Teamwork predicts performance even when performance is a count of mistakes
If you have ever sat through a presentation on team effectiveness with your arms folded, you probably had a version of this objection ready.
Most of the evidence, you might think, comes from surveys. People rate how well their team works together, then rate how well their team performs, and unsurprisingly the two move together. That isn't a finding about teams. It's a finding about how people answer questionnaires. It's a good objection. It's also answerable.
Look where the outcome is recorded by someone else
In 2019, Jan Schmutz, Laurenz Meier and Tanja Manser published a systematic review and meta-analysis in BMJ Open asking a deliberately blunt question: how effective is teamwork, really?
They searched for studies in acute healthcare — emergency departments, operating theatres, intensive care — and applied a strict inclusion rule. A study only counted if it linked a teamwork process, such as coordination or non-technical skills, to a performance measure such as a checklist-based expert rating or an error.
That rule matters. In these settings, performance is not a self-assessment. It's whether the right thing happened, judged by an observer who wasn't part of the team.
They screened 2,002 articles, included 31, and pooled data across 1,390 teams.
Photo: John Crawford. In acute care, performance is an error or an expert's rating, recorded by someone outside the team.
The result
Pooled across those studies, teams with better teamwork processes were around 2.8 times more likely to perform well — the authors' own translation of the relationship they found. It appeared for both outcome measures and process measures, and it held regardless of the characteristics of the team or task.
Across a large and varied body of studies, in settings where the outcome is unambiguous, how a team works and what it achieves are reliably related.
What to do with it
Use it when someone raises the self-report objection, which they should. When you take the surveys away and measure errors instead, the relationship is still there.
And notice what the studies were actually measuring on the input side: coordination, and what the literature calls non-technical skills — communicating, sharing information, anticipating what a colleague needs. Not personality. Not seniority. Observable behaviour, of the kind a team can change.
Where this shows up in what we built at TeamHive
That last point is a design decision for us, not a footnote. TeamHive 360 measures observable patterns of behaviour — how a team coordinates, communicates and backs each other up — gathered from the team and from the people around it. Not personality types. The studies in this review looked at behaviour because behaviour is where the signal lives; so do we.
