How big should randomly generated groups be?

Three or four people for discussion, four to six for a build or problem-solving task. Below three, quiet people can disengage; above six, someone always ends up spectating.

Should managers be in the same random groups as their reports?

For candid discussion, keep them apart — pre-assign managers to different groups, then randomise everyone else. People filter what they say in front of their boss.

Using a Random Group Generator for Team-Building at Work

Put a room of colleagues into groups the obvious way — “sort yourselves out” — and you get the same clusters every time: same team, same desk neighbours, same friends. Random grouping breaks that up. It is a small facilitation move with a real payoff: people talk to colleagues they normally would not, and nobody is picked last.

When random grouping is the right call

Cross-team workshops where the point is to mix perspectives.
Offsites and onboarding where people don’t know each other yet.
Brainstorming, where familiar groups tend to converge too fast on familiar ideas.
Icebreakers and warm-ups, where the stakes are low and novelty is the goal.
Rotating pairs for stand-ups, retros or peer reviews, so feedback doesn’t ossify.

When it is the wrong call

Delivery work with dependencies. If the task needs specific skills together, assign deliberately.
Sensitive conversations. Keep reporting lines apart by design, not by chance.
Groups that must be balanced on seniority or domain knowledge — randomise within bands instead (all the seniors dealt out first, then everyone else).
When people have real accessibility or language needs that grouping should account for.

Random is for mixing, not for optimising.

A method that keeps groups even

Using the random team generator:

Put every attendee’s name in the entries box, one per line.
Decide how many groups you need. For 20 people and groups of four, that’s five groups.
Turn on Remove the winner after each spin so no one is drawn twice.
Spin, and deal results round-robin: first spin to Group 1, second to Group 2, … sixth back to Group 1. Keep going until the wheel is empty.
Read the final groups off the Results tab.

Round-robin dealing keeps the sizes within one of each other automatically, which “spin a group label for each person” does not.

Running it in the room

Do it live on the screen. Watching the draw is part of what makes it feel fair.
Have a landing point ready — table numbers, breakout room names, a slide — so people move straight into the group.
Pre-seat the exceptions. If two people genuinely should not be together, assign them first and randomise the rest around them. A quiet pre-assignment is better than an awkward public swap.
Re-randomise between sessions so the afternoon groups aren’t the morning groups.

What to expect

Some groups will be stronger than others. Random does not mean balanced, and that’s usually fine for a workshop — the goal is fresh conversations, not a fair fight. If one group is clearly struggling, join it yourself for a few minutes rather than reshuffling the whole room.

Used in the right moments, a group generator is one of the cheapest ways to make a session feel less cliquey and more open — and it takes about a minute to run.