Your psychological safety survey is lying to you

You send out the engagement survey. The safety scores come back green. Then three months later, a senior person quits and tells you on the way out that nobody has been saying what they actually think for the better part of a year.

The survey did not lie. It just asked the wrong question to the wrong audience.

Research from Harvard, Gallup, and McKinsey keeps landing in the same place. Psychological safety in distributed teams depends on leader behaviors, not proximity, and those behaviors leave observable traces long before they show up in a survey. The problem with asking a team whether they feel safe to speak up is that answering honestly is itself a safety test. If they do not feel safe, they will not tell you they do not feel safe.

Here is what to do instead: stop asking. Start watching.

Two things to watch closely:

One, response latency on uncomfortable input. When someone raises a concern in Slack, a meeting, or a one-on-one, how long before a leader engages with it, and what is the tone when they do. Delay and deflection are the tells.

Two, cross-channel voice consistency. The same people who lead in meetings should be showing up in written channels, async threads, and informal spaces. When the activity lessens as the channel gets more public, you have a safety problem hiding in plain sight.

You can run this diagnostic on your own team this week. No survey required.

Thanks for reading,
Sarah