Team dynamics
What to do when planning poker estimates differ.
A spread between cards is evidence that the team sees the work differently. That is valuable information, not a broken meeting.
Start with curiosity, not an average
Averaging 2 and 13 hides the reason they are far apart. Ask the lowest and highest voters to describe the path they imagined. The high card may expose a migration, a dependency, or an uncertain edge case. The low card may reveal a reusable component, a simpler slice, or prior knowledge the group can share.
Sort the discovery before choosing a card
Listen for a concrete next step. Clarify an acceptance criterion when the outcome is vague. Split the item when it contains several independent behaviours. Create a timeboxed experiment when a technical unknown dominates. Bring in a decision-maker when a business choice is blocking the estimate. The number comes after the work is understandable enough to compare.
Know when consensus is sufficient
Consensus is not unanimity or a contest of seniority. It is a shared estimate the team can use, with important concerns heard and addressed. After a focused discussion, vote again. If the item is still too ambiguous, deferring it is often more honest and less expensive than manufacturing agreement.
When the item remains broad rather than unclear, an alternative estimation technique may be a better fit for the decision.
Watch for recurring patterns
If the same kind of disagreement appears repeatedly, treat it as a process signal. Your references may be stale, a dependency may need better visibility, or refinement may be happening too late. Retrospectives are a better place to improve that system than an endless debate over a single card.