Estimation field guide

How to Run a Better Refinement Meeting

A productive refinement meeting leaves the team with clearer work and fewer hidden assumptions. Estimating is one part of that outcome, not the agenda’s only purpose.

Prepare decisions, not a queue of tickets

Select a small set of upcoming backlog items with the Product Owner. For each item, state the user problem, the intended outcome and the decision the team needs to make. A short list of relevant stories is more useful than reading through the entire backlog without a purpose.

Share draft acceptance criteria, designs and known dependencies in advance. Invite people who can clarify important constraints. Developers should have the opportunity to examine unfamiliar areas; asking for an instant estimate on unseen work is a weak substitute for preparation.

An example 45-minute agenda

  1. First 5 minutes: confirm the purpose and which items matter most. Agree on the questions that need answers today.
  2. Next 25 minutes: work through the selected stories. Clarify expected behavior, boundaries, acceptance examples and dependencies. Split items that contain several independently valuable outcomes.
  3. Next 10 minutes: estimate items that are sufficiently understood. Discuss differences in complexity, effort and risk rather than demanding immediate agreement.
  4. Final 5 minutes: record decisions, unresolved questions, owners and follow-up dates. Confirm which items need another conversation.

This is a facilitation example, not a Scrum requirement. Refinement is ongoing work; Scrum does not prescribe a dedicated refinement event or this timebox. Adjust the meeting to the questions and people involved.

Make uncertainty discussable

Ask for concrete examples: What happens when permission is denied? What is the largest expected dataset? What should the user be able to do after an error? Encourage participants to say “we do not know yet”. When an answer needs investigation, record a bounded question instead of using the whole meeting to speculate.

For the CSV export story, separate the familiar file-generation work from uncertain data-volume behavior. That distinction helps the Product Owner understand whether the next step is a scope decision, implementation or investigation. A smaller story is useful when it still delivers a coherent outcome, not merely because it fits a target estimate.

Estimate after clarifying the work

Complexity

How difficult is the solution? Consider business rules, technical interactions and the reasoning needed to implement and test them.

Effort / scope

How much work is involved? Include implementation, testing, review and documentation. A large amount of familiar work can still be straightforward.

Risk

What remains uncertain? Consider untested assumptions, external dependencies and unknown behavior. Record what would reduce that uncertainty.

Let the developers doing the work own the sizing. A facilitator can structure the discussion without choosing the estimate for them. With Consonus, review the three rating distributions and ask about the widest disagreement. Agree on final ratings only after sharing the relevant assumptions.

Leave a useful record

  • Update the story and its acceptance examples so the learning survives the meeting.
  • Give each open question an owner and an agreed next step. An unassigned question is easy to rediscover next week.
  • Check whether the session clarified upcoming work. Counting estimated tickets alone does not tell you whether the refinement helped.

Ready for clearer estimates?

Bring one story to Consonus. Compare your team’s complexity, effort and risk ratings, discuss the different assumptions and agree on your next step.

Free during public preview. No payment details required.