Backlog refinement — sometimes called grooming — is the most consistently under-executed of all Scrum practices. When done well, sprint planning takes 30 minutes. When done poorly, sprint planning becomes a 3-hour ordeal of story clarification and last-minute estimation. Here is how to do it right.
When and How Often?
The Scrum Guide does not mandate a specific refinement cadence — it simply says refinement should consume no more than 10% of the team's capacity. In practice, most high-performing teams run:
- One dedicated 60–90 minute refinement session per sprint (mid-sprint, around day 6–7 of a 2-week sprint)
- Async pre-refinement: PO shares stories 2 days early so engineers can read them before the session
- Brief "ready check" at sprint planning start to confirm the top 10 backlog items are refined and estimated
The Readiness Criteria (DEEP + Ready)
| Criterion | What it means | Checked by |
|---|---|---|
| Detailed (top items) | Next 2 sprint's worth detailed; future items high-level | PO |
| Estimated | All sprint-candidate items have story point estimates | Dev Team |
| Emergent | New items can be added as learning happens | PO + SM |
| Prioritised | Order reflects current business value | PO |
| Has acceptance criteria | Team knows when the story is done | PO |
| Dependencies identified | Blockers and external dependencies flagged | Dev Team + SM |
| Fits in one sprint | Estimate ≤ team's sprint capacity | Dev Team |
Estimation Techniques Compared
Story Points + Planning Poker: Most widely used. Relative sizing using Fibonacci numbers (1, 2, 3, 5, 8, 13, 21). Planning Poker ensures all team members vote independently, surfaces estimation disagreements, and generates useful discussion about complexity.
T-Shirt Sizing (XS, S, M, L, XL): Faster than Planning Poker. Good for high-level roadmap estimation. Less precise than story points but avoids false precision on uncertain work.
#NoEstimates: Replace point estimation entirely with throughput — how many items does the team complete per sprint. Track throughput, split large items, and use probabilistic forecasting (Monte Carlo) for roadmap commitments. Works well for mature teams with consistent story sizes.
Running an Effective Refinement Session
- Pre-read (2 days before): PO shares top 5–8 candidate stories in Jira or equivalent. Team reads them before the session.
- Open session: SM confirms the agenda — today we are refining stories for next sprint and re-checking two from last session that need splitting.
- PO explains context: For each story, 2 minutes on why it matters (business value, not implementation).
- Team asks clarifying questions: SM holds space. Every question is valid. PO answers or flags as a follow-up.
- Estimate and record: Planning Poker or dot voting. Record in Jira immediately.
- Close: SM confirms the "ready" list — stories that meet all readiness criteria and are sprint-planning-ready.
Scrum Master's Role in Refinement
The SM does not own the backlog — the PO does. The SM's job in refinement is to facilitate (keep time, ensure all voices are heard), coach the PO on story quality (INVEST criteria), protect the team from scope injection, and ensure refinement doesn't become sprint planning in disguise. If stakeholders try to commit to features during refinement, the SM redirects.