Technical documentation

The contents vary by scheme but the shape does not. A general description of the product and its intended use. Design and manufacturing drawings, schematics, and the descriptions needed to understand them. A list of the standards applied, and where a standard was applied only in part, a description of what was done instead. The risk assessment. The test reports. The instructions and safety information supplied to the user. And the declaration or certificate itself.

The risk assessment is the part most often missing and the hardest to reconstruct. It is a record of hazards identified and decisions taken — which risks were designed out, which were guarded, which were left to the instructions and why. Written at design time it is a few pages of engineering judgement. Written two years later by someone reading a schematic, it is a guess dressed as a record, and an assessor who has read a hundred of them can tell.

The other structural requirement is traceability. The file has to describe the product that is actually shipping. A file that describes revision A while the factory ships revision C is a nonconformity in itself, independent of whether revision C is safe, and it is the finding that most often turns a surveillance visit into a problem.

Retention is not optional and it is not the consultant’s to hold. In the European Union the file is kept available for ten years after the last unit was placed on the market, and where the manufacturer is outside the Union the authorised representative holds it or is able to produce it. The United Kingdom takes the same shape through its own responsible person. The Eurasian Economic Union places the obligation on the applicant established in a member state.

What this means in practice is that the file has to be a maintained document rather than a folder assembled for an application. Every design change, every supplier substitution, every label revision is an update to it. Manufacturers who treat it as a one-time deliverable discover the gap at the first market surveillance request, which arrives without warning and with a deadline attached.

User instructions and safety information are assessed as part of conformity, not as marketing collateral, and they are a common source of findings. Several regulations require them in the official language of every member state where the product is sold, and a translation that softens a safety warning to read better is a nonconformity rather than an editorial choice.

The residual risks identified in the risk assessment have to be addressed in the instructions, and the two documents have to agree. Where the risk assessment says a hazard is controlled by a warning and the manual carries no such warning, the file contradicts itself — which is a worse finding than either document being weak on its own, because it shows the process was not followed.

We specify what the destination will ask for, review what exists against that, identify the gaps, and structure the file so an assessor can navigate it. Where the underlying evidence exists but is scattered across engineering, purchasing and quality, that is most of the work and it is worth doing properly once.

Where the underlying evidence does not exist, that is the real project, and it belongs to the manufacturer. A risk assessment, a design rationale and manufacturing records are produced by the people who made the decisions. Identifying that in week one is uncomfortable; identifying it in month four, when a certification body asks and the launch date is fixed, is expensive.

The two rows that cannot be produced retrospectively are the two that decide whether a surveillance request is answered in a week or becomes a withdrawal. Both are cheap at design time and neither can be bought later.