How to break ties in planning poker votes?
Our planning poker sessions often end up with split votes. For example, yesterday we had 3 developers vote 5 and 3 vote 8 for the same story.
We ended up spending 20 minutes debating it before finally settling on 8. But this happens almost every session and it's slowing us down.
How do other teams handle ties? Do you:
- Always round up to the higher estimate?
- Take the average (6.5 → 8)?
- Let the person doing the work decide?
- Force more discussion until consensus?
What's the best practice here?
After facilitating planning poker for 10+ teams, here's my recommended tie-breaking protocol:
Default Rule: Round Up
If half the team votes 5 and half votes 8, go with 8. Here's why:
- The people voting 8 are seeing complexity or risks the 5-voters aren't
- It's better to overestimate and deliver early than underestimate and miss commitments
- Story points have diminishing precision at higher numbers anyway
Quick Discussion Round
Before automatically rounding up, do this:
- High voter speaks: "I voted 8 because I'm worried about X"
- Low voter speaks: "I voted 5 because I think we can Y"
- Revote once: Now that everyone heard both perspectives
- If still tied: Take the higher number
This whole process should take 2-3 minutes max, not 20.
When to Defer to Specific People
Sometimes you should weight certain votes more heavily:
- Implementation expert: If Sarah has built 10 payment integrations and she votes 13 while others vote 8, go with 13
- Tech lead on that domain: Backend lead votes higher for database work? Listen to them
- Person doing the work: If someone is assigned already, their estimate matters more
Avoid Averaging
Don't average 5 and 8 to get 6.5. Fibonacci doesn't have a 6 or 7. You'll end up with weird numbers. Just pick the higher Fibonacci number (8).
Time-Box Discussions
Set a rule: "If we can't converge in 3 minutes, we round up and move on." Planning poker is about speed and consensus, not perfection.
We use a "highest bidder speaks first" rule. The person with the highest vote explains their reasoning. Often the lower voters realize they missed something and adjust up.
Rarely does the high voter hear the low explanation and think "oh yeah, I was wrong." Usually the high voter has spotted a real complexity issue.
Love the "round up + 3-minute discussion" approach. We were letting debates go on way too long trying to get perfect consensus.
I'll propose this protocol to the team for our next sprint planning. Thanks!
One thing to watch: if your team is consistently tied between adjacent Fibonacci numbers (5 vs 8, 8 vs 13), you might need to recalibrate your reference stories.
Ties are normal occasionally, but if every story is a tie, your team doesn't have shared understanding of the scale.