An audit trail is a form of respect
A working paper is written for a stranger who intends to disagree with you and does not owe you the benefit of the doubt.
Contents
A working paper has a physical quality that takes a while to appreciate. Every figure on it carries a mark saying where it came from and what was done to it. A reference to the source. A note on what was tested and against what. The result of the test. The whole apparatus exists so that a person who was not there, and who has no reason to take your word for anything, can rebuild your conclusion without asking you a single question.
I prepared those before I understood what they were for. The first explanation anybody gives you is protective: keep a record, because somebody may come looking. That explanation is not wrong and I think it is the reason the practice travels so badly into other kinds of work.
It is an offer, not a shield
The defensive framing gets the direction backwards. A shield is aimed at somebody. What a trail actually says is: here is everything I did, in the order I did it, arranged so that checking me is cheap. Come and check. I will not take it personally.
That is not caution, it is regard. It treats the reader as a competent adult who is entitled to their own conclusion rather than as an audience who should be persuaded and left there. There is a version of work that arrives finished and smooth, with the working removed, and it is asking to be believed. Belief and verification are different transactions and only one of them is offered to everybody equally.
The second thing it does is quieter and I think it matters more. Without a trail, the only way to check somebody is to ask them to explain themselves. That is a social act, not a technical one, and it is expensive. A junior person will not do it to a senior one. A new person will not do it in their first month. Somebody who has already been difficult once this week will not do it a second time.
So the absence of a trail silently converts a technical question into an interpersonal one, and interpersonal questions do not get asked. The figure then goes unchecked, not because anybody decided it was fine, but because checking it had a social price and nobody wanted to pay it. A trail removes the price. It lets somebody disagree with your number without disagreeing with you.
The objection is never really time
I have heard the same objection everywhere, in accounting and later in delivery, and I no longer believe it.
Producing a trail while you are doing the work costs almost nothing. You are already holding the reference and the reason. Writing it down is a few seconds and the few seconds are inside the work. Producing one afterwards costs an enormous amount, because you have to reconstruct a state of mind you no longer have. So when somebody says there was no time, they are describing the second case, and the second case only exists because of a decision made during the first.
What is actually being declined, most of the time, is the invitation. A trail puts a stranger on equal footing with you and it removes the part of your authority that comes from being the only person who knows how the number was made. That is a real loss to a certain kind of position and it is never stated out loud, because there is no acceptable way to say it.
The two ways I got it wrong
I over-produced. For a stretch I wrote trails that recorded everything, in the sequence I happened to do it, with no concession to a reader at all, and I thought volume was rigour. It is not. A trail nobody can navigate is not an offer, it is a wall with paperwork on it, and it discourages checking exactly as effectively as having no trail while looking like the opposite. The good ones are indexed for the question the reader is likely to arrive with, which means you have to think about the reader, which is the whole point and the part I skipped.
The other one is worse and took longer to admit. Asking somebody to show their working can itself be a status move. I have done it, more than once, and what I actually meant on those occasions was that I did not believe them. Dressing that up as a process request let me avoid saying so, and let the other person be corrected by a procedure rather than by me. That is not respect. It is the same gesture used in the opposite direction, and it is harder to see because it wears the language of good practice.
What stays with me
The strangest thing about working papers is what they preserve.
A trail records the answer, but it also records where the person was unsure. You can see it. Somebody tested five items where three would have done, or wrote a longer note than the item deserved, and years later you can tell that they were uneasy about that line without being told. It is the only part of most work that outlives the person who did it and stays legible.
Nobody prepares them for that reason. It is just what happens when you write for somebody who is not you.