The psychological safety survey report

What a team actually receives: distributions rather than a score, why people held back, what's already working, a set of experiments to try, and — when you measure again — what changed. With a full example report you can read.

Read the example reportBuild a team survey

The spread is the finding

Most psychological safety surveys hand back a number. A team scores 4.1 out of 5, and everyone nods, and nothing happens — because 4.1 is the average of a team where four people feel safe and two feel silenced, and it is also the average of a team where everyone feels roughly fine. Those are completely different teams with completely different problems, and the mean erases the difference.

So every item in this report is shown as a distribution: how many people picked each point on the scale. You can see the shape. A cluster with two outliers looks nothing like an even spread, and the report shows you which one you have.

Different answers are the interesting data

When a team splits on an item — half saying it's easy to raise problems, half saying it isn't — the instinct is to treat the split as measurement error to be averaged away. It's the opposite. A split means the team is not one habitat. Something differs by role, by tenure, by how close someone sits to whoever holds power, and the people at the quiet end are experiencing a different workplace from the people at the loud end.

The report says this in as many words when it sees a split, and points at it as the most interesting thing on the page.

Why people hold back — and what makes speaking up possible

Knowing that people stay quiet is not much use on its own. What helps is knowing why. When someone reports finding it hard to speak up, the survey asks a short follow-up: is it that you can't predict how it'll land, that you expect it to go badly, or that you don't think it'll change anything? Ambiguity, valence, futility. Those three have different remedies, and treating one as another is how well-meaning interventions fail.

Where power is involved, it asks what kind: formal authority, informal influence, expertise, or who's in the room. A team whose voice depends on the most senior person present needs something different from a team where it depends on who is considered the expert.

The report also probes the high scores, not only the low ones. When something is working, it asks what makes it possible — because a team that knows why speaking up is easy on Tuesday has something to protect and repeat, and most surveys only ever ask what's broken.

How predictable is speaking up — and what does it cost?

Two things govern whether someone speaks: can they predict what will happen, and what does it tend to cost them. The report puts the team on both at once. A team where speaking up is unpredictable but usually fine needs clarity — visible norms, reactions people can foresee. A team where it's predictable and predictably costly needs something harder: what actually happens to people who speak has to change. Clarity won't fix a consequences problem, and the report says which one you have.

Not knowing is data

If a lot of people answer "I don't know" to whether it's clear what the team's priorities are, that is not a gap in the data. It is the finding. The report treats not-knowing as a signal in its own right rather than discarding it or folding it into a midpoint.

What people wrote, verbatim and shuffled

Comments come back exactly as written, unedited and unsummarised. They are never processed by AI, never paraphrased into themes, and they arrive shuffled and detached from the ratings their author gave, so no one's set of answers can be reassembled from the page.

People can also leave a comment against a particular statement. Those sit under that statement in the report — so "you can disagree with the lead in private; in the room, nobody does" appears right beside the item it explains — still unattributed, and never linked to the score its author gave.

A plain-language picture

Alongside the charts, the report writes a short paragraph describing the team in ordinary words: what comes easily, what takes more, and where people diverge. It is written by fixed rules from the numbers — no AI, and none of anyone's comments — and it is marked experimental, because a paragraph can be wrong in ways a chart can't. So the report asks: does this ring true? Those answers are how the wording gets better.

Your own statements, measured the same way

Every team has something the standard questions don't quite reach — handovers, on-call, a particular meeting. When you build the survey you can add up to five statements in your own words, each placed under a theme. They're answered, charted and read exactly like the rest, and if you ask the same one next time, it's compared like the rest too.

Suggested experiments, matched to what the team said

The report ends with a small number of concrete things to try, chosen by fixed rules from the pattern in the answers — not a generic list of best practice, and not an AI's impression of the team. If the picture is one of futility, the suggestions are about closing loops. If it's ambiguity, they're about making reactions predictable. The point of the whole exercise is the conversation the team has next, so the report is built to hand them a starting place rather than a verdict.

Measure again, and see what changed

Surveys of the same team join up automatically. The second report carries a section on change since the previous one, theme by theme — read, as it says, as a prompt for a conversation about what happened in between, not as a scoreboard. One honest caution comes with it: scores on the who's-in-the-room questions can worsen as awareness improves, because people start noticing a gradient they'd stopped seeing. The example report includes an earlier wave, so you can see how it reads.

You read it first, then share it

When the survey closes, the person who set it up reads the report before anyone else. Their copy has a few notes the team's doesn't — what to do next, and how to read the hard parts — and they can preview exactly what the team will see. Then one click releases it, and everyone with the survey link sees the same report. People answered on the understanding they'd see the result, so the report nudges towards sharing it; it can also be saved as a PDF for the conversation that follows.

What the report doesn't do

How anonymity holds up

Responses carry no name, no email and no IP address. Only the date of submission is kept, never the time. Results stay sealed until the survey closes and at least four people have answered, and once anyone has read the report the survey can't be reopened for more answers. Those rules exist to stop anyone reading the report at four responses, then again at five, and deducing what the fifth person said. The organiser reads it first and then releases that same report to the team.

The honest limit, which the survey states above every comment box: in a team of four or five, a colleague who knows you well may recognise your writing. Architecture can protect the data; it cannot protect a turn of phrase. Full detail is on the privacy page.

Read the example

The example report describes a team called Meridian Crew. Meridian Crew doesn't exist — the six responses, and the earlier wave they're compared with, are synthetic, written to show a team with real disagreement in it, because an example where everyone agrees would demonstrate nothing. Everything else is real: the same report builder, the same rules, the same layout your team would receive.

Read the example reportBuild a team surveySee the survey questions

Your first team survey is free, and the whole report comes with it — nothing here is held back for a paid plan. Measuring again, or running several teams, is what the plans cover. The individual self-check is free and always will be. If you'd like to talk about using this across an organisation, get in touch.