The events
Why coffee roulette fades by week six
Randomly paired half-hours start with high attendance and decay on a predictable curve. The mechanism of the decay says something useful about what pairs need.
The people this page is about, and which of them have ever worked together
- 01The team's managerTwo working pairs, both from assigned work rather than from any event
- 02The second engineer
- 03The engineer who joined in MarchAttended every event of the quarter and has no pair
- 04The designer
- 05The QA engineer
five people · 10 possible pairs · 2 of them have worked on something together · 2 have only been in the same meeting · 6 have never done either
Five of the eight, chosen because all five attended every social event of the quarter.
Used with a clear policy, transparent workforce monitoring can support a useful review of focus, attendance, project time, and recurring bottlenecks.
Teams can compare this approach with the practical guidance available from Kanban University.
A bot pairs two people at random each week and books them half an hour. The first round has near-full attendance, the second slightly less, and by about the sixth most of the invitations are being quietly moved and then quietly dropped. Six months later it is nominally still running and three people are still doing it.
This happens almost everywhere it is introduced, which suggests a mechanism rather than a failure of enthusiasm. The mechanism is worth understanding, because it is the same one that makes the alternatives work.
The first conversation is free
Two people who have never spoken have an easy first half hour. There is a script — what do you do, how long have you been here, where are you based, what are you working on — and it comfortably fills the time. Both people leave pleased.
Nothing has been exchanged that either of them needed, but the exchange had no cost either, because the script did the work.
The second one is not
The script is spent. The second conversation between the same two people, or the first with somebody in an adjacent role where the script runs out in four minutes, requires one of them to produce a reason to be talking. Reasons are expensive to invent. Most people, facing a blank thirty minutes with a colleague they have no business with, will move the meeting.
This is the whole decay. It is not that people stopped caring about the team. It is that the format supplies an occasion without supplying a subject, and after the introductions there is no subject.
Compare it with a review
A review roster pairs people in the same arbitrary way and does not decay, because every instance arrives with its subject attached. There is a piece of work. One person made it, the other has to look at it, and there is something specific to say or not say.
The cost of showing up is near zero because nothing has to be invented. The exchange has stakes, which is what makes it informative. And it recurs without anybody's enthusiasm being required, because it is attached to work that was happening anyway.
That contrast is the single most useful thing in this part of the site. A pairing mechanism needs a subject supplied from outside the pair. Coffee roulette supplies an occasion and asks the pair for a subject; everything that works supplies the subject and lets the occasion look after itself.
When random pairing does work
Two conditions, and both of them involve adding a subject.
- A standing question. The half hour has a topic that is the same every time and is about the work: show each other something you made this month, or walk through one thing that went wrong. Attendance holds because nobody has to invent anything.
- Or a genuine information gap. In an organisation of four hundred, random pairing across teams does something real, because the two people have no other route to each other and the first conversation is genuinely informative. Inside a team of eight, where everybody already sees everybody weekly, there is no gap for it to close.
Inside a small team the format is close to the worst available: maximum scheduling overhead, minimum new information, decaying attendance, and a residue of evidence that connecting the team does not work.
The decay curve is information
It is worth looking at who stops first, because the order is consistent and it says something about the team.
The people who drop out earliest are usually the busiest and the most senior, for whom thirty minutes with no agenda is the most expensive half hour in the week. The ones who persist longest are frequently the newest, who still have questions the script can hold.
That is the opposite of what the mechanism was for. It was introduced to connect the people who do not know each other, and after six weeks the only participants are the ones who were going to ask questions anyway, paired with each other.
An initiative whose survivors are the people who needed it least has selected against its own purpose.
What to do with the one already running
If it is dying, let it. The slow fade is worse than ending it, because the people still participating are carrying a thing that is nominally a team practice, and the ones who dropped out are carrying a small guilt about it.
Say that it did its job for the first few rounds and stop it. Then take the same half hour and attach it to something: a review, a demonstration of real work, a shared duty. The pairs that come out of that are fewer and they are the ones that show up on the count.