Key takeaways
- →Split the room into teams when the task genuinely needs several people and the whole group is too large to work on it at once - not because the session feels flat.
- →Three to five people per team is the working range; above six, per-person contribution falls away and a passenger seat appears.
- →Assign teams yourself rather than letting people self-select, unless the point of the exercise is precisely who chooses whom.
- →Give every team the same brief, the same clock and a named output, or you will get four groups doing four different exercises.
- →Team scoring builds cohesion but hides free riders; a visible individual contribution inside a team score fixes most of that without turning it into a contest.
- →The re-merge is where team work is won or lost - budget as much time for bringing findings back as you did for producing them.
Team activities are worth running when the task is too big for one person to do alone and the whole room is too large to do it together. That is the entire test. If a single person could produce the output in the time available, teams add coordination cost for nothing; if twelve people could genuinely work on it as one group, splitting them just fragments a conversation that was about to be useful.
The reason facilitators over-use teams is that splitting the room reliably raises the noise level, and noise is easy to mistake for progress. Four groups talking energetically look better than one group thinking quietly, but the noise tells you nothing about whether four usable outputs are being produced. Some of the best sessions have no team phase at all.
This guide covers what a split actually buys you, the right team size and why it is smaller than most people run, how to allocate people without wasting five minutes, how to score team work without breeding resentment, what to do about the passenger who lets the others carry it, how teams work in hybrid rooms, and how to bring the outputs back together so the split was worth making.
What splitting the room actually buys you
The first thing you buy is airtime. In a room of thirty working as one group, one person speaks at a time, so a forty-minute discussion gives each participant an average of eighty seconds and in practice gives four people thirty minutes and everyone else nothing. Split the same room into six teams of five and the airtime per person rises roughly sixfold, immediately, with no change to the agenda.
The second is parallelism. Six teams can cover six different problems, or the same problem from six angles, in the time one group would take on one. That only pays off if the outputs are genuinely different and you have a way of combining them - six teams producing six near-identical lists is worse than one group producing one good one, because you now have to reconcile duplicates in front of everybody.
The third is social. Small groups lower the cost of a half-formed thought. People who will not float an idea to thirty colleagues will float it to four, and the team then owns it collectively when it comes back to the room. That is the mechanism behind think-pair-share and most structured participation formats, and it is the strongest argument for splitting even when the task itself does not require it.
What you pay for all of that is coherence. The room stops sharing a single conversation, you cannot hear what most of it is saying, and the facilitator's grip loosens for the whole team phase. Everything below is about keeping that cost proportionate.
The numbers that decide a split
These are structural defaults, not research findings - adjust them to your room. The one worth defending hardest is team size. Facilitators routinely run teams of eight because it produces fewer groups to manage, and eight is where the well-documented tendency for individual effort to fall as a group grows starts doing real damage.
The named output is the other non-negotiable. A team told to 'discuss the onboarding problems' will discuss them and return with a summary of the discussion. A team told to 'come back with the three problems you would fix first, ranked, in that order' returns with something the room can act on. The instruction takes the same ten seconds to give.
When teams are the wrong call
The last item on each list is the honest one. Splitting the room because energy has dipped does work - it will lift the noise level within ninety seconds - but it treats a symptom and costs you the thread of whatever the group was working on. A two-minute stand-up, a pulse check or simply a break usually fixes energy more cheaply.
The under-eight rule catches people out. With seven participants, breaking into two teams of three and four gains you very little airtime and loses you the one thing a small group does well, which is hold a single conversation everyone can follow. Below eight people, run structured turn-taking instead - silent generation, then a round where each person contributes one item - and keep the group whole.
