Tool
Random Lunch Partner Picker
Pair people across teams — never same team, never a repeat from last round.
Random Lunch Partner Picker
6 people across 3 teams, avoiding 0 previous pairings.
Pair colleagues under two constraints at once: never the same team, and never a repeat of last round's partner. Paste the roster with teams, paste last round's pairs, generate.
Both constraints matter for the same reason — the whole point of a pairing programme is to connect people who would not otherwise talk, and the two things that defeat it are pairing people who already work together and pairing the same two people twice.
When the constraints cannot both be satisfied, the tool relaxes the cross-team rule and says so. It never silently ignores the history.
How the Random Lunch Partner Picker works — and why it's fair
Each attempt shuffles the roster uniformly and pairs it positionally, then tests every pair against both rules: are they on the same team, and were they paired last round? A single violation rejects the whole arrangement and the search shuffles again, up to three thousand attempts.
The two constraints are not equal in importance, and the fallback order encodes that. If no arrangement satisfies both, the tool retries with the cross-team rule dropped while keeping the history rule — because pairing two people from the same team is a mildly worse outcome, and pairing the same two people twice in a row is the failure that makes people stop participating.
The relaxation is reported on the result. That is the difference between a tool that degrades honestly and one that degrades quietly: an arrangement that broke the cross-team rule looks identical to one that did not, so the only way you can know is if the tool tells you.
History is supplied rather than remembered. You paste last round's pairs, which is more work than automatic storage and considerably more robust — the person running the programme is usually not on the same device or even the same browser from month to month, and a history that silently failed to load would be worse than no history at all.
The exhaustion case is real for small groups. Six people can be paired in fifteen distinct ways; after a handful of rounds excluding previous pairs, the combinations run out and the tool reports that no arrangement avoids every previous match. The honest response is to clear the history and start the rotation over, which the tool suggests.
Odd numbers leave one person unpaired, named explicitly in the output rather than dropped. That person should be the first to pair next round, which is bookkeeping the tool does not do for you.
When the Random Lunch Partner Picker is fair — and when it is not
What it does guarantee
- Every arrangement satisfying both rules is reachable, since each attempt starts from a fresh uniform shuffle.
- The history rule is never dropped, even when the cross-team rule has to be.
- Any relaxation is reported on the result rather than left invisible.
What it does not
- It does not track history itself. You supply last round's pairs, and a forgotten paste means an unconstrained draw.
- It only knows about the immediately previous round, not the whole history of the programme.
- Randomised restart gives up after three thousand attempts, so a very tight roster can be refused even when an arrangement exists.
Two worked examples
6 people in 3 teams of 2, no history
- The cross-team rule alone removes the 3 same-team pairings from consideration.
- 8 of the 15 possible arrangements satisfy it, so a valid one appears within a couple of attempts.
- Both rules are satisfied and no relaxation is reported.
The same 6 people, three rounds in
- Nine previous pairings are now excluded, plus the three same-team pairings.
- Very few arrangements remain, and the cross-team rule usually has to be relaxed.
- By round four the tool typically reports that no arrangement avoids every previous match — the group is exhausted and the history should be cleared.
Pairing programmes that actually run
Random-coffee and donut-style programmes are the direct case, and the reason they fail is almost always one of the two things this tool constrains: people get paired with their own team, or with the same person again, and after the second occurrence they stop signing up.
Onboarding programmes use it with a slightly different emphasis, pairing new starters across teams deliberately. Same mechanic, and the cross-team rule is doing the primary work rather than the secondary.
Cross-functional projects use it to seed conversations before the work starts, where the exhaustion behaviour is a feature — a small project group that has run out of new pairings has, by definition, connected everybody.
For pairing with no constraints at all, the pair generator handles the odd-number case explicitly and runs instantly. For splitting into groups larger than two, the team generator and group role assigner are the right tools.
Frequently asked questions
Why does it drop the cross-team rule rather than the history rule?
Because pairing two colleagues is mildly disappointing and pairing the same two people twice is the failure that makes people stop participating.
Does it remember previous rounds automatically?
No — you paste last round's pairs. Automatic storage would fail silently when the organiser changes device, which is worse than asking.
Why did it say no arrangement is possible?
Because the group has run out of new pairings. Six people have only fifteen possible pairs; after a few rounds excluding previous ones, there is nothing left. Clear the history and restart.
Does it consider the whole history or just last round?
Only what you paste. Paste more rounds to exclude more pairings, at the cost of exhausting the group faster.
What happens with an odd number of people?
One person is left unpaired and named explicitly. Pair them first next round — the tool does not track that for you.
Can I have more than one person per team?
Yes, any number. The team label is just a string, so teams can be any size.
What if someone has no team?
They are given a placeholder team of their own, so the cross-team rule never blocks them.
How do I know if a rule was relaxed?
The result says so explicitly, in bold, above the pairings. A silent relaxation would be indistinguishable from a clean result.
Can it pair three people instead of two?
No. For groups of three or more, the team generator splits a list into groups of any size.
Related tools
Random Pair Generator
Split a list into pairs — with an honest answer for what happens on an odd count.
Group Role Assigner
Split into groups, then assign roles inside each group — two levels, one draw.
Round-Robin Schedule Generator
Everyone plays everyone once, in the fewest rounds — by the circle method.