Trust & Safety: what it is and how to measure it

A product review is ending. The lead asks whether anyone sees a problem with the plan. Nobody does. Two people saw the problem a week ago and decided it was not worth the discomfort of raising it. The plan ships, the problem surfaces in month three, and the retrospective concludes that communication could have been better. It could not have been: the information was in the room. The belief that it was safe to share it was not.

That gap between what a team knows and what a team says is what the Trust & Safety KPI measures.

What Trust & Safety means

Trust & Safety is the PEAKTEAMS KPI for psychological safety and interpersonal trust. Amy Edmondson, whose 1999 study of 51 work teams gave the field its working definition, describes psychological safety as a shared belief held by members of a team that the team is safe for interpersonal risk taking. In plain language: on this team, can I ask a question that might sound naive, admit a mistake, disagree with a senior person or flag a risk, without expecting to be embarrassed or punished for it? Trust is the sibling belief: will my colleagues do what they said, and will they assume good intent when something goes wrong?

Three things it is not. It is not niceness: a polite team where nobody says anything difficult has low safety, not high. It is not low standards: Edmondson is explicit that safety and accountability are separate dimensions, and the teams that learn fastest are high on both. And it is not a personality trait of individuals. It is a property of the group, which is why it has to be measured at team level rather than inferred from who seems confident.

Why it matters

Three findings carry most of the weight. First, Edmondson (1999) found that teams with higher psychological safety engaged in more learning behaviour, such as asking for feedback, discussing errors and experimenting, and that this learning behaviour was in turn associated with better team performance as rated by independent observers. Second, Google's re:Work study of its own teams, known as Project Aristotle, placed psychological safety first among the five dynamics that distinguished its effective teams, ahead of dependability, structure and clarity, meaning, and impact. Third, William Kahn's 1990 study of personal engagement identified psychological safety as one of three conditions under which people bring their full selves to their work; without it, they hold back and protect themselves.

The mechanism behind all three is the same. A team's decisions can only be as good as the share of what its members know that actually reaches the decision. Safety is the valve on that flow.

What high and low look like in a real team

You can hear the difference in ordinary meetings. In a team that scores high on Trust & Safety:

• The most junior person in the room asks the clarifying question everyone else was also wondering about, and gets a straight answer.
• Someone says "I got that wrong last week" in a stand-up, and the conversation moves to what to change, not to whose fault it was.
• Disagreement with the lead happens in the meeting, not in the corridor afterwards.
• People ask for help early, while the problem is still small.
• Commitments are kept, and when one cannot be, the person says so before the deadline rather than after.

In a team that scores low:

• Questions are saved for one-to-ones or direct messages, so the group never hears them.
• Mistakes surface late, through their consequences, and the first discussion is about blame.
• Decisions get nodded through, then quietly worked around.
• The same two or three people do all the talking; the rest have learned that speaking up costs more than it returns.
• People keep private notes "just in case", a sign that they expect to need a defence.

How PEAKTEAMS measures it

Trust & Safety sits in the Process pillar of the Team P&L, alongside Communication Culture and Conflict Navigation. The Snapshot survey (about 12 minutes per member, all 18 KPIs) asks three statements for it. One is reverse-keyed, so a member who simply agrees with everything does not inflate the score.

The Deep-Dive survey adds twelve further items that take the construct apart: admitting mistakes without fear, asking for help, keeping commitments, taking risks without blame, giving each other the benefit of the doubt, handling sensitive information with care. Read together, they show whether a low score comes from fear of speaking up, from colleagues who do not deliver, or from both, which call for different responses.

Two reading rules apply. Results appear only once at least five team members have answered; the owner's own answer does not count. And every item is plotted on the Consensus Matrix, which shows its average against how much the team agrees on it. For Trust & Safety this matters more than for almost any other KPI, so read its items together. A high average with low agreement usually means the people with the most standing feel safe and the rest do not. The methodology page calls that quadrant delusion, and it is the most common way this KPI misleads a team lead who reads only the average.

How to improve it

• Ask real questions and wait. Safety grows when a leader asks "what are we missing?" and tolerates the silence, because the team learns that the question was not rhetorical.
• Respond to bad news with curiosity first. The first thirty seconds after someone reports a mistake teach the whole team whether reporting the next one is worth it.
• Name your own errors. When the person with the most authority admits a miss, the cost of admitting one drops for everyone else.
• Separate the post-mortem from the performance review. People will not analyse a failure honestly in a forum that also decides their rating.
• Keep small commitments visibly. Trust is built on dozens of kept promises about minor things, which is why a team that is sloppy about follow-through scores low here even when nobody is unkind.
• Rotate who speaks first. If the lead always opens, the lead's view anchors the room; letting others go first produces opinions rather than agreement.

Related KPIs

Trust & Safety is the condition that makes two neighbouring Process KPIs possible. Communication Culture measures whether information and opinions actually circulate; without safety, the channels exist but stay quiet. Conflict Navigation measures whether disagreement is handled openly and in time; a team that does not feel safe avoids conflict rather than navigating it. Social Cohesion is the warmth of the bonds, which is related but distinct: a close team can still be one where nobody dares to disagree.