Key takeaways
- →Four to six is the working default. Three collapses if one person is quiet; seven produces two conversations and a spectator.
- →Choose the allocation method from the goal: random for cross-pollination, stratified for balanced expertise, self-selection only when motivation matters more than mix.
- →Give every team a named output, a named reporter and a visible clock before they leave the main room, not after.
- →Announce a five-minute warning. Teams reliably do the last third of the work in the final two minutes.
- →In hybrid rooms, put remote people together rather than splitting them one-per-table — a lone remote participant in an in-person group is effectively silenced.
- →Report-back is the bottleneck: six teams at three minutes each is eighteen minutes. Use a written channel and take two verbal reports instead.
When you split a large group into teams, three decisions determine whether anything useful comes back: how big each team is, how people are allocated, and what they are told to produce. Everything else — the icebreaker, the tooling, the room layout — is downstream of those three, and most breakouts that go badly went badly at the sizing stage.
The common failure looks like this. Forty people are put into eight groups of five with the instruction 'discuss this and we will hear back'. Three groups do good work, two drift onto something else, one has a dominant voice and four passengers, and the report-back takes twenty-five minutes of the forty you allocated. Nobody thinks the session was terrible; nobody can name what it produced either.
This guide gives you the sizing rules, five allocation methods with the situations each suits, the ninety-second launch script, the recovery moves when a group stalls, and how to get the output back without spending the rest of the session listening to summaries.
Pick the size from the task, not the room
Group size is the highest-leverage decision and the one most often made by dividing attendee count by the number of breakout rooms. Work the other way. Ask what the team has to do, and size for that.
Pairs are for disclosure and speed — checking understanding, practising a conversation, generating a first reaction. Nobody hides in a pair, which is exactly why it is the right unit for warming up a cold room. Threes are for structured roles: speaker, listener, observer. Fours to sixes are the general-purpose working unit and where most workshop tasks belong. Above six, per-person contribution drops sharply and the group naturally splits into a main conversation and a side one.
There is a real effect underneath this. As a group grows, individual effort tends to fall — each person contributes less than they would alone, because responsibility is shared and nobody's slack is visible. Six is roughly where it becomes noticeable in a forty-minute task; eight is where it becomes the defining feature of the group.
Five ways to allocate, and when each is right
Allocation is not a neutral administrative step. It decides which conversations are possible. Choosing the method from the goal takes ten seconds of thought and changes the output more than any facilitation technique you apply afterwards.
The default should be deliberate mixing rather than random assignment, unless you specifically want the surprise of random. And whatever you choose, say why out loud — 'I have mixed the teams across functions on purpose' pre-empts the grumbling that otherwise consumes the first three minutes of every breakout.
| Method | Use when | Cost |
|---|---|---|
| Random assignment | You want cross-pollination and no politics in who sits with whom | Occasionally produces a group of five quiet people and one of five loud ones |
| Stratified (one per function) | The task needs several perspectives to be answerable at all | Takes preparation; falls apart when eight people do not turn up |
| Self-selection by topic | Motivation matters more than balance and topics are genuinely distinct | Uneven sizes; popular topics get twelve people and one gets two |
| Existing teams | The output has to be owned and implemented by that team afterwards | You get the team's existing dynamic, including its existing silences |
| Deliberate seeding | You need to separate two people, or spread scarce expertise | Only workable if you know the room; do it before the session, never live |
The ninety-second launch brief
Say the output first
'Come back with three risks written on one card' before any context. Teams that hear the deliverable first organise around it; teams that hear it last spend eight minutes discussing and two minutes writing.
Name the reporter now
'Whoever joined the company most recently reports back.' Assigning it up front stops the last-minute panic and stops the most senior person defaulting into it, which is what happens every time you leave it open.
State the clock and the warning
'You have twelve minutes and I will give you a two-minute warning.' The warning is not a courtesy — it is what triggers the writing-down phase that otherwise never happens.
