Most organisations have documented procedures that a substantial share of their people do not follow. The gap is rarely defiance, and understanding why it opens is the useful part.
Procedures compete with the shortest route
People under time pressure take the path that completes the task. If the documented route is longer than an informal one, the informal one wins by default.
This happens even where everybody agrees the procedure is sensible. Agreement operates at the level of principle, while behaviour operates at the level of the next twenty minutes.
The practical consequence is that a procedure must be at least as convenient as the workaround it replaces. Otherwise it is a statement of intent rather than a control.
Written by people who do not do the work
Procedures are frequently drafted centrally, by someone who has observed the task rather than performed it. The document then describes an idealised version of the job.
Real work contains exceptions the author never encountered: the supplier who cannot supply a reference number, the system that times out, the customer whose case fits no category.
When the document has no route for those, staff invent one. The invented route is undocumented, inconsistent between people, and invisible until something goes wrong.
Where the procedure actually lives
A procedure stored in a document library is consulted during induction and then rarely again. Retention fades, and the version people remember drifts from the version that exists.
Procedures embedded in the tool are followed far more reliably, because they are not optional. A form that will not submit without a field enforces itself without supervision.
That is the underlying pattern: the more a rule is built into the system, the less it depends on anybody remembering it. Documentation is the weakest form of enforcement.
Why version control matters more than content
Organisations accumulate procedures faster than they retire them. Superseded versions survive in shared drives, printed copies and individual habits long after the official update.
People then follow different vintages of the same instruction, and the resulting inconsistency looks like carelessness rather than what it is: a document management failure.
Retiring the old version is therefore part of publishing the new one. A change that only adds a document has not replaced anything.
What audits actually reveal
An audit that compares practice against the document usually finds divergence. Treating every instance as non-compliance misses the more useful question of which direction is right.
Sometimes the practice is a genuine shortcut that should be corrected. Often it is an improvement developed by people closer to the work than the author was.
Organisations that harvest the second category improve their procedures continuously. Those that treat all divergence as error simply teach staff to hide it during the next audit.