Votes of 3, 5 and 13: resolving disagreement
A password-recovery case to distinguish differences in scope, uncertainty and team references.
The case: one sentence, three different jobs
“Allow password recovery by email.” In this fictional case, Ana votes 3, Bruno 5 and Cris 13. Mean 7 and median 5 explain neither person's understanding.
Ana assumes a reusable screen and existing email. Bruno includes expired-link handling and tests. Cris includes selecting and integrating a provider. Another vote without clarification repeats the issue.
Facilitation uncovers what estimates contain. Rather than “Why so high?”, which can pressure someone to lower a vote, ask which steps and risks were included.
Three questions that turn numbers into information
First, ask what already exists. Review components, permissions, services and data. Confirming the email service removes a major scope difference.
Second, ask what must happen to finish the story. Specify expiry, single use and clear messages. A screen-only estimate differs from one including these checks.
Third, ask what is unknown and how to learn it. If invalidating links is unclear, assign a bounded investigation rather than hiding uncertainty behind a precise number.
- What exists and can be reused
- What the team's Definition of Done includes
- What is unknown and who will clarify it
Choose whether to clarify, split or investigate
Clarify when interpretations differ. Confirm the existing service and add tests to the description so everyone re-estimates the same work.
Split separable useful outcomes: email recovery for existing accounts could precede another channel if each meets quality criteria. Separating coding from testing does not itself reduce delivery size.
Investigate technical or product questions that prevent sensible estimation. Define evidence, an owner and a review time. Investigation should enable a decision, not postpone the story indefinitely.
Repeat when new information appears
After clarifying, choose cards privately again. Votes of 5, 5 and 5 suggest closer understanding, not a guarantee against implementation surprises.
If votes become 5, 5 and 8, check remaining assumptions. Do not repeat until someone gives in. Agree how to proceed, document risks or split again.
Use distribution and history in StoryFlop to recall the discussion. Do not automatically adopt the mean: it can fall between cards and is not a shared decision.
What to write in the task afterwards
Illustrative note: existing email; single-use link, expiry and error tests included; provider replacement excluded; agreed estimate 5; review if the service changes. It is not a universal estimate.
Using points to compare individuals or demanding identical initial votes removes useful information. Differences uncover hidden work and support decisions.
If questions recur next session, add them to preparation. A concrete process improvement matters more than matching cards.
Put it into practice
Sources and further reading
The examples and StoryFlop instructions were written for this section. These sources provide context for the method and refinement: