Satisfaction Histogram

When to use

The data-gathering block of a retro, when you want the team's honest read on one specific dimension — how we work together, the release process, the on-call rota. Pick it over a temperature check when the number deserves the meeting: temperature check is a five-minute barometer at the door, this is an anchored scale you spend twenty minutes reading. Pick niko-niko when you want the day-by-day trend instead of one deep read. Skip it under five people — four dots don't have a shape; just ask them.

How it runs

1. Name the dimension

   'Satisfaction' bare is too vague to plot. Satisfaction with what — how the team works together, the deploy process, code review? One dimension per chart. If you can't finish the sentence 'satisfaction with…', you're not ready to vote.

2. Write the anchors

   A bare 1-5 measures each person's private calibration, not the team. Write a sentence for 5 ('I'd change nothing about how we work'), for 3 ('it's fine, and fine is the ceiling'), and for 1 ('I'm actively working around us'). Post them beside the chart before anyone votes.

3. Vote anonymously, simultaneously

   Stickies written face-down and placed in one pass, or a hidden poll revealed at once. Public dots anchor on the first score up — you'll get a tidy cluster that isn't real.

4. Read the shape aloud, never the average

   Four shapes, four meetings. Tight cluster: consensus — go where it sits. Wide spread: no shared reality; the retro is about calibration. Bimodal: two teams wearing one name. Single outlier: one story the room hasn't heard.

5. Ask the shape's question

   Bimodal — find the fault line: role, tenure, pod, on-call. Cluster at 2 — skip the debate about whether there's a problem and ask what one notch better looks like. Outlier — invite the story, opt-in only. The shape names the topic; the rest of the retro works it.

Why it works

The written anchors make the numbers commensurable — without them every 3 means something different and the chart is noise. The shape then surfaces what an average launders: a 3.2 could be seven people mostly fine, or half the team at 5 and half at 1. Same number, opposite meetings. A bimodal chart is the format telling you you're running two teams; an outlier is it telling you someone's carrying a story alone. Surveys report the mean because they can't hold a conversation. You can.

Variations

• Two charts, same anchors: satisfaction with the team, satisfaction with the organisation. The gap between the two shapes is its own finding.
• Same anchored question every retro, photograph the chart each time. The drift across a quarter is a planning input no survey tool gives you.
• One quick chart per practice — standup, code review, deploys — then discuss only the ugliest shape.
• Fist-of-five as the live version: fingers instead of stickies when you need the read in two minutes and can wear the anchoring.

Facilitator notes

Vote too, honestly, and place yours last. Someone will compute the mean — when they do, ask which person the 3.2 describes. Don't name the bimodal fault line yourself even when you can see it; the split the team names is the one they'll own. And hold the anonymity line: the vote was anonymous, so the invitation to explain a low score is opt-in, with silence an acceptable answer. Break that once and next retro's chart is fiction.

Pitfalls

• Reporting the average. A 3.2 in a room of 1s and 5s describes nobody who voted.
• A bare, unanchored scale. Every 3 means something different; the shape becomes noise.
• Public voting. The first dot anchors the rest — a clustered chart that reads as agreement and isn't.
• Outing the outlier. Anonymous vote, opt-in story. Demand a defence and you've run your last honest histogram.

Remote tips

Hidden voting (Miro private mode, a poll bot) with a single reveal beats dots trickling onto a shared board — trickle is public voting in slow motion. Keep the anchor sentences on screen while people vote, not in chat scrollback. Screenshot the chart before closing; the trend across retros is half the format's value and remote teams get it for free.

Example outputs

• How we work together, 1-5: two 1s, a 2, four 4s, a 5. Bimodal — the fault line turned out to be on-call versus everyone else.
• Deploy process: 2, 2, 2, 3, 2, 2. A tight cluster at 2 is consensus. Skip the is-there-a-problem debate; ask what a 3 would look like.
• Same anchored question, four retros: the cluster slides 4 → 4 → 3 → 3. Nobody has said anything is wrong. The chart disagrees.
• Team satisfaction: 4, 4, 5, 4, 4 — and one 1. The 1 chose to speak. The chart knew before the manager did.

FAQ

Satisfaction histogram or temperature check?

Temperature check is the door: five minutes, an unanchored mood number, a trend line across meetings. Satisfaction histogram is an agenda block: written anchors, one named dimension, and the discussion of the shape is the point. If you'd glance at the number and move on, run the temperature check. If the distribution deserves twenty minutes, run this.

What does a bimodal histogram mean?

Two teams wearing one name. Half the room is having a different experience of the same sprint, and the fault line — role, tenure, project pod, who carries on-call — is the retro topic. Averaging a bimodal chart to 3 is precisely how that split stays invisible for two more quarters.

Should I average the scores to track over time?

Track the shape, not the mean. Keep the question and anchors identical, photograph the chart each retro, and compare pictures. A cluster sliding from 4 to 3 over a quarter is a real signal; a mean smooths a forming bimodal split into a steady 3 — hiding exactly the thing the format exists to catch.

Related activities

Recommended use cases