I have watched companies lose customers for the wrong reason three times in my career. None of them was pricing, product or competition. Each was one person leaving.
The mechanism
When the knowledge of a process lives in one person's head, you do not have a documentation problem. You have a process that depends on a person, and people leave.
You find out at the worst possible moment: a campaign, a client scaling up, or a seasonal peak. The person is not there, nobody knows why it worked that way, and the thing that worked for three years stops working.
Reconstructing it almost always takes far longer than writing it down would have.
Why it does not happen
It is not lack of will. It is three concrete reasons.
Tacit knowledge is easier to use than to write down. For someone who has the process in their head, explaining it to a colleague costs time they do not have today. Writing it down costs time they do not have. The list of reasons it did not happen is full of people who never had two free hours in three months.
There is never a good moment. The right moment was three months ago. Today there is a campaign, tomorrow there is a client. So it does not happen.
Nobody warns you. The file was never written, so nobody notices it is missing. The only time you find out is when you need it.
What actually works
I have tried three approaches. Only one held.
The wrong format is the main cause. I have watched documentation attempts fail because they asked people to "write the procedures manual". Nobody writes a procedures manual, not out of laziness: because a manual is an enormous job and it tells nobody what to do on Monday morning.
What worked was the opposite: a sheet, one row per decision, not one page per process. Column A: what we do. Column B: why. Column C: who decides. Column D: when we last changed it. Four columns, one row each.
Four columns, and column B is the one that matters. A process without a "why" is a process somebody will redo differently the moment that person is gone.
Do not delegate the writing to someone who does not know the process. The writer has always been someone who already knew. The real cost is not writing: it is explaining, and explaining needs protected time, not "if there's time".
Let it happen as a side effect. The opposite of what I expected worked: documentation grew when we started recording meetings and putting the decisions made inside them. Not because anyone became more disciplined, but because that was already the moment when we were talking.
The metric I use
One thing only: how many decisions from the last six months cannot be found anywhere.
Not how many procedures are written: how many decisions are traceable. Once measured, the number is always higher than the company thought, and that finding is usually enough to decide to intervene.
The second metric is more uncomfortable: how many times in the last three months someone asked a person "how did we used to do this?". Those questions are the signatures of knowledge about to walk out.