There is a folder in most firms that everyone knows about and almost nobody opens. It holds the process documents. Some are good. Some were written by someone who has since left. Between what is in that folder and what people do, there is often a wide gap.
The instinct when that gap is noticed is to write more, or to write it better. That instinct is wrong often enough to be worth naming. If a firm has more documented process than it follows, the problem is not documentation. It is adoption, and adoption is a design question rather than a writing one.
Why a good system stops being used
Written systems are rarely abandoned in a moment. They stop being used the way anything stops being used, which is one exception at a time.
The system is slower than the workaround. This is the biggest one by a distance. If the documented route takes eleven steps and the shortcut takes three, the shortcut wins, and it wins with people who are conscientious and busy rather than people who are careless. Anything that asks someone to do more work than the alternative is competing against the alternative every single day, and it will eventually lose.
It was written for a firm that no longer exists. The process was accurate when it was written. Then the software changed, or a role split in two, or a client type arrived that it was never designed for. Nobody updated it, because updating it was nobody's job. It is now slightly wrong, and slightly wrong is functionally the same as wrong, because a person who follows it and gets a bad result will not follow it twice.
It never had a named owner. A system with no owner has no route for correction. Somebody notices the step that no longer works, has nowhere to send that, and works around it instead. The workaround spreads faster than any document, because it spreads by demonstration.
It was handed over rather than built with. A process written for people rather than with them starts its life without anyone in the firm feeling responsible for it. It reads as an instruction from elsewhere, and the first time it is inconvenient it gets treated as one.
What the used ones have in common
The systems that survive in a firm tend to share four properties, and none of them is about the quality of the writing.
The documented way is the easiest way. Not the most correct, the easiest. This means removing steps rather than describing them better, or building the step into the tool so it happens by default. A template that opens pre-filled beats a page explaining what to fill in. A form that will not submit without the missing field beats a reminder to include it.
They live where the work happens. A process that lives in a folder someone has to go and find is already losing. The same process attached to the job, in the system people are already in, gets followed without anyone deciding to follow it.
They have a person's name on them. Not a department. One person who owns that process, who is the route for corrections, and who is expected to change it when it stops being true. This is the cheapest of the four and the most commonly skipped.
They get used in front of people. A system that the owner visibly works around is finished, whatever the document says. Consistency at the top is not a nice-to-have here; it is the entire signal about whether this is real.
The three-person test
The test we run takes one process the firm believes it has and asks three people, separately, to describe how that job gets done.
If the three descriptions match each other, the firm has a system. If they match each other but not the document, the document is out of date and the real system is fine, which is a five-minute fix. If the three descriptions do not match each other, there is no system, only three habits, and the document was never the thing holding it together.
That last case is the common one, and it matters because it changes what comes next. Writing a better document will not close it. Sitting with the three people and agreeing one route will.
What the good version looks like
In a firm where this works, new people are useful in days rather than months, because the way things are done is visible rather than remembered. Nobody has to ask permission to find out how something works.
The documents are often shorter than expected, and there are fewer of them. A firm with forty processes written down and nine of them true is in a worse position than a firm with nine that are all true, because the first one has taught everybody that the folder cannot be trusted.
Where we start
We start with the process that costs the most when it goes wrong. We run the three-person test on it first, then fix whichever of the four properties it is missing, and in most firms that is the first one.
A system is the version of the work that is easiest to do correctly, and it stays true because somebody's name is on keeping it that way.