How to Handle Disagreements During Estimation

When half the team votes 2 and the other half votes 13, most facilitators panic. They shouldn't. That gap is exactly the information planning poker is designed to surface.

Disagreement means something is unknown

A wide spread usually means at least one person sees a complexity or dependency that others don't. The goal isn't to average the votes - it's to figure out who's right and why.

Ask the outliers in the right order

Always ask the highest estimator first. They've spotted something. Once they explain, the low estimators either update their mental model or identify why the risk doesn't apply.

Then ask the lowest estimator. They might have prior experience that makes the task simpler than it looks, or they might be underestimating a risk they haven't considered.

Time-box the discussion

Set a timer. Two minutes for high, two minutes for low. Then re-vote. This keeps disagreement productive instead of letting it spiral into a design meeting.

When to split the story instead

If after two rounds the spread is still wide, that's a signal the story itself is too vague. Wide disagreement on a clearly-described story means hidden complexity. Wide disagreement on a vague story means the story needs more refinement.

Park it, add a refinement task, and move on.

When consensus doesn't mean accuracy

Be careful: quick consensus isn't the same as accurate estimation. Teams that always agree fast are sometimes just following social pressure rather than genuinely sharing a view. Simultaneous reveal (using a tool like Play Scrum Poker Online) helps prevent this by ensuring no one's vote is visible before their own.

The most valuable planning poker sessions are the ones where people disagree. That's where you find the hidden assumptions.