Reading results
Why an average gets psychological safety wrong
Two teams can have the same average and need opposite conversations. Here is why, and what to look at instead.
Most survey tools hand a team back a number. It is the natural thing to do with a column of answers, and it is the single most misleading thing you can do with a psychological safety survey.
Two teams, one average
| Team | Four answers to “On this team I can raise problems and challenge ideas” | Average |
|---|---|---|
| A | 3, 3, 3, 3 | 3.0 |
| B | 1, 2, 4, 5 | 3.0 |
Answers on a five-point scale, 1 strongly disagree to 5 strongly agree.
Team A has settled into a shrug: nobody feels unsafe, nobody feels able to push. Team B contains someone who strongly disagrees, sitting next to someone who strongly agrees. One of them is experiencing a different workplace from the other. These teams need opposite conversations, and the average says they are the same.
Now make it six people: four agree, two strongly disagree. The average is 3.2 out of 5, a team that looks fine. The two at the bottom are the finding, and the average has buried them. The least-safe experience can’t be averaged away.
Why it happens
An average assumes the people in the team are samples of the same thing, so their differences are noise to be smoothed out. In a team, the differences are the signal. Psychological safety varies by role, by tenure, by how close someone sits to whoever holds power, and by who is in the room. When a team splits on a statement, the split is telling you the team is not one habitat.
There is a second problem. People manage what is measured. Once a team has an average, the average becomes a thing to raise, and the quickest way to raise it is for the quiet people to answer a little more generously next time. The number improves and nothing else does.
What to look at instead
The spread. How many people chose each point on the scale. A cluster with two outliers looks nothing like an even spread, and they mean different things.
The lowest answer. Someone gave it. A team is only as safe as that person’s experience of it, whatever the mean says.
Where the team disagrees with itself. A statement where answers are far apart is the best place to start a conversation, because the people at the quiet end know something the people at the loud end don’t.
How many couldn’t say. “I don’t have enough information to say” is an answer, and several of them means parts of the team are working out of each other’s sight.
Why. Ask the people who hold back what is closest to the reason. A split caused by futility needs a different response from one caused by fear of the consequences.
This is what every Measure report shows. There is no overall score. Each statement is drawn as a distribution, one mark per person, with the lowest answer and a spread figure beside it, and the report names a wide split as the most interesting thing on the page. Theme scores exist, as a summary, and the report tells you to read the spreads before trusting them. See it in the sample report.
What to do with a split
Don’t argue about whose answer is right. Both are. Ask instead what is different about the experience of the people at each end, and what would have to change for the lowest answer to move. That is a conversation a team can have without anyone having to identify themselves, and it is usually the most useful thing the survey produces.
Questions
Should I average Likert-scale answers in a psychological safety survey?
Not as the headline. An average hides the people at the quiet end, and two teams with opposite problems can share the same mean. Show the distribution of answers, the lowest answer and where the team disagrees with itself, and use an average, if at all, as a labelled summary.
What is a good psychological safety score?
There isn’t one. A team’s only meaningful comparison is with its own past, and the useful questions are where people disagree, what the lowest answer was and why people hold back. A high average with two people at the bottom is not a good result.
How do I report psychological safety results without an average?
Show the spread for each statement, name the splits, say what people gave as their reasons for holding back, and lead with what the team will try next. That is more useful to a leader than a number, and harder to game.
Why does the average hide the least-safe person?
Because an average treats differences between people as noise to smooth out. In a team, those differences are the finding: one person strongly disagreeing beside three who agree means the team is not one habitat, and the average describes nobody in it.