A working method for producing a Standard Operating Procedure that people actually follow. Not the section list, the process of getting from an undocumented task to a controlled document.
The most common reason an SOP becomes unusable is that it covers too much. Someone sets out to document "receiving" and ends up with fourteen pages spanning unloading, inspection, put-away and the ERP transaction, none of which the person doing any one of those jobs will read.
Fix the boundary first, in one sentence, and be specific about both ends: where the process starts, and what condition the output is in when it ends. "Starts when the pallet is on the dock, ends when the material is in a bin with a printed label" is a boundary. "Receiving" is a topic.
Then find the process owner, meaning the person accountable for the outcome, not the person who happens to perform the task. You will need them twice: once for facts you cannot determine yourself, and once for approval. Identifying them at the start prevents a finished draft sitting unapproved for a month.
This is the step almost everyone skips, and skipping it is why so many procedures describe a process nobody recognizes.
Go and watch the task performed, start to finish, without interrupting. Take notes on what happens, not on what should happen. Pay particular attention to three things:
If the process is too slow to observe end to end, watch the parts with the most handoffs. Handoffs are where consistency breaks.
After watching, ask. Two questions produce more useful material than any structured interview:
One rule while you draft: never invent a value. Torque specs, tolerances, temperatures, dwell times and cure times either come from a released document or from the process owner in writing. If you do not have the number, leave a visible placeholder rather than a plausible guess. A wrong tolerance in a controlled document is worse than a missing one, because the missing one gets questioned and the wrong one gets followed.
Writing an SOP from the first section to the last is slow, because purpose and scope are much easier to write once you know what the procedure actually contains. Work in this order instead:
A procedure lives or dies on its numbered steps. Four habits do most of the work:
Start every step with a verb. "Measure the bore at three positions" is an instruction. "The bore should be measured" is a description, and descriptions are read as optional.
One action per step. If a step contains "and", check whether it is two steps. Combined steps are the ones that get half done, and they cannot be cited cleanly when something goes wrong.
Put the acceptance criteria in the step. Not in an appendix, not in a specification stored somewhere else. If the tolerance is plus or minus 0.05 mm, it belongs on the same line as the measurement, because that is where the person is looking.
Write the failure path immediately after the check. Every verification step needs a companion step stating what happens when it fails and who is told. A procedure that only covers the good outcome is abandoned the first time the outcome is bad.
Weak: The operator should ensure the part meets print requirements before packing.
Better: 5.3 VERIFY: bore diameter is 25.00 mm plus or minus 0.05 mm at three positions. Record each reading on FORM-014.
5.4 If any reading is out of tolerance, quarantine the part on the red rack and raise a nonconformance report. Do not rework without supervisor authorization.
Describe the process and get a complete SOP, purpose through records, that you can then correct against what you observed. Free, no account needed.
Generate an SOP freeHand the draft to someone who does not perform the task and ask them to follow it while the regular operator watches. Every place they stop, ask a question, or do the wrong thing is a defect in the document, not in the person.
This takes about an hour and it is the highest-value hour in the whole exercise. It is also the step that reveals the assumed knowledge you could not see yourself, because you learned the process by watching an expert.
Then route it for approval. At minimum the process owner, who is accountable for the content, and one approver independent of the author, usually quality. Record both names and dates in the document control block. Approval evidence is the first thing an auditor checks, and an unapproved procedure is not a controlled document at all.
Issuing a procedure is not the same as implementing one. Three things finish the job:
Train against it and record that you did. The training record is what connects the document to the people expected to follow it. Without it, an auditor has a procedure and no evidence anyone knows it exists.
Control the copies. Put the document number and revision in the footer of every page, and mark printed copies as uncontrolled. The copy taped to a machine is the one that will still be at revision 1 in two years.
Set a review trigger, not just an interval. An annual review is a floor. The triggers that matter are a process change, a nonconformance the procedure failed to prevent, and any corrective action that changes the method. A document only ever revised on its anniversary is usually a document nobody is using.
For a process you can observe in a single session, budget half a day: an hour watching, an hour interviewing, two hours drafting, an hour piloting. Approval routing usually takes longer than the writing. The estimates that go badly wrong are the ones that skip observation and try to write from a meeting.
Neither alone. The operator has the content and usually not the time or the document conventions. Quality has the conventions and usually not the detail. The workable split is that quality drafts from observation and the operator corrects it, which also means the operator has read it before it goes live.
For anything with a physical setup or a software screen, yes, and they earn their maintenance cost. Keep them small, number them to the step they belong to, and be aware that a screenshot is a revision trigger the moment the interface changes.
Then you have found something more valuable than a document gap. Decide which way is correct before writing, because an SOP that papers over a real difference between shifts will be followed by neither. If both ways are legitimate, write the branch explicitly.