I just finished a week on teams and leadership at Wharton. I went in expecting soft-skills filler.
What I got was a pile of research with effect sizes bigger than a lot of the medicine we take without thinking about it.
Here's the part that stuck. Almost everything I believed about why teams work points at the person leading them. The evidence points somewhere else.
Eight findings, roughly in the order they landed.
The leader matters less than the setup
One study separated two things: how a team was designed, and how well the leader coached it day to day. Design predicted performance about forty times better.
Good coaching helped teams that were already well designed. It did almost nothing for the ones that weren't. And bad coaching, meaning diagnosing the problem and telling people what to fix, actively made struggling teams worse.
Hackman has a name for our instinct to credit whoever is standing out front. The leader attribution error.
Half the room isn't really participating
Managers were asked how often they had privately held back their real input on a team. The answer was 56% of teams.
The reason they gave most often, at 73%, wasn't fear of the boss. It was that someone with more expertise was in the room. Second was that somebody had already made a decent argument.
Both of those feel like good judgment while you're doing them.
Brainstorming out loud makes the group worse
Forty years of replication here. Groups generating ideas together produce fewer and lower-quality ideas than the same people working alone.
Three reasons. Only one person can talk at a time. Everyone is managing how they come across. And people drift toward whatever has already been said.
The fix isn't a longer meeting or a bigger room. Both have been tested. It's having people generate alone first, then combining and ranking together.
You're arguing about conclusions, not data
Argyris called it the ladder of inference. You observe something, select a few details, add meaning, form an assumption, reach a conclusion, act.
The whole climb takes about a second. And only the bottom rung and the top one are visible to anyone else. So two people trade conclusions, and neither has seen a single step of the other's reasoning.
The move is to say what you actually saw that got you there, then ask what they saw.
There is a good kind of conflict
A meta-analysis across 116 studies found that task conflict, meaning arguing about the work itself, can improve performance. Relationship conflict never does.
Process conflict is the sneaky one. Disagreement about roles, ownership, how things get run. Left unnamed, it turns into the personal kind, because if nobody says "we disagree about the process," the only explanation left for the friction is each other.
Your mood is the culture
Every organization has an emotional culture, including the ones where the culture is suppression. It travels through faces and tone rather than words, so you're broadcasting it whether you meant to or not.
One executive in the research was so reliably angry that his staff had text message codes for the stages of it. Red for the face, vein for the forehead.
He was building a culture very effectively. Just not the one he thought.
Charisma is a skill
Managers were randomly assigned to a short course on inspirational leadership behaviours. Three months later their bosses, peers and direct reports rated them 22% more charismatic.
Personality barely predicts any of this. Transformational leadership is a set of behaviours you can be taught, which is either encouraging or slightly deflating depending on how you thought you got here.
Culture is execution, not values
Chatman's version is the one I keep coming back to. Rules only cover situations you can predict, so they can't produce anything outstanding, because outstanding is defined by the situation nobody anticipated. Norms can.
And a strong culture needs two things, not one. Agreement about what matters, and intensity about it. Plenty of places have the first. Everyone can recite the values and nobody will stay late for them.
The thread
Every one of these points the same direction. We overweight the person at the front and underweight the conditions they set before anyone showed up.
Which is oddly good news. Charisma is trainable. Conflict is typeable. Silence is predictable. The variables with the biggest effect sizes are mostly the ones you decide in advance, on purpose, usually in the first week.
I'm still bad at most of it. But I've stopped trying to fix teams by working harder inside them.
Sources: Hackman (2004) and Wageman (2001) on team design; Mulvey, Veiga & Elsass (1996) on self-limiting behaviour; Diehl & Stroebe (1987) on brainstorming; Ross (1994) on the ladder of inference; De Wit, Greer & Jehn (2012) on conflict types; O'Neill & Barsade (2016) on emotional culture; Chatman & Cha (2003) on culture and execution.
