When the process lives in one person
The hardest documentation problem is not writing. It is extracting a procedure from someone who has done it for nine years and no longer notices half of what they do.
How to get a procedure out of the one person who knows it, when that person has no time and does not think it is complicated.
Why the expert leaves things out
Ask someone who has done a job for nine years to describe it and you will get six steps. Watch them do it and there are thirty. They are not withholding anything — the missing twenty-four have become automatic, and automatic actions are genuinely hard to report.
The steps that vanish are the ones that matter most. They are the checks, the exceptions, the “obviously you would look at whether it came from the other branch first”, and the small judgements that separate the work being done right from being done to the letter.
So the interview cannot be “tell me how you do this”. That question reliably returns the six steps, and a document built from them fails the first time someone hits case twenty-five.
Watch first, ask second
Sit with them while they do it and write down what happens, not what they say happens. You are not interviewing yet; you are taking a record.
Note every point where they hesitate, check something, open a second window, or make a decision. Those are the branches, and branches are where documents fail. A step performed without pause is usually genuinely simple; a pause almost always means a rule that has never been written down.
One pass of observation produces more usable material than an hour of discussion, and it costs the expert nothing beyond being watched.
Then ask about the exceptions, specifically
The productive questions are narrow and concrete. “What happens when the field is blank?” “When was the last time this went wrong?” “What would a new person get wrong here?” “Why did you check that before you did this?”
That last question is the most valuable one in the whole exercise. The answer is nearly always a rule the expert did not know they were following, and it is nearly always the reason the procedure works.
Avoid “is there anything else”. It returns nothing. Specific questions about specific moments return everything.
Have them correct a draft rather than write one
Asking a busy expert to write the SOP produces nothing for three weeks. Asking them to correct a wrong draft produces detailed corrections in twenty minutes, because reacting is far easier than composing, and because an error in front of someone is intolerable in a way a blank page is not.
So write the draft yourself from the observation, deliberately including the parts you are unsure about, and hand it over. Being slightly wrong is a feature here. The mistakes are what draw out the corrections.
This is also the step that gets you their agreement, which you will need when the document becomes the standard.
Test it with someone who does not know the job
Give the draft to a person who has never done the task and have them attempt it while the expert watches without helping. Not speaking is the hard part and it is the point of the exercise.
Every place the new person stops is a gap in the document. Every place the expert wants to interrupt is a step that was left out. You will find more in twenty minutes of this than in any amount of re-reading.
The version that survives this test is the one worth numbering and putting into circulation. The version that has only been reviewed by people who already know the job has not been tested at all.
Doing it before it is urgent
The reason this gets done at all is usually that the person is leaving, and a notice period is the worst possible time to attempt it. The work competes with a handover, and the expert has stopped caring.
The cheap version, done early, is one observation pass and one draft per procedure — a couple of hours, not a project. Start with the procedures where exactly one person can do the work and the business stops if they are away. That list is usually short, and it is the whole risk.
If several procedures qualify, do not write them all at once. Writing too many at one time is one of the reliable ways these projects die — covered in why SOP projects fail.
The blank SOP template is the structure to pour the result into, and the method guide covers what belongs in each section.
Editor's notes
Written by Free Biz Docs about the document above — not reader submissions.
The whole method assumes a cooperative expert
Observation, narrow questions, a draft to correct, a test with a newcomer — every step needs the person to be broadly willing. The post says the work is usually triggered by someone leaving and then, in the same section, that a notice period is the worst time to attempt it.
Both are true and together they describe the common case: the method is at its weakest exactly when it is most often reached for. That is an argument for doing it early rather than a technique for doing it late, and the post has no technique for doing it late.
The deliberately wrong draft can misfire
Handing over a draft with mistakes in it works because correcting is cheaper than composing, and it works on the assumption that the expert reads it as you having made an honest attempt.
Read by someone who already feels their knowledge is being extracted before they leave, the same draft reads as a test they are being set. If that is the relationship, say what you are doing and why before you hand anything over. The technique survives being explained; it does not survive being resented.
Watching someone work is not a neutral act
The observation pass is described as costing the expert nothing beyond being watched, which understates it. Being observed changes how people work, and in most workplaces it also carries a history — of appraisal, of time-and-motion study, of someone being replaced.
Ask first, say what the record is for and who will see it, and it is usually fine. Turn up with a notebook unannounced and you will get a performance rather than the procedure.
These documents are starting structures for you to build on, not professional or legal advice. Anything that will have legal or contractual effect, or that varies by jurisdiction, should be reviewed by someone qualified in the relevant place.
Common questions
- How do I document a process only one person knows?
- Watch them do it before you interview them, and write down what happens rather than what they say happens. Note every point where they hesitate, check something, or open a second window — those pauses are the unwritten rules, and they are what the six-step version leaves out.
- Why does an expert leave steps out when describing their job?
- Because after years the steps have become automatic, and automatic actions are genuinely hard to report. The ones that vanish are the checks, the exceptions and the small judgements — which is to say, the ones that separate doing the work right from doing it to the letter.
- What questions actually get the detail out?
- Narrow, concrete ones tied to specific moments: what happens when the field is blank, when did this last go wrong, what would a new person get wrong here, and why did you check that before doing this. That last one usually surfaces a rule they did not know they followed.
- How do I get a busy expert to write the SOP?
- Do not. Write a draft yourself from observation — deliberately including the parts you are unsure about — and ask them to correct it. Reacting takes twenty minutes where composing takes three weeks, and being slightly wrong is what draws the corrections out.
- How do I know the document is good enough?
- Give it to someone who has never done the task and have them attempt it while the expert watches without helping. Every place the new person stops is a gap; every place the expert wants to interrupt is a missing step. A document reviewed only by people who know the job has not been tested.