✓ Pros
- •The task has genuinely separable parts, or the same task benefits from several independent attempts
- •The room is large enough that whole-group discussion gives most people no airtime at all
- •You want half-formed ideas surfaced before they have to survive the full room
- •The group is new to each other and needs a low-stakes reason to talk
- •You have a clean way to combine the outputs and time budgeted to do it
- •The energy in the room needs to move from listening to producing
✗ Cons
- •The conversation is one everybody needs to hear in full - a decision, a disagreement, a sensitive update
- •You have fewer than about eight people, where the whole room is already a small group
- •There is less than fifteen minutes left, which is the split plus a rushed report-back and nothing else
- •The output is a single shared decision that will have to be renegotiated in plenary anyway
- •Nobody has a plan for what happens to four flipcharts of notes afterwards
- •You are splitting because the session feels flat rather than because the task requires it
How to split the room without losing five minutes
The whole allocation should take under sixty seconds of session time. Session Flo's team mode handles the allocation and the shared clock in the background - people who have already joined the room get put into named teams and see their own team's brief and countdown on their phone, which removes the two moments where splits usually stall: working out who is with whom, and remembering what the task was.
Decide the composition before the session, not during it
Work out how many teams and who is in them while you are building the agenda. Allocating live in front of thirty people costs three to five minutes of dead air and always produces one team of two and one of nine.
Assign rather than let people choose
Self-selection reliably reproduces the existing social map: the same functions sit together, the newest person ends up wherever there is a spare chair, and the split you wanted for cross-pollination does not happen. Random or deliberate allocation both beat it.
Mix seniority deliberately, or separate it deliberately
Both are valid, but pick one. A mixed team gets context and decision authority in the room; a peer-level team gets candour. What does not work is accidentally putting one director in a team of four graduates and hoping for an argument.
Give teams a name and a place, in one sentence
'Team one, this corner. Team two, the window. Team three, the corridor door.' Named teams with a location start working forty seconds after the instruction. Unnamed teams stand around working out who they are with.
Repeat the brief after the split, not before it
People stop listening the moment they are told to move. Give the task, let them move, then say the task again in one sentence once they have settled - or better, leave it on the screen where every team can read it for the whole slot.
Name a scribe and a reporter in each team
Two roles, assigned in ten seconds - 'whoever arrived last is writing, whoever has been here longest reports back'. Without them, teams spend their first minute negotiating roles and their last minute discovering nobody wrote anything down.
Put a visible clock on the shared screen
A team that can see three minutes left starts converging on an answer. A team that cannot is still generating options when you call time, and its report-back becomes an apology for not having finished.
Scoring: team totals, individual points, or neither
Scoring changes behaviour inside a team more than facilitators expect. A pure team score produces cohesion and cover: people help each other, and the person who does not contribute is invisible and stays that way. A pure individual score inside a nominally team activity produces the opposite - people optimise their own answer and stop helping, which makes the team a seating arrangement rather than a unit.
The version that works for most sessions is a team total with visible individual contribution. Everyone answers on their own device, the team score is the sum, and each person can see what they added. Cohesion survives because the total is what is displayed and compared; free-riding becomes visible enough to be self-correcting, without anyone being singled out on a screen in front of the room.
For a quiz specifically, the team total is the right headline number and speed is a reasonable tie-break. Avoid making speed the primary metric in a team format, because it pushes the fastest person to answer for everyone and turns four people into one player with an audience.
And sometimes the answer is not to score at all. If the team task is to produce a ranked list of risks, adding points to it converts a thinking exercise into a competition and you will get the answer that scores rather than the answer that is true. Score quizzes and games; leave analysis, generation and prioritisation unscored.
The passenger problem
Free-riding in group work is well documented and entirely predictable - as a group grows, individual effort tends to fall, and the people who reduce theirs are not necessarily the disengaged ones. Often it is the person who cannot get a word in, or who has decided the two confident colleagues have it covered. Treating it as a motivation problem leads to pep talks; treating it as a structural problem leads to the fixes above, which work.
The most effective of them is the silent start. Two minutes of individual writing before any team discussion changes who speaks first, gives the quieter members something concrete to read out, and stops the team's output being whatever the first speaker said with three people agreeing. It costs two minutes and improves almost every team activity you might run.
Shrink the team
Going from six to four removes most of the hiding space without changing anything else about the activity. It is the single highest-leverage fix and it costs nothing.
Start with silent individual work
Two minutes where everyone writes their own answer before the team talks means every member arrives with something, and the loudest voice no longer sets the agenda.
