In 2012, Julia Rozovsky had a problem most people would envy. A researcher inside Google’s People Operations division, she had access to performance data on thousands of employees across hundreds of teams. Leadership wanted a simple answer: What makes some teams brilliant and others mediocre?
Rozovsky and her colleagues launched Project Aristotle — named after the philosopher’s observation that the whole is greater than the sum of its parts. They studied 180 teams across engineering and sales, analyzed 250 team attributes, and spent two full years looking for the formula.
What they expected to find was a recipe built on talent, seniority, or personality combinations. What they actually found upended a decade of conventional hiring wisdom.
The Surprise at the Top of the List
The single strongest predictor of high performance was not who was on the team. It was how the team interacted.
Specifically, it was psychological safety — a concept Harvard Business School professor Amy Edmondson first defined in 1999 as “a belief that one will not be punished or humiliated for speaking up with ideas, questions, concerns, or mistakes.
Teams where members felt safe to take interpersonal risks outperformed every other configuration Google could test. According to the company’s internal data, psychologically safe teams outperformed others by 27%, showing higher productivity, more innovation, and greater employee satisfaction.
“We were pretty confident that we’d find the perfect mix of individual traits and skills necessary for a stellar team,” Rozovsky told Google’s re:Work publication. “We were dead wrong.”
Five Dynamics, One Clear Winner
Project Aristotle identified five dynamics that distinguished high-performing teams from the rest. Psychological safety sat at the foundation, but four other factors completed the picture.
Dependability came second. Can team members count on each other to deliver quality work on time? When the answer was yes, trust compounded. When it was no, even psychologically safe teams stalled.
Structure and clarity followed. High-performing teams had clearly defined goals, roles, and execution plans. Everyone knew what was expected and how their work connected to the larger mission.
Meaning was fourth. Team members needed to find personal significance in the work itself — not just the paycheck attached to it. This aligns closely with what researchers have found about how intrinsic motivation disappears the moment you add incentives.
Impact rounded out the five. Teams performed better when they genuinely believed their work mattered beyond their own department.
But here is the critical nuance: psychological safety was not just the most important factor. It was the prerequisite for the other four. Without it, dependability eroded, clarity felt performative, and meaning never took root.
What Psychological Safety Actually Looks Like
Edmondson, whose research inspired Rozovsky’s breakthrough, is careful to distinguish psychological safety from niceness. “It’s not about being warm and fuzzy,” she has said in multiple interviews. “It’s about candor. It’s about being direct, taking risks, and being willing to say, ‘I screwed up.'”
On Google’s highest-performing teams, two behaviors appeared consistently. First, members spoke in roughly equal proportions during meetings — a pattern researchers call “conversational turn-taking.” Second, members demonstrated high “social sensitivity,” meaning they could intuit how colleagues felt based on tone, expression, and body language.
When those two norms were present, even teams composed of average performers punched well above their weight.
When they were absent, teams stacked with top individual performers underdelivered. Talent, it turned out, was necessary but nowhere near sufficient.
Why Most Managers Get This Wrong
The instinct in most organizations is to optimize for individual capability. Hire the best people, give them the best tools, and get out of the way.
Project Aristotle’s data suggests that approach misses the point. A team of A-players operating in a low-safety environment will consistently lose to a team of B-players who trust each other enough to be honest.
Dr. Timothy Clark, author of The 4 Stages of Psychological Safety, puts it bluntly: “If you punish vulnerability, you get theater. If you reward it, you get innovation.”
The research supports him. A 2017 Gallup study found that only 3 in 10 U.S. employees strongly agreed that their opinions counted at work. Gallup estimated that moving that ratio to 6 in 10 would reduce turnover by 27% and increase productivity by 12%.
The math is not subtle. Psychological safety is not a soft skill. It is a performance variable with a measurable dollar value.
How to Build It Without Making It Weird
The biggest misconception about psychological safety is that it requires a grand cultural overhaul. It does not. The research points to small, repeatable behaviors that leaders can start using immediately.
Ask before you tell. When a manager opens a meeting by sharing their own opinion first, the team’s effective IQ drops. Edmondson’s research shows that leaders who ask genuine questions before stating positions get 40% more input from their teams.
Respond to bad news with curiosity, not blame. The first time a team member reports a failure and is met with “What can we learn from this?” instead of “How did you let this happen?” — the tone shifts for everyone watching.
Name your own mistakes publicly. When leaders model fallibility, it gives permission. This is what the best managers do when they give feedback that actually changes behavior — they start from a position of shared vulnerability.
Make it safe to disagree with you specifically. Psychological safety is not about making everyone comfortable. It is about making dissent survivable — especially dissent directed upward.
The Limits of the Research
Project Aristotle is not without critics. The study was conducted inside Google, a company with resources, hiring standards, and cultural norms that are far from universal. Extrapolating from 180 Google teams to a 50-person manufacturing firm requires caution.
Additionally, the study relied heavily on self-reported survey data. Teams that already felt safe were also more likely to report feeling safe, creating a potential circularity in the findings.
But the broader body of evidence extends well beyond Mountain View. Edmondson’s original 1999 research, conducted across hospitals, found that units with higher psychological safety reported more errors — not because they made more mistakes, but because they were honest about the ones they made. That honesty led to faster learning and fewer repeated errors over time.
The Takeaway for Leaders
Google spent two years and significant analytical resources looking for the secret ingredient of team performance. They tested for IQ, tenure, personality type, educational background, and dozens of other variables.
The answer was simpler and harder than any of those: Do people on this team feel safe enough to be themselves?
That is not a question you answer with a policy memo. It is a question you answer with behavior — specifically, the behavior of whoever holds the most power in the room.
As Rozovsky’s team concluded in their final report, “Who is on a team matters less than how that team’s members interact, structure their work, and view their contributions.”
The best teams are not built from the best parts. They are built from the best conditions.