What leaves when one person leaves

The arithmetic of a hub departure

On the team counted here, five of the six working pairs include the manager. If that person leaves, the grid goes from six pairs to one.

Seven people remain. They have between them a single working relationship — the two engineers who shared an on-call week — and twenty other possible pairs that have never existed. Every route between any other two of them went through somebody who is no longer there.

Nothing about this appears in a handover. Handovers transfer responsibilities: who owns what, what is in flight, what to watch. They do not transfer relationships, because relationships cannot be transferred. The new manager arrives and has to build five pairs from nothing, which takes two quarters, during which the team is seven people doing individual work.

What else goes

Four things, roughly in order of how long they take to notice.

• The map. One person knew what all eight were doing in enough detail to route a question to the right place. That routing now has to be done by search, and the search mostly does not happen.
• The reasons. Not the decisions — those are usually written somewhere — but why. Why the limit is three. Why that customer is handled differently. Why the obvious improvement was tried in April and abandoned. These live in the people who were there and are the first thing lost.
• The translations. On a team with pockets of private vocabulary, somebody is usually the person who can tell both sides what the other means. That role is never assigned and is always held by whoever has pairs on both sides.
• The unofficial ownership. Everybody owns something that is not in their title. Nobody knows what the departing person's version of that was until it stops being done.

Why the usual mitigations do not work

Documentation helps with the second item and nothing else, and it helps less than people expect, because the useful reasons are the ones nobody knew were worth writing down.

A long notice period helps if it is used to form pairs rather than to write documents. Four weeks in which the departing person works jointly with two colleagues on real things transfers more than four weeks of them writing a handover document that will be read once.

Hiring a replacement quickly does not help at all in the short term. It adds a person with zero pairs to a network that just lost most of its pairs.

What actually reduces the loss

Reduce the concentration before anybody is leaving. This is the whole case for the pairs that bypass the hub, and it is the only mitigation that has to be done in advance.

A team with twelve pairs spread evenly loses three or four when somebody goes. A team with six pairs concentrated on one person loses five. The difference is decided a year earlier, by whether anybody arranged work between people who were not the manager.

Using the notice period for pairs

Four weeks is usually spent on documents. It is better spent on work.

Take the two or three areas the departing person holds alone and put a colleague on each of them, jointly, for a real task with a date inside the notice period. Not shadowing, which transfers almost nothing, and not a walkthrough, which transfers the parts the departing person thinks are important.

Joint work transfers the rest: the shortcuts, the thing that always goes wrong, the reason the obvious approach was abandoned. It also leaves a colleague who has done the work rather than one who has read about it, which is the difference between a handover and a document.

Write the document too, and write it second.

The one thing to do on the day

When a departure is announced, before the handover document is started, take the pair count and mark which pairs are about to disappear.

It takes ten minutes and it changes what the notice period is used for. A team that can see that it is about to lose five of six relationships will spend the four weeks differently from one that thinks it is losing one person out of eight.