• Timebox: 60–90 minutes; 5–8 stories max
• Inputs: Current backlog, PRD/epics, carry-over from last grooming

2. Per backlog item — walk team voices

For each epic or story, answer these questions (synthesize; do not guess):

Voice	Question
Product	Why this? Priority? In/out of scope this sprint?
Business	Business rules clear? Domain edge cases?
Architecture	Feasible? Boundaries? Major dependencies?
Team Lead	Task count? Rough dev days? Blockers?
Eng Manager	Fits capacity? Calendar realistic?
QA	AC testable? Happy + unhappy paths?
Security	PII, authz, abuse? Early red flags?

Open questions get an owner and due date — never assume an answer.

3. Clarify and split ([[spec-first]])

• Rewrite: As a … I want … so that …
• Add testable acceptance criteria per story
• Split if too large — must fit one sprint slice
• Mark out of scope explicitly

4. Estimate and sequence ([[work-planning]])

• Rough estimate (days or points) with stated assumptions
• Dependencies: schema → API → UI, flags, migrations
• Flag spikes before commitment if unknowns remain

5. Definition of Ready gate

Ready only when all true:

• User story + testable AC
• No blocking open questions (or spike scheduled)
• Rough estimate with assumptions
• Dependencies identified
• Priority confirmed
• QA confirms AC are testable

Mark each item Ready or Not ready with what's missing.

6. Close the session

• Prioritized list for sprint planning
• Open questions with owners
• Deferred items with reason
• Hand off to scrum-master and engineering-manager

Use grooming checklist alongside this process.

Common Rationalizations

• "We'll figure it out in the sprint" — Ungroomed stories become mid-sprint surprises.
• "Only product needs to attend" — Engineering and QA discover impossibilities late.
• "Estimates can wait" — Sprint planning without estimates is guessing.
• "One big epic is fine" — Split in grooming, not on day 8.

Red Flags

• Stories enter sprint planning without AC
• Open questions have no owner
• Estimates without assumptions
• Grooming becomes deep solution design — defer LLD to architects
• Same stories groomed repeatedly with no progress toward Ready

Verification

• Every discussed item has Ready / Not ready verdict
• Ready items have story, AC, estimate, dependencies, priority
• Not ready items list what's missing and who owns it
• Session notes captured for sprint planning
• No story committed with unresolved blocking questions