A remote all-hands day lives or dies on the breakout rooms. The keynote is fine, the roadmap slides are fine, and then people go into small rooms with colleagues they’ve never spoken to, and that’s the part they remember. The goal is usually stated as “get people mixing”, and the usual result is that the same few people keep landing together, because nobody planned for round two. This guide plans a day for a 30-person remote company with three breakout rounds and eight weeks of follow-up coffee pairs, and works out how much mixing each setup actually delivers.
The day we’re planning
Thirty people across four time zones, one video call, a facilitator with the host controls. The agenda:
- 10:00 Welcome and company update
- 10:30 Breakout round 1, 20 minutes
- 11:30 Breakout round 2, 25 minutes
- 14:00 Breakout round 3, 25 minutes
- 15:00 Close, and the coffee-pair programme announced for the following eight weeks
Three tools do the work: the Breakout Room Assigner for the rounds, the Random Icebreaker Picker for the opening question in each round, and the Random Pair Generator for the weekly coffees afterwards.
The first decision: room size
Before touching a tool, decide how big the rooms are, because room size decides how much mixing is even possible.
With 30 people there are 435 possible pairs of colleagues. A room of six contains 15 pairs; five such rooms make 75 pairs per round. A room of three contains 3 pairs; ten such rooms make 30 per round. Bigger rooms meet more people per round, but they also collide more. In any round, a given pair of colleagues lands in the same room of six with probability 5 in 29, about 17%. In rooms of three it’s 2 in 29, about 7%.
The collisions get worse in later rounds, and there’s a hard floor. After round one, each room of six in round two needs six people, but there are only five round-one rooms to draw them from. So at least two people in every new room must have shared a room before. With five rooms, round two can never have fewer than five repeat pairs, no matter how cleverly it’s drawn.
The Breakout Room Assigner has an Avoid repeat pairings option, covered below. Here’s what simulation of 4,000 days gives, counting pairs who were already together in an earlier round:
| Room setup | Avoid repeat pairings | Repeats in round 2 | Repeats in round 3 |
|---|---|---|---|
| 5 rooms of 6 | off | about 13 | about 24 |
| 5 rooms of 6 | on | about 8 | about 18 |
| 6 rooms of 5 | off | about 8 | about 15 |
| 6 rooms of 5 | on | about 4 | about 10 |
| 10 rooms of 3 | off | about 2 | about 4 |
| 10 rooms of 3 | on | under 0.1 | about 1 |
The lesson is plain: if mixing is the point, use small rooms. The avoid option helps at every size, but it can only work with the room the arithmetic leaves it. With rooms of three and the option on, three rounds give almost everyone six different people to talk to. With rooms of six, someone will say “oh, you again” by lunchtime.
We’ll use rooms of three for all three rounds. If your agenda needs a bigger discussion group for one session, put that session last, so the small-room rounds have already done the mixing.
Setting up the Breakout Room Assigner
Paste all 30 names, one per line. The list is saved in your browser, so you can set it up the day before.
In Options:
- Mode: By size per room. (The other choice, By # rooms, splits into a fixed number of rooms instead.)
- Number: 3.
- Avoid repeat pairings: ticked.
Click Assign rooms. You get ten rooms listed as Room 1 to Room 10, and Copy gives you a plain-text version, one room per line, to paste into the host’s room assignments or the chat.
How the avoid option works: with the box ticked, the tool makes 25 different random arrangements and keeps the one with the most pairs that haven’t been in a room together before. Then it remembers this round’s pairs, in your browser, for next time. It isn’t a perfect optimiser, which is why rooms of six still show repeats, but at three per room 25 tries is plenty.
Three details catch people out:
- Round one has to be run with the box ticked. Only rounds run with the option on are remembered. If you run round one without it, round two has no idea who was together and can’t avoid anything.
- The box stays as you left it. Options are saved in your browser along with the list, so a box ticked for round one is still ticked after a reload. That cuts both ways: if you unticked it for a one-off split last week, it is still unticked today. Check it before round one.
- Last month’s pairings are still in memory. The tool remembers the most recent 500 pairs, which at 30 pairs per round is about 16 rounds. If you ran rooms for a workshop last month, those pairs will count as “already met” today. That might be what you want. If not, click Clear history before round one; it clears the remembered pairings along with the Recent results list.