Give every seat a role
Scribe, reporter, timekeeper, devil's advocate. Four roles for four people, assigned in ten seconds, and nobody is a spectator by default.
Collect from every device
When each person submits their own response and the team total is the sum, a non-contributor shows up as a gap rather than being absorbed silently by the group.
Rotate the reporter
If the same person reports back every round, the others stop preparing to. Naming a different reporter each round keeps everyone close enough to the output to explain it.
Teams in hybrid and remote sessions
The rule that saves hybrid team work is simple: no mixed teams. A team of three people in the room and two on a call is not one team - it is a table conversation with two spectators, because the in-room three will drift into talking to each other and the audio will not carry the side comments that carry the meaning. Make teams either fully co-located or fully remote.
That sometimes means a slightly uneven split, and that is fine. Two remote teams of three and two in-room teams of four beats four balanced-but-mixed teams every time. If you have only one or two remote attendees and cannot form a remote team, pair them and give them a task suited to two, or keep them with you as a review pair rather than pretending they are inside a room team.
For fully remote sessions, the friction is the transition. Sending fifteen people into breakout rooms costs about thirty seconds each way, plus the pause where people work out who is talking first. Budget it. A ten-minute breakout is really an eight-minute breakout, and a five-minute one is barely worth the round trip.
Keep the brief visible inside the breakout - on their own device rather than on a slide they can no longer see. The most common failure in remote team work is a group spending its first ninety seconds reconstructing the task from memory, and it is entirely avoidable.
Bringing the teams back together
The re-merge is where most team activities are wasted. Four teams working brilliantly for fifteen minutes produce nothing if the recombination is four sequential monologues, because the room cannot hold four lists in its head and the last team always gets the worst attention.
Submit first, then discuss. If every team enters its output before any of them presents, the whole room sees all four at once, overlap is obvious, and each team only has to speak to what is distinctive about theirs. That turns a twelve-minute report-back into a four-minute one and leaves time to do something with the result.
Then move straight into a vote or a ranking while attention is still there. The combined list is the raw material, not the outcome, and a session that stops at 'here are the four lists' has produced a document rather than a decision. Dot voting on the merged set takes ninety seconds and gives the group something it has actually chosen.
Budget the re-merge properly at the planning stage: roughly as long as the team slot itself for four or more teams, and never less than half. If the agenda cannot afford both, run fewer teams on a narrower task rather than squeezing the recombination.
✗ A report-back that wastes the split
- •Each team reads its full list aloud, in order, for three minutes each
- •The fourth team says 'ours is basically the same as theirs'
- •Duplicates are discovered live and reconciled in front of everyone
- •The room's attention is gone by the third team
- •Notes exist on four flipcharts and one laptop nobody can find
- •No decision is made, because time ran out during the reading
✓ A re-merge that pays for the split
- •Each team submits its output before anyone speaks, so it is all on screen at once
- •Every team gets sixty seconds to add only what is not already up there
- •Duplicates are visibly merged, which shows the room where it already agrees
- •The combined list goes straight into a vote while attention is still high
- •One artefact exists at the end, not four sets of notes
- •The session ends with a ranked, owned shortlist
A worked example: forty people, sixty minutes
Notice how much of the hour is not team time. Eight minutes of silent individual work at the front and twenty minutes of recombination at the back bracket a fifteen-minute team slot, and that ratio is roughly right for most sessions. The team phase is the engine, but it is the framing and the merge that determine whether anything comes out of it.
0-8 min - Frame and generate silently
State the question and the artefact you want. Everyone writes their own three answers privately on their device. No discussion at all, so nobody's list is contaminated by the first confident voice in the room.
8-10 min - Split into ten teams of four
Pre-assigned, named, with a location or breakout room each. Roles announced in one sentence: newest person scribes, longest-serving reports, and the timer is on the main screen for everybody.
10-25 min - Teams converge on three
Each team merges its members' individual lists into exactly three items, ranked. The forced number does the work - four people with twelve items between them must argue their way to three, which is the actual thinking you wanted.
25-30 min - Submit before anyone speaks
Every team enters its ranked three. Thirty items appear on the shared screen at once and the duplicates are immediately visible, which is itself a useful finding about where the room already agrees.
