High-Performing Team · Project Aristotle
High-performing Team
What is a high-performing team?
- It’s not who is on the team, but how they work together
- Psychological safety is the foundation; the other four factors rest on it
- Safety means room for disagreement, not the absence of conflict
What this research really changes is what an organization targets when a team underperforms. In the past, a low-performing team meant replacing people or adding one more star. After Project Aristotle, you look first at what happens in the room rather than who is sitting in it. How speaking time is distributed in meetings, how many hours it takes for bad news to surface, what reaction follows a dissenting opinion: these become manageable variables.
The five factors are not a flat checklist; they build on one another. Psychological safety has to be the foundation for the other four to work. However clear the roles are, if no one can say 'this schedule isn’t realistic,' structure and clarity exist only on paper. Edmondson divided teams into four quadrants along two axes, safety and performance standards, and learning happens only when both are high. High safety with low standards produces a team that gets along well but never improves. Safety is also different from trusting a particular colleague. It is closer to a shared belief that it is safe to take risks on this team.
When measuring, fix the unit and the cycle first. Run the diagnostic not at the company or division level but with groups of 5 to 8 people who work together every day, or the numbers won’t be readable. Division averages let well-run teams and struggling teams mask each other. Four to five items per factor is enough, and scores begin to move only at least three months after you intervene.
The most common mistake in practice is treating this as 'improving the mood.' Companies hold workshops, add team dinners, and have leaders announce that people should speak freely. But if someone actually voices dissent at the next meeting and meets a stiff face or 'that’s already been decided,' the team remembers not the announcement but those three seconds of reaction. Telling people they can speak up and then penalizing them is worse than an organization that never invited them to speak. The same goes for diagnostic surveys: if you pull the scores and change nothing for six months, response rates fall in the next round and the survey itself becomes a target of cynicism.
Eun-a ChoExpert in communication, stress management, career development, and Gen Z/Millennial training
John KimCEO of Nanum Consulting and author of B2B Marketing DesignThese are the results of Google’s 'Project Aristotle,' which analyzed about 180 internal teams over several years starting in 2012. The concept of psychological safety itself builds on earlier research by Amy Edmondson.
Consider a hypothetical case: the development division of Company A, an industrial software company with 240 employees. Three teams, 42 people, and releases slipped by an average of three weeks over the past year. Incidents were reported long after they hit, and two team leads did most of the talking in weekly meetings.
- The head of development turned the five factors into 22 questions and ran an anonymous survey; 38 of 42 people responded. Psychological safety scored lowest, at 2.6 out of 5.
- In interviews with the three team leads, the same sentence came up all three times: “If you report bad news, the first thing you hear is why you’re only telling us now.” The division head admitted the problem was their own reaction.
- The division head made the first 10 minutes of the weekly meeting a fixed 'share what's blocking you' slot. In the first session, he opened with the time he had set the wrong priority last month and lost two weeks.
- Team leads decided to call on people who had never spoken in a meeting, asking 'What do you think about this one, ○○?', and to record who spoke in the meeting minutes for three months.
- They also revised the incident response policy so that incidents are shared within 30 minutes, even before root-cause analysis, and stated in writing that no one is blamed for reporting quickly.
- Six months later, re-measuring with the same items showed psychological safety rising from 2.6 to 3.8, release delays dropping from an average of three weeks to four days, and the time to first incident report falling from an average of 9 hours to 40 minutes.