When the votes disagree

A spread of votes isn't a problem with planning poker: it's the point of it. When one person votes 2 and another votes 13, they almost always know different things about the work. The job is to find out what.

What a spread usually means

• Someone has spotted a risk. A hidden dependency, a fragile part of the code or a lot of testing.
• Someone knows a shortcut. An existing component, a library or a similar story done before.
• People read the story differently. One person included the database change and another didn't.
• The story isn't ready. If nobody is sure what "done" means, the votes will be all over the place.

Running the discussion

1. Ask the highest and lowest voters first. Ask what they were thinking, not why they are wrong. Keep it short, a minute each.
2. Let the product owner clarify. Many spreads disappear once the scope is clear.
3. Vote again. After the discussion, everyone votes again in secret. Don't just ask people to agree out loud, because the loudest person will win.
4. Agree a number. If the second round is close, the facilitator picks the agreed number, often the median or the most common card.

When the team still can't agree

• Split the story. Big spreads on big numbers usually mean the story is too large to estimate well.
• Take the higher number. If the difference is one card, choosing the higher one is usually safer than another round.
• Park it. If the team doesn't know enough, add a short spike to find out and come back to the story later.
• Timebox it. If an item takes more than about ten minutes, stop and move on.

How EstimateIt helps

Votes stay hidden until the facilitator reveals them, so nobody is anchored by the first number they hear. When a section is revealed, EstimateIt shows the average and median of number votes and says whether everyone agrees, the votes are close (one card apart) or they are split. When they are split, the cards at each end are highlighted and it suggests that the highest and lowest explain. The facilitator can press Vote again to start a new round, or pick the agreed estimate.

Common questions

What should you do when planning poker estimates differ?

Ask the people with the highest and lowest votes to explain, clarify the story, then vote again in secret. Two rounds are usually enough.

Should you take the average of planning poker votes?

Not automatically. The average hides the reason for the spread. Use the discussion to agree a number, and treat the average and median as a guide.

What if the team still disagrees after two rounds?

The story is probably too big or not clear enough. Split it, add a spike to find out more, or take the higher of the two numbers and move on.

Try it with your team

EstimateIt is free planning poker for agile teams, with nothing to install and no sign-up.

Start a planning poker room