The company with the best people data expected composition to win — it didn't

The most interesting thing about Google's "Project Aristotle" isn't the finding. It's the prior expectation it overturned, and who held it.

What Google set out to do

In 2012, Google's People Analytics group began a two-year internal project to work out why some of its teams thrived and others struggled. They studied 180 teams — 115 engineering project teams and 65 sales pods.

Given the company, you would expect them to be hunting for the right mix of individual stars. And they were. The researchers looked at personality traits, sales skills, and the demographic composition of teams. Google is an organisation built on the premise that hiring exceptional individuals is the highest-leverage thing you can do. Finding that the right blend of people explained team performance would have been the comfortable result.

They didn't find it. Their stated conclusion: what really mattered was less about who is on the team, and more about how the team worked together.

What they found instead

Instead, their research identified five dynamics, which Google listed in order of importance:

Psychological safety — whether people feel able to take an interpersonal risk

Dependability — members reliably complete quality work on time

Structure and clarity — clear expectations, process and consequences

Meaning — a sense of purpose in the work or its output

Impact — the judgement that the work makes a difference

They also reported a set of null results that get quoted less often but are just as informative: team size, seniority and colocation showed no meaningful relationship with effectiveness at Google.

Why the expectation is the story

Plenty of organisations conclude that teams matter. Fewer of them arrive there having spent two years and considerable analytical firepower trying to prove that hiring is the answer.

That's the version of this study worth telling. Not "Google says psychological safety is important" — which sounds like something a poster might say — but: the organisation with the best people data in the world went looking for the perfect mix of individuals, and reported that the mix wasn't what separated their good teams from their bad ones.

The same conclusion, reached the hard way

TeamHive's founding observation is this study in miniature. Kimberly spent more than a decade working in and alongside leadership teams — some in organisations with excellent reputations — and kept seeing the same pattern Google's analysts eventually found in their data: talented people working next to each other did not reliably produce a good team, and how the team worked together was what separated the ones that hummed from the ones that didn't.

Google could check the assumption against 180 teams' worth of data. Most organisations can't — they hold the same assumption, that team quality is a hiring outcome, with no equivalent check on whether it's true. That check is what the TeamHive 360 is: a measured look at how each team actually works, for organisations that don't have a people analytics division to build one.

You can read more about Google's Project Aristotle here.