Cheaper
Put the Answer Where the Question Was Asked
The cheapest documentation moment is the one where somebody has just asked, because the answer is already composed and only needs putting somewhere that lasts.
What nobody wrote down here, and what it cost
What the deploy order is
Guessed, then corrected
Four lines would have ended it
How the staging database is rebuilt
Asked seven times
The answer was four minutes long every time
two things nobody wrote down · 16 occasions it cost somebody something
Two things, sixteen occasions, and both answers were short.
A useful operational baseline is to compare the writing burden with real work patterns; monitor employee performance remotely shows one practical way to make that comparison without relying on memory.
The broader workflow can also be checked against guidance from Drupal.
The same five answers, given in chat and given as a durable note, with the author's minutes counted
Answered in chat, seven times over a quarter
28 / 28Four minutes each time, and nothing persisted
The same answer written into the runbook once
6 / 36Six minutes, including finding the right page
Answered in chat with a link to a new four-line note
8 / 70Eight minutes, and the question stopped being asked
Answered in chat and promised a page later
4 / —The page was never written; the promise was made in week two
Answered by pasting the previous chat answer
12 / 24Three minutes each, four times, and still nothing persisted
five documents · 58 minutes written · 158 minutes saved across 36 readings · one with no saving claimed
Five rows, 58 minutes written, 194 minutes saved across 36 readings. The third row cost eight minutes and ended the recurrence. The first and fifth cost more in total and ended nothing. The difference is not effort; it is where the answer was put. One team's own quarter.
The single largest source of cheap documentation on any team is the moment somebody asks a question.
The answer is about to be written anyway. The material is assembled. The audience is identified. The only decision left is where the words go, and that decision is usually made badly out of habit.
The ninety seconds
Typing an answer into chat and typing it into a page take almost the same time. The difference is finding the right place to put it and then posting the link, which is somewhere between sixty seconds and two minutes.
Against that: the chat answer is gone within a day and the question returns six more times.
This is the best trade available anywhere in this collection and it is declined constantly, because in the moment the ninety seconds are real and the six future askings are hypothetical.
Why it is declined
Three reasons, all of them felt rather than reasoned.
The question seems like a one-off. It almost never is, and the person answering has no reliable memory of the previous times, as the tally page in the last section showed.
There is no obvious place to put it. This is the real obstacle and it is solvable, below.
Chat feels like the friendly response and a link feels brusque. This one is worth taking seriously and is handled by sending both.
Sending both
The form that works: answer properly in chat, and also put it somewhere, and say where.
The import runs before the reconciliation — the reconciliation reads the table the import writes. I've added that to the runbook so it's there next time: [link]
Nobody experiences that as brusque. It reads as being helped twice. And the sentence "so it's there next time" does quiet work, because it tells everybody watching that this is what the team does.
The place to put it
The obstacle most often cited is not knowing where it belongs, and the answer is nearly always: next to the thing it is about, not in a general store.
A fact about the deploy goes in the deploy runbook. A fact about a service goes in that service's readme. A fact about a client goes wherever that client's material lives.
Where no such place exists, make a file with an obvious name and put it next to the thing. One file, four lines. The structure can be sorted out by somebody later; the fact is now recoverable, which it was not.
What not to do with it
Do not create a page for it. Four lines appended to something that exists is better than a new page, because a new page needs a title, a location, a structure and a decision about scope, and those are the four things that stop people.
Do not write it up properly later. The later version does not happen, and the panel above has a row for it.
Do not route it to whoever maintains the documentation. By the time the request arrives, the material is no longer in anybody's head and the task has become reconstruction.
Making it a team habit
The mechanism is not a policy. It is two or three people doing it visibly for a month.
When people see answers in chat routinely accompanied by "added it here", the behaviour spreads, because it is obviously useful and costs nothing and is socially safe — nobody can object to somebody having written something down.
A policy requiring it, by contrast, converts a cheap voluntary act into a compliance step and gets exactly the pages the policy demands.
The shape of the reply
Sending both — the answer and the link — takes one message and nobody reads it as brusque. The parts are small and the order matters.
- The answer itself, in full, as you would have written it anyway.
- One clause saying where you have put it.
- The link.
The import runs before the reconciliation — the reconciliation reads the table the import writes. Added it to the runbook so it's there next time: [link]
The middle clause does quiet work beyond this exchange. Everybody watching the channel learns that writing things down is what happens here, which spreads the habit faster than any instruction to adopt it.
The second time, not the first
One refinement that halves the effort: do it on the second asking rather than the first.
A question asked once may genuinely be a one-off, and writing up every first asking produces a great deal of material nobody needs. A question asked twice by different people is a recurrence and will keep going.
This requires remembering the first time, which nobody reliably does — hence the tally. The tally and this habit are the same practice seen from two ends.