Give one worked example
Show a filled-in version of the output card. Ambiguity about format costs more breakout time than ambiguity about content, and an example resolves it in five seconds.
Put the brief somewhere they can reread it
On the shared board, in the breakout description, on a printed card on each table. A verbal-only brief is forgotten by minute four and one person always has to come and ask.
Send them without a countdown ritual
Do not do the 'off you go in three, two, one'. Just say go. The first thirty seconds of a breakout are the awkward ones, and drawing attention to them makes them longer.
The first sixty seconds decide the rest
Every newly formed group spends its opening minute negotiating who speaks and what exactly is being asked. Left alone, that negotiation takes two to four minutes of a twelve-minute slot, and in a video breakout with strangers it can take longer.
Remove it with a structure. The cheapest is a round: each person gets thirty seconds, uninterrupted, no discussion until everyone has been. It costs three minutes in a group of six and buys you a group where all six have already spoken — which changes the odds that the quiet two contribute anything for the rest of the session.
The alternative is silent-first. Two minutes of individual writing before anyone talks, then share. This is better when the task is idea generation, because it stops the first idea anchoring everything that follows and prevents the strongest talker setting the frame while others are still thinking.
Odd numbers are fine
Do not reshuffle forty-three people to make even groups. Two groups of five and the rest of four works. Rebalancing live costs three minutes and visibly signals that the plan did not survive contact.
Plan for the no-shows
Build allocations at eighty per cent of registrations for optional sessions. If you assign eight groups of five and twelve people miss it, you get several groups of two who cannot do the task.
Never leave one person remote
A single remote participant in an otherwise in-person group of five contributes almost nothing. Either group the remote people together, or make the whole group remote including the in-room people.
Give teams a name, not a number
'Table 4' produces no identity. A short name — even a silly one chosen in ten seconds — measurably improves how much a group commits, and makes the report-back easier to follow.
Visit, do not hover
Drop into each group for under a minute, listen, leave. A facilitator who settles in changes the group into an audience and the conversation performs for you instead of continuing.
What to do when a group stalls
Stalls have three causes and each has a different fix. If the group has misunderstood the task, they will be discussing something adjacent and confidently — restate the deliverable in one sentence and leave. If they have finished early, they are usually not finished, they have stopped at the obvious answers — ask them for the second-best option, or for what would have to be true for their answer to be wrong.
If one person is doing all the talking, do not address it in front of the group. Change the format instead: 'Take ninety seconds each, write your own version, then compare.' Switching to writing removes the airtime problem without anyone being told off.
The one to catch early is the group that never got started because nobody wanted to speak first. Two minutes of silence in a video breakout is enough for people to give up and open their email. This is why the round or the silent-write is worth the time it costs — it makes the stall structurally impossible.
Getting the output back without losing the session
Report-back is where large-group breakouts most often collapse. Eight teams at three minutes each is twenty-four minutes of the room listening to summaries, of which they retain the first two and the last one. Meanwhile the teams reporting fourth onwards find everything they wanted to say has already been said.
Collect in writing, discuss selectively. Have every team post their output to a shared board or submit it as a structured response, then put all of it on screen at once. Ask two teams to speak — the one with the most unusual answer and the one that disagreed with the emerging consensus — and take the rest as read. Ten minutes instead of twenty-four, and the room sees more of the content rather than less.
If the teams voted or ranked something, aggregate it rather than reading it out. Session Flo will collate team submissions into a single view and let the whole room vote across them, which turns eight separate lists into one ranked list the group can actually act on. The point is not the tooling — it is that the aggregation happens in seconds rather than being narrated table by table.
Decide the output, then the size
Three risks needs four people and twelve minutes. A full plan needs six people and forty. Size follows the deliverable.
Choose the allocation method deliberately
Random, stratified, self-selected or existing teams — pick from the goal and say why out loud.
Brief in ninety seconds, output first
Deliverable, reporter, clock, worked example, and the brief written somewhere they can reread it.
