Running a survey

How many people a psychological safety survey needs

Four at the very least, and the reasons are the same reasons the results stay sealed until the survey closes. On team size, response rates, and what anonymity can and cannot protect.

Two questions get asked together: how many people do we need, and can anyone work out who said what? They have the same answer, because in a small team anonymity is arithmetic.

Four is the minimum, and here is why

A psychological safety survey asks people whether they feel safe to say what they think. If they suspect their answer can be traced, the people who feel least safe will answer most generously, and the survey will tell you the opposite of the truth. So anonymity is not a courtesy; it is the condition for the data meaning anything.

With three responses, one strongly-disagree answer is a guess away from a name, especially once the team talks about the result. With four, the arithmetic starts to protect people, and it gets stronger with every answer after that. Measure will not build a report until a survey has closed with at least four responses, and that rule is in the database, not in a setting someone can change.

Why the report stays sealed until the survey closes

If an organiser could read results as they arrived, they could compare the report at four responses with the report at five and deduce the fifth person’s answers. So nothing opens while the survey is open. It opens once, when the organiser closes it, and once anyone has read the report the survey cannot be reopened for more answers. The organiser reads first, and then the identical report goes to everyone: three working days after that first read, unless they share it sooner or keep it private.

Teams of two or three

Don’t survey them. A conversation serves a team that small better than a form does, and no architecture can make three answers anonymous from each other. The same goes for a short-lived project group: by the time the report is ready, the group has moved on.

What about response rate?

In a team, response rate matters more than headcount. Eight answers from a team of nine tells you about the team. Eight from a team of thirty tells you about the eight who felt like answering, and the people who didn’t answer may be exactly the ones the survey is for. So the report shows how many answered against the size the organiser declared, and that gap is itself worth a conversation.

You don’t need everyone. You do need the quiet people, and the best way to get them is to say, before the survey opens, that nobody will see who said what, that the team will see the whole report, and what will happen afterwards. Then do what you said.

Why the report never splits by role or group

It is tempting to ask how the junior members answered, or the people who joined this year. In a team of eight, that is a way of asking two people by name. Measure collects no role, tenure or demographic information with a team response and never splits a team report by any group. If the gradient between groups is the question, the voice and power statements ask about it directly, and the follow-up asks which kind of power is in play, with the answers counted across the whole team.

The honest limit

Architecture protects the numbers. It cannot protect a turn of phrase. In a team of four or five, a colleague who knows you well may recognise your writing in a comment, and the survey says so above every comment box. Write what you want the whole team to read, and say what you mean in ways only you would say if you are happy for that to be known.

What is never stored: a name, an email address, an account, an IP address or the time of day. Only the date of submission is kept. How the data is handled.

↑ Back to the top