# An exhaustive procedure is not a compliant procedure. The cautious writer's reflex is to add: an edge case, a regulatory reminder, a precaution. By the end, the document covers everything and **guides nobody**. The GMP guide asks for the opposite: documents that are clear, unambiguous, and written to be followed. Two readers An operator who executes, an auditor who verifies. They are not looking for the same thing, they do not read at the same moment, and they do not have the same time. Writing one document for both means failing both. What stays in the procedure at the workstation, and the four contents that live elsewhere.At the workstationThe real sequenceThe criteriaWhat to do on deviationElsewhere, by referenceThe regulatory reminderThe justification of choicesThe very rare casesThe gesture, on video Confusion of readership is the most frequent cause of unreadable documentation. A procedure swells because it carries content that is not addressed to whoever executes it: the regulatory framing, the justification of choices, the revision history, the vanishingly rare cases. **All of that content is legitimate — elsewhere.** Leaving it in the body of the text means asking the operator to sort it out themselves, at the workstation, under time pressure. "If it isn't written in the procedure, we are not covered." That sentence drives documentation bloat, and it rests on a reasoning error. **What covers you is that the practice is controlled and demonstrable** — not that a sentence exists somewhere. Worse: a procedure your operators cannot fully apply exposes you more than a short one that holds. It creates a permanent gap between what is written and what is done, and that gap is exactly what an auditor looks for. ## Write from the task, not from the standard. ### 1\. Observe before writing A procedure is written at the workstation, not at a desk. One hour of observation reveals the real order of operations, the switching points, and the two or three places where the error actually happens. That material should structure the text — not the layout of the standard. ### 2\. Separate the stable from the variable The principle rarely changes; the parameter changes often. Mixed into one document, they force you to reopen the whole text for a single value. Separated, a revision touches only what moves — and the procedure stops ageing badly. ### 3\. One verb, one action, one actor The GMP guide asks for an imperative style. "Care should be taken to ensure that" indicates neither who acts, nor when, nor how you would observe it. "The operator checks X before Y" does. This writing discipline alone removes a large share of interpretation ambiguities. ### 4\. Test before approving Have the text executed by someone who did not write it, without helping them. Anything requiring a spoken explanation is an ambiguity to fix. The test costs half an hour and it beats three rounds of review in the approval loop. ## Operators' cognitive load is a quality parameter, not a question of comfort. A long text consumes attention before it is even applied. **And attention is a finite resource**: whatever a paragraph of superfluous caution consumes will not be available at the moment of the critical action. Writing short is not a stylistic convenience, it is a risk control decision — and it is argued as such in front of an auditor. ## Four contents to move out of the document. ### The regulatory reminder It reassures the writer and does not help the person executing. Its place is in the higher-level document, which links the requirement to the internal arrangement — once, not in every procedure. ### The justification of choices Why this threshold, why this interval: valuable, and it belongs in the validation file or the risk analysis. The operator needs the value, not its history. ### The very rare cases Handling inline a case that occurs once a year weighs down the reading three hundred times a year. A dedicated procedure, called by a cross-reference, serves both situations better. ### The gesture, when it can be shown Some know-how does not write down usefully. A short video sequence attached to the document grain conveys in thirty seconds what two pages describe badly — and updates just as fast. None of this content is deleted: it is **moved to where it serves**. That is the difference between lightening a system and impoverishing it, and it is what an auditor will check if the question arises. ## Go further. [ ### Why they age badly Over-revision or fossilisation. Read →](/en/blog/quality-procedures-why-they-age/)[ ### Remove without losing the trail Two hundred documents become twenty. Read →](/en/blog/documentation-simplification-auditable-traceability/)[ ### From 30% to 3% Deviations caused by doc and training. Read →](/en/blog/documentation-training-deviations-30-to-3/) ## Writing a procedure, plainly. What structure should you use? + Purpose and scope in a few lines, responsibilities as a table rather than paragraphs, the sequence of operations in the real order of execution, control points and acceptance criteria, what to do in case of deviation, cross-references. Everything else — regulatory framing, justifications, detailed history — attaches by reference. How long should it be? + No standard sets a length. The right criterion is functional: an operator must be able to find what they need without leaving their workstation for more than a few seconds. If the document has to be read in full to be usable, it is too long whatever its page count. Do you need a flowchart? + As soon as there are branches, yes: a flowchart makes visible at a glance what a text forces the reader to reconstruct mentally. It does not replace the text for criteria and values; it complements it for the path. How do you know a procedure is applied? + By going to look, and comparing the real order of operations with the written one. The gap is nearly always instructive: either the practice has drifted, or — more often — the text never described the reality of the workstation. Either way, observation costs less than the deviation that will eventually reveal it. ## Do your procedures describe what your teams actually do? The gap between the text and the workstation is measured in one day of observation. It often explains most of your deviations.