Structure the opening minute
A thirty-second round each, or two minutes of silent writing. Never leave the first minute to chance.
Warn at two minutes remaining
This is the trigger for writing down. Without it, a third of teams return with nothing recorded.
Collect written, discuss two
All outputs on screen at once; two teams speak. Aggregate any rankings rather than reading them out.
Sizing the whole exercise
For a group of forty, eight teams of five, twelve-minute breakout, ten-minute collection: thirty-five minutes including the brief and the transitions. Transitions are the part people forget — moving forty people into and out of breakouts costs three to four minutes in a physical room and about ninety seconds online.
For a hundred, do not run twenty breakout groups and try to hear from all of them. Run the breakouts, collect everything in writing, put the aggregate on screen, and hear from three. Twenty groups is fine as a working structure; twenty groups is impossible as a reporting structure, and that mismatch is what makes large-group breakouts feel chaotic.
If you have less than twelve minutes, do not use breakouts at all. Use pairs for ninety seconds instead — a think-pair-share round gives you most of the participation benefit at a tenth of the coordination cost, and it does not require anyone to move.
Reshuffling across a longer session
Top Tips
- Change the teams at least once in a full-day workshop. Groups that stay together all day settle into fixed roles by mid-afternoon and stop generating anything new.
- Keep teams stable when the output builds — if the afternoon depends on the morning's work, reshuffling means re-explaining, which eats the first ten minutes of every session after the break.
- Rotate one person rather than reforming everyone. Moving a single member per team preserves continuity and still breaks the settled dynamic.
- Use the same teams for scoring across a day if you are running any competitive element, then reshuffle for the discussion work. Mixing the two makes the leaderboard meaningless.
- Announce reshuffles before the break, not after. People come back to the room already knowing where to sit, which saves several minutes of milling about.
- After the second reshuffle, drop the introductions round. By then people know each other and repeating it reads as filler.
Frequently asked questions
What is the ideal team size for a breakout?
Four to six for most working tasks. Pairs for disclosure, practice or a fast warm-up. Three when you want explicit roles such as speaker, listener and observer. Above six, per-person contribution drops noticeably and the group tends to split into a main conversation and a side one within about ten minutes.
Should breakout groups be random or chosen?
Choose from the goal. Random is good for cross-pollination and avoids politics about who sits with whom. Stratified assignment — one person per function — is better when the task genuinely needs several perspectives. Self-selection by topic maximises motivation but produces wildly uneven group sizes, so cap each topic in advance.
How long should a breakout be?
Twelve minutes is the practical minimum for a group of five: roughly two minutes to organise, eight to work, two to write it down. Under twelve, use pairs instead. Above about twenty-five minutes, give the group a mid-point checkpoint or a second sub-task, because most groups reach their natural conclusion and then drift.
How do I split a hybrid group into teams?
Do not put one remote person into an in-person group — they cannot compete with a room and will effectively be a spectator. Either form all-remote groups and all-in-person groups separately, or make every group fully remote with in-room people on their own devices and headphones. The second is fairer and is worth the awkwardness.
How do I stop the report-back taking over the session?
Collect every team's output in writing to a shared board, display all of it at once, and ask only two teams to speak: the one with the most unusual answer and the one that dissented. For six or more teams this turns twenty-plus minutes of sequential summaries into about ten minutes with more content visible, not less.
What if some groups finish early?
They usually have not finished, they have stopped at the obvious answers. Give them a sharpening prompt: what is the second-best option, or what would have to be true for your answer to be wrong. Keep one such prompt ready for every breakout you run, because at least one group will need it.
Keep reading
- breakout rooms versus whole-group discussion — When splitting the room is worth the coordination cost and when it is not.
- when to switch an activity into team mode — Team scoring changes how groups behave, not just how they are counted.
- running a session for 100+ people — Above a hundred attendees, reporting back has to change shape entirely.
- why contribution falls as groups grow — The research behind the four-to-six rule.
- Session Flo pricing — Team modes and aggregated team submissions are available on all plans.