One more option the tool has: a line wrapped in parentheses, like (Ana, Raj), keeps those people together. It’s useful for a new hire and their onboarding buddy in round one. In By size per room mode, though, a locked group counts as one entry, so with a size of three, their room gets three entries and therefore four people. That’s usually fine for one room. Don’t lock five pairs and expect rooms of three.
Icebreakers: one question per round, sent to every room
Small rooms of people who don’t know each other need a first question, or they spend four minutes on the weather. The Random Icebreaker Picker has three categories: Classroom, Team and Training. Team has 32 questions, things like “What’s a project you’re proud of and rarely get to talk about?” and “Two truths and a lie about your career.”
Leave the names box empty, which gives you a question on its own, set Category to Team, and click Pick an icebreaker. Then broadcast that one question to every room as the round opens.
Why one question for all rooms rather than a different one each? Partly logistics: it’s one broadcast message. But also repetition. The picker has no no-repeat option, so each click is an independent draw from the 32. Drawing ten questions for ten rooms gives a 79% chance that at least two rooms get the same question, which is harmless but looks careless. Drawing three questions for the whole day, one per round, has a 9% chance of a repeat. If it happens, click again.
If you do type names into the picker, it picks a name at random along with the question, which suits “who answers first” in a single room. It picks independently every time, so the same name can come up twice in a row; don’t use it as a turn order.
The post on icebreakers that don’t make people cringe has more on choosing questions that work for a mixed group.
After the day: eight weeks of coffee pairs
The offsite gets people talking; a coffee-pair programme keeps it going. Every Monday for eight weeks, everyone is paired with one colleague for a 20-minute call.
Paste the 30 names into the Random Pair Generator, tick Avoid repeat pairs in Options, and click Make pairs. You get 15 pairs. With the option on, the tool tries 30 random pairings and keeps the one with the most new pairs, then adds this week’s pairs to a memory in your browser. The memory holds the last 200 pairs, which is more than 13 weeks of 15, so it covers the whole programme.
Simulating the eight weeks shows what the option is worth:
| Week | Repeat pairs, option off | Repeat pairs, option on |
|---|---|---|
| 2 | about 0.5 | 0 |
| 4 | about 1.5 | 0 |
| 6 | about 2.4 | about 0.2 |
| 8 | about 3.2 | about 0.6 |
| Total, weeks 2–8 | about 13.6 | about 1.2 |
Without it, around 14 of the 105 coffees after week one are people meeting for the second time. With it, about one. Pairing all 30 people with everyone else would take 29 weeks, so eight weeks is nowhere near exhausting the 435 possible pairs; the repeats you avoid are pure chance.
Two limits worth knowing:
- The two tools don’t share memory. The pair generator doesn’t know who shared a breakout room on the day. Each person met about six colleagues in their rooms, so in week one each coffee pair has about a 1 in 5 chance of being two people who already talked at the offsite, which works out at about three of the 15 pairs. Usually a fine reunion; if you’d rather avoid it, there’s no switch for it.
- Odd numbers make a trio. If 31 people sign up, the last pair becomes a group of three. All three partnerships inside a trio go into the memory, so a trio counts as three pairs met, not one.
The post on pairing without repeat partners goes deeper into the pair generator by itself.
Whose laptop?
Everything in this plan is stored in the browser that ran it: the name lists, the breakout memory, the coffee memory. Nothing is uploaded and nothing syncs. If the facilitator on the day isn’t the person who’ll run the Monday coffee pairs, that’s fine; those memories are separate anyway. But the three breakout rounds must all be run from the same browser, or the avoid option is working blind.
The plan in one table
| When | Step | Tool | Settings |
|---|---|---|---|
| Day before | Paste names, clear old pairings if wanted | Breakout Room Assigner | Clear history |
| 10:30, 11:30, 14:00 | Assign rooms | Breakout Room Assigner | By size per room, 3, Avoid repeat pairings ticked |
| Each round | One opening question to all rooms | Random Icebreaker Picker | Team, names left empty |
| Mondays, 8 weeks | Coffee pairs | Random Pair Generator | Avoid repeat pairs ticked |
What it won’t fix
A random room assignment spreads people around; it doesn’t make anyone talk. Rooms of three with one quiet person and one person who is also quiet will have a slow first five minutes, and the icebreaker is there to cover exactly that. Coffee pairs only work if people book the calls, which is a reminder in the calendar invite, not a setting in a tool.
What the tools can do is take away the two things that most often spoil the mixing: the same faces in every room, and the organiser’s own instincts about who should meet whom. Small rooms, the avoid option ticked every round, one browser running the day. That gets most of the way.