30-40 min - Sixty seconds each, additions only
Each team speaks only to what is distinctive or contested in its three. Ten teams at a minute each fits, and nobody sits through the same point being made four times.
40-50 min - Vote on the merged set
Dot voting across the deduplicated list, three votes each, results hidden until everyone has voted so the early tally does not pull the later votes towards it.
50-60 min - Assign and close
Take the top five, name an owner for each out loud, and confirm the next checkpoint date. Then send the recap while people are still in the room so the output leaves with them.
Running teams with the same group repeatedly
Top Tips
- Reshuffle the teams every session for the first month, then stop. Early rotation spreads relationships across the group; permanent rotation stops any team getting good at working together.
- Keep a note of who has been in a team with whom. Two or three sessions in, you will otherwise recreate the same grouping by accident and wonder why the cross-team conversations have dried up.
- Retire the running leaderboard after about three sessions. Cumulative team scores across a series stop motivating the teams at the bottom long before they stop motivating the teams at the top.
- Vary what teams are for. If every team phase is a quiz, teams become a game mechanic. Alternate scored quizzes with unscored analysis so the format keeps carrying real work.
- Watch for a team that has quietly become one person plus three. It shows up as a report-back always delivered by the same voice with the same phrasing, and it is fixed by rotating the reporter, not by a conversation about participation.
- Ask the group occasionally whether the split is helping. A one-question poll at the end - 'was the team slot worth the time?' - will tell you within two sessions whether you are splitting out of habit.
Frequently asked questions
What is the ideal team size for a workshop activity?
Three to five for anything with a single output. Four is the safest default: it is big enough for genuine disagreement and small enough that nobody can sit out for fifteen minutes without it being obvious. Go to five or six only when the task has clearly separable parts. Above six, per-person contribution reliably falls and you are running a small meeting rather than a team, with all the same airtime problems you split the room to avoid.
Should I let people pick their own teams?
Usually not. Self-selection reproduces the existing social and functional map, which is normally the opposite of what you split the room to achieve, and it burns two or three minutes of session time while people work out where to go. Assign teams in advance. The exception is a session where who chooses whom is itself the information you want - an early-stage project where people are picking what to work on rather than who to sit with.
How do I stop one person doing all the work in a team activity?
Structure, not encouragement. Shrink the team to four, start with two minutes of silent individual writing so everyone arrives with something, give each seat a named role, and collect responses from every device so the team total is visibly a sum of contributions. Those four changes remove most free-riding without any conversation about it. Rotating the reporter each round handles the residual case where one person has become the team's public voice.
Do team scores damage the atmosphere?
Team scores are far safer than individual ones - the competition sits between groups rather than between colleagues, and losing is shared. The risk is a cumulative leaderboard across several sessions, where the teams at the bottom disengage. Keep scores within a session, celebrate the winner briefly, and reset. And avoid scoring analytical work at all: points push a team towards the answer that scores rather than the answer they believe.
How do team activities work when some people are remote?
Keep every team either fully in the room or fully remote. Mixed teams turn into a table conversation with remote spectators, because side comments and body language do not cross the audio boundary. An uneven split into co-located and remote teams beats a balanced but mixed one. If you have only one or two remote participants, pair them on a task sized for two rather than attaching them to an in-room group.
How long should a team slot be?
Eight to twenty minutes for most tasks. Under eight, the team spends the whole slot organising itself and produces nothing; over twenty, it needs a mid-point prompt or it drifts. In remote sessions subtract about a minute for the round trip into and out of breakouts. Whatever you allocate, budget at least half as long again for the recombination, because a team phase with no time to merge the outputs is a phase that produced nothing.
Keep reading
- splitting a large group into effective teams — The allocation mechanics in more depth, including uneven numbers.
- breakout rooms versus whole-group discussion — The decision that comes before choosing a team size.
- team scores versus individual scores — How each scoring model changes behaviour inside a team.
- why contribution falls as groups get bigger — The evidence behind keeping teams to four or five people.
- team modes in Session Flo — Team allocation, shared timers and team scoring on one room code.