What to do instead
Pair on one real task this week
The core intervention, in full. Two named people, one piece of work that needs both, a date near enough to force a conversation, and nothing else.
The people this page is about, and which of them have ever worked together
- 01The senior engineerThe pair that came out of a shared on-call week
- 02The second engineer
- 03The designer
- 04The data analyst
- 05The product manager
five people · 10 possible pairs · 1 of them has worked on something together · 3 have only been in the same meeting · 6 have never done either
Five of the eight, with the manager left out so that what remains is what the team has without them.
For a distributed group, teams can view the available options to review where work travels, where it stalls, and whether responsibility is concentrated in one person.
An independent source worth comparing with this example is Cambridge Judge Business School.
Everything in this part of the site is a variation on one thing. Two people who have no working relationship are given a piece of work that needs both of them, with a date. This note is that thing in detail, because the details are where it usually goes wrong.
Choosing the pair
From the grid, not from intuition.
The best candidates are near-misses: two people who have been in the same meetings and never produced anything together. They already know the other exists and is reasonable, so the task does not also have to carry an introduction.
Second best is anybody with an empty row. The gain is largest there and the task has to be chosen more carefully, because the pair is cold.
Worst is a pair that already works well, which is what gets chosen when nobody looks at a grid, because those two are the obvious candidates for anything joint.
If there is a seam — a place where the work actually crosses between two people and has been crossing badly — that pair beats all of the above. The task is already justified and the benefit is immediate.
Choosing the task
Four properties, and it needs all four.
- It genuinely needs both of them. Not a task one person could do with the other watching. There has to be something each of them has that the other does not: knowledge, access, judgement, or a part of the system.
- Somebody wants the output. A made-up exercise produces a made-up relationship. If nothing happens when it is finished, both people will know within an hour that it is an exercise, and they will treat it as one.
- It is small. One to five days of elapsed time. Large tasks produce months of careful politeness before anybody says anything real; small ones produce the real conversation on the first morning because there is no time for anything else.
- It has a date close enough to force contact this week. This is the property that gets dropped and it is the one that makes the whole thing work. Two busy people with a vague joint objective will exchange two messages and let it drift, and both will conclude afterwards that the pairing was artificial.
How to assign it
Directly, to both of them, in the same message, naming the output and the date.
Not as an invitation. "It would be good if you two worked together on something" transfers the work of finding a task to the two people least able to find one, and nothing happens.
Not with the manager in it. If the manager is a participant, the exercise creates two hub pairs rather than one bypass pair, which is the opposite of the intention.
And without ceremony. Do not announce that this is a team-building exercise, do not call it a pairing initiative, and do not tell the two people that the point is their relationship. The work is the point; the relationship is the by-product, and naming it makes both people self-conscious and the task feel like a pretext, which it then becomes.
What to do while it runs
Nothing. Specifically, do not check in halfway, because a check-in routes the exchange back through the manager at exactly the moment the two of them would otherwise have had to resolve something themselves.
If it goes wrong, let it go wrong and ask afterwards. A pair that disagreed and settled it has transferred more than one that was rescued.
Afterwards
Two things, six weeks apart.
Immediately: give them the second task. This is the step that is almost always missed. One shared task produces a pair that will decay in a few months without an occasion; two, spaced a month apart, produce one that holds, because by the second one the preamble is gone and both of them know the exchange is cheap.
At six weeks: ask each of them, separately, whether they would now go to the other with a question. That is the outcome. Whether the task went well is a different question with a different answer, and a task that went badly can still produce a working pair — sometimes faster, because disagreement transfers more than agreement does.
When it does not work
Three cases, worth recognising so they are not repeated.
The seniority gap. If one of the two is markedly more senior, the task can turn into supervision, and what forms is a reviewing relationship rather than a pair. It is fixable by choosing work where the junior person holds something the senior one needs.
The separable task. If there is any way to cut it in half, two busy people will find it in the first hour. This is the commonest failure and it is a failure of task selection rather than of the people.
And the pretext. If either of them concludes the work exists so that they will bond, they will do it politely and nothing will transfer. This is why the exercise is not announced as what it is.
How many per quarter
Four. Any more and you are rearranging the team's work around the measurement, which is the failure mode the whole method is supposed to avoid.
Four deliberate pairings a quarter on a team of eight moves a grid from six pairs to nine or ten, fills the empty rows, and roughly doubles the number of relationships that do not run through the manager. That is as much as this is worth doing.