Decisions in retrospectives
A retrospective rarely fails for want of ideas; it fails at the selection: time runs out before the team agrees on an action. Have the actions rated instead — everyone assigns each action 0 to 10 resistance points — and you have the result in three minutes. The lowest total starts, and the remaining time belongs to the conversation about the reasons.
Every team knows the pattern: the list of improvements grows, the discussion goes in circles, and in the end the round decides either nothing or everything. Both lead to the same points being on the wall in the next retrospective. Which other methods come into question is shown in the comparison of decision-making methods.
The timebox has a say
One hour of retrospective splits roughly into five phases. The numbers are a common division, not a law — what matters is which phase gets cut at the end.
| Phase of the retrospective | Usual time | What it costs when it slips |
|---|---|---|
| Check-in and read on the mood | 5 minutes | Little. This phase rarely runs off the rails. |
| Gather data: what worked, what jammed? | 15 minutes | The list gets longer, and that is fine — nothing is rated yet. |
| Look for causes | 15 minutes | This is where the round most often wanders into side topics. |
| Pick actions | 15 minutes | The most expensive point: without a method the team discusses until the time is gone. |
| Close and assign ownership | 10 minutes | The first thing to be dropped — and without it nothing happens before the next round. |
Picking the actions is the only point at which the round decides anything. It is exactly the one that eats the most time, because an open discussion has no natural end. A rating round does: when everyone has rated, it is over.
Rating instead of discussing — in four steps
- Phrase the actions. Every action needs a verb and a period of time. "Better communication" is not an action; "mornings without meetings for four weeks" is.
- Add the do-nothing option. "We change nothing" belongs on the list. It is the yardstick for everything else.
- Rate resistance. Everyone assigns each action 0 to 10 points. That takes two to three minutes, even with eight people.
- Read the result and assign ownership. The lowest total starts, with a name and a date. Where peaks show up, talk first.
Why the scale reveals more than a thumbs up is on the page about resistance voting. The method behind it is called systemic consensing.
An invented example, worked all the way through
Six people, four actions and the do-nothing option. Names and numbers are invented; they show how the arithmetic behaves. Everyone assigns each action 0 to 10 resistance points.
| Action | Ben | Eva | Timo | Nina | Ole | Ruth | Total resistance |
|---|---|---|---|---|---|---|---|
| Mornings without meetings | 0 | 2 | 1 | 0 | 2 | 1 | 6 |
| Cut the daily sync to 10 minutes | 1 | 0 | 2 | 1 | 3 | 1 | 8 |
| Code reviews within one working day | 3 | 2 | 4 | 5 | 3 | 4 | 21 |
| On-call duty in rotation | 8 | 2 | 9 | 3 | 7 | 4 | 33 |
| Change nothing | 6 | 7 | 5 | 8 | 6 | 7 | 39 |
The team starts with mornings without meetings (6 points). No score is above 2 — this action costs nobody anything and can begin right away.
The shorter daily sync follows closely at 8 pointsand is the candidate for the next round. Two actions at once are fine too — no more, or in four weeks there will be five half-finished ones on the wall.
On-call duty in rotation is the interesting case. At 33 points it sits far back, but the scores diverge: Timo gives 9, Ben 8, while Eva gives 2. That is not a voting result, that is a conflict. It belongs on the agenda, not in an action.
"Change nothing" comes last at 39 points. Good: the team wants to change something. If the do-nothing option were in front, that would be the real finding of the retrospective.
Five options and six participants fit into the free plan: it allows 5 options per session and any number of participants. What the paid plans unlock on top is in the pricing overview.
Where the method does not help
- If the team does not speak openly at all, the round delivers clean numbers about a false picture. Safety first, then the scale.
- If only one action is up for choice, the rating only measures mood. From three options on it plays to its strength.
- If a conflict between two people sits behind an action, no number replaces the conversation.
In Punkst3r the round runs in two phases: in the riot phase the actions go into the session, in the slam phase everyone rates from 0 to 10. The totals appear live in the result chart. More cases are in the use cases overview, such as feature prioritisation in a product team.
Frequently asked questions about decisions in retrospectives
How does a team decide in a retrospective which action to take?
Collect actions first, then rate instead of discussing. Everyone assigns each action 0 to 10 resistance points; the lowest total starts. That takes three minutes and leaves the rest of the timebox for talking about the reasons.
How many actions should a team start at once?
One or two. Decide on five and the next retrospective inherits five half-finished ones. The resistance totals hand you the order for free: whatever lands second today is the candidate for the next round.
Should the do-nothing option be on the list?
Yes. "We change nothing" is the yardstick for everything else. If it lands in front, either the question was framed badly or the team is not ready to change right now — both matter more than any action.
What if one action collects extreme scores?
Do not vote, talk. An action with a 9 and a 2 in the same column shows a real conflict. The number is the way into the conversation, not a substitute for it.
Decide something at last.
Collect actions, rate resistance, take the result with you — in three minutes instead of half the timebox. No account needed.