Quality documentation: getting the right revision into the operator's hands

Documents instruct, records prove — and the controls differ

A drawing, specification, control plan, work instruction or inspection method tells someone what to do; it has a revision, an approver and a distribution list, and old versions must be withdrawn. A completed inspection sheet or certificate proves what was done; it is never revised, only retained and protected. Confusing the two produces two recognisable failures: instruction sets that are freely edited without approval, and completed results that somebody tidies up after the fact. Splitting the register in half, with different rules for each side, removes most of the ambiguity that document systems accumulate over time.

The uncontrolled copy is where the money is lost

Every plant generates them: a photocopy taped inside a cupboard door, a scan on a supervisor's phone, a laminated sheet from a previous product family, a marked-up print the setter trusts more than the official one. Each represents a frozen revision that will eventually diverge from the approved one. The countermeasures are practical rather than procedural — make the current version genuinely easier to reach than the copy, print with a visible revision and issue marking, physically remove superseded prints during changeovers, and audit the shop floor for paper rather than auditing the register in the office.

Approval is about competence, not signature count

Requiring several signatures on every instruction slows change and lowers the quality of review, because each approver assumes another has read it properly. Define, by document type, who must approve on technical grounds and who must approve on manufacturability grounds, and keep the list short. An inspection method that a metrology-competent person has not read is a liability regardless of how many managers signed it. Where a customer holds approval rights over certain documents — typically drawings, control plans and anything affecting a validated process — that constraint must be visible on the document itself, not buried in a contract.

Issuing a change is not the same as telling people about it

The transition is where documentation causes defects. A revision that appears silently in a folder leaves the previous version running on the night shift, half-finished batches built to two specifications, and material already cut to the old dimension. A managed change states the effective point, what happens to work in progress and existing stock, whether the change is retrospective, who briefs each shift, and what training or re-qualification is required before the new instruction can be followed. It also names someone responsible for confirming afterwards that the old version has actually disappeared from the floor.

Writing instructions people will actually use

Volume is the enemy of compliance. Instructions written to satisfy an auditor read as continuous prose, cover every conceivable circumstance and get ignored in favour of asking a colleague. Instructions written for the person doing the job are short, sequenced, and heavy on images of the correct and incorrect condition — particularly for anything judged by eye. The practical test is to hand the document to a competent person who has not run that job and watch what they do. Whatever they ask about, or get wrong, is the part of the document that needs rewriting.

Frequently asked questions

Does moving documentation onto screens at the line solve the revision problem?

It solves distribution and removes the photocopy, provided the screen shows the current version by default and cannot display a cached older one. It does not solve approval discipline, change notification, or the habit of working from memory. It also introduces a new failure: when the network or the terminal is down, people revert to whatever paper they can find, so the plan for that outage needs to exist before it happens.

How do we handle red-lined drawings and temporary deviations at the line?

Treat a deviation as a controlled document in its own right, with a stated scope, an expiry, an approver and a reference to the parent drawing, and attach it physically to the affected job rather than to the drawing file. Hand-marked prints with no approval are the single most common source of parts made to a specification nobody authorised. If a mark-up survives more than a couple of runs, it is a change request, not a deviation.

Who should own the document register — quality or engineering?

Ownership of content sits with whoever is technically competent for that document type: engineering for drawings and specifications, manufacturing engineering for methods and setups, quality for inspection and control documents. Ownership of the register, the revision discipline and the withdrawal of superseded copies is an administrative function that works best in one place, usually quality, precisely because that team has a reason to check the shop floor rather than the folder.