Disagreement is the point
When votes spread across the scale, it can feel like failure. It is not. Planning Poker is designed to surface different assumptions before a sprint starts. That saves you from surprises later.
A wide spread means someone sees hidden complexity, missing scope, or a dependency others missed. That is exactly the signal you want.
Understanding vote spread
Not all spreads are equal. The pattern tells you what is going on:
| Spread | What it means | Action |
|---|---|---|
| Tight (3, 3, 5, 3) | Shared understanding | Accept the majority or average |
| One step (5, 5, 8, 5) | Minor disagreement | Brief discussion, then accept either value |
| Two steps (3, 5, 8, 5) | Different assumptions | Discuss, clarify, re-vote |
| Wide (3, 5, 13, 8) | Fundamental misalignment | Deep discussion or split the story |
| Bimodal (3, 3, 13, 13) | Two interpretations | Clarify scope before proceeding |
The outlier discussion protocol
When votes diverge, hear from the outliers first. It keeps the discussion grounded.
- Identify the outliers - "We have a 3 from Alex and a 13 from Jordan. Everyone else is 5 to 8."
- Low voter speaks - "Alex, what are you seeing?" Example: "I assumed we can reuse the component."
- High voter speaks - "Jordan, what is driving the 13?" Example: "That component does not support the new data model."
- Group discusses - Now the real question is visible.
- Re-vote - With shared context, votes usually converge.
Neither person was wrong. They were working from different assumptions.
When to keep discussing vs move on
Not every disagreement needs a full debate. Use these heuristics.
Keep discussing when:
- The spread is 3+ steps (3 vs 13, 5 vs 21)
- Votes are bimodal
- The story is high priority for this sprint
- The team has not worked in this area recently
- Someone says "Wait, are we including X?"
Move on when:
- The spread is 1 to 2 adjacent values (5 vs 8)
- Discussion is circling without new information
- The story is low priority or far out
- You have hit the timebox
- SprintJam's Judge suggests a value everyone accepts
Using The Judge effectively
The Judge proposes a consensus value based on the distribution. It helps when:
- Breaking ties - Split between 5 and 8? The Judge suggests one.
- Starting discussion - "The Judge suggests 8. Strong objections?"
- Ending debate - If you cannot converge, accept it and move on.
It is a starting point, not an authority. Override it when the team has a good reason.
Techniques for stubborn disagreement
- Split the story - Separate backend from frontend, or break it into thin slices.
- Make assumptions explicit - "Estimate assuming we can reuse the component."
- Default to the higher value - When uncertain, the cautious estimate is safer.
- Take it offline - A short spike can beat a long debate.
- Accept the uncertainty - "We will learn once we start; call it an 8."
Psychological safety matters
Consensus only works if people feel safe to disagree. Watch for these signs:
- Anchoring to the first voice - Estimates cluster around whoever speaks first.
- Silent participants - Some people never explain their votes.
- Quick capitulation - "Oh, you said 5? I guess I agree."
- Seniority bias - Junior developers defer to seniors.
Building safety
- Celebrate outliers - "Good catch, I had not thought of that."
- Ask quiet voices first - Give them space before louder voices dominate.
- Model vulnerability - Leaders should admit when they do not know.
- Separate estimate from identity - Critique the estimate, not the person.
The goal is shared understanding
Consensus does not mean everyone agrees. It means differences were aired and the team can move forward with a shared estimate. That is enough. The estimate will evolve as you learn more.