Ask around most design offices and you will hear the technical file described as "the paperwork for the notified body" or "the folder we assembled for certification". Both descriptions are wrong, and the second is dangerous. Under the EU machinery, EMC, low-voltage and pressure-equipment regimes, the technical file is the manufacturer's evidence that the product meets the applicable essential requirements. It is the reasoning behind the CE mark, not a by-product of it.

I have opened technical files that were immaculate on the day of assessment and factually wrong within eighteen months. The risk assessment still referenced a guard interlock replaced during a value-engineering round. The harmonised standard cited in the Declaration of Conformity had been superseded twice. Nobody committed a crime; everybody assumed somebody else owned the file. That assumption is where compliance quietly rots.

The legal position is unforgiving. When a market surveillance authority knocks, and under the EU Market Surveillance Regulation they can knock with little notice, the file must be assembled and presented within a reasonable period, often ten working days depending on the member state and directive. A file you cannot reconstruct quickly is, in practice, a file you do not have.

Harmonised standards: cited, superseded, forgotten

Every Declaration of Conformity I have reviewed cites harmonised standards: EN ISO 12100 for machinery safety, EN 60204-1 for electrical equipment of machines, the EN 61000 series for EMC, and EN 61508-derived standards where functional safety applies. Citing them is easy. Tracking what happens to them afterwards is the part manufacturers neglect.

Harmonised standards are revised, withdrawn or replaced in the Official Journal on a rolling basis. A new edition of EN 60204-1 changes stop-category requirements, wiring colour conventions and verification clauses. If your declaration cites the 2007 edition and your design documentation was validated against it, you may continue under the presumption of conformity for a transition window, but that window closes, and the file must show you know it exists.

The failure mode is subtle: the design is still safe, but the evidence chain no longer supports the declaration, and the presumption of conformity evaporates on a technicality. Practically, this means the file needs a standards register: each cited standard, its edition, its status in the Official Journal, the date of the last check, and the named person who checked it. I ask for this register first when auditing a technical file. Where it does not exist, I know within ten minutes that the declaration was written once, laminated, and never read again.

Design changes that never reach the file

Engineering change is where technical files age fastest. Consider a typical industrial machine over a seven-year production life: a servo drive is substituted because the original went end-of-life, a light curtain replaces a physical guard to improve cycle time, the control software gains a new safety function, and a cheaper supplier provides the emergency-stop rope switch. Each change can alter the hazard analysis, the standard applicability, and the validation evidence.

Under the machinery framework, a change that affects compliance, and virtually any safety-related change does, requires the risk assessment to be revisited and the file updated. In my experience the engineering change process and the technical file live in separate buildings, sometimes literally. The change board approves the light-curtain substitution; the file still contains photographs and calculations for the interlocked guard.

If an incident occurs, the authority will compare the machine as built against the machine as documented, and the gap becomes the story. The fix is procedural and cheap: a single field on the change request form asking whether the change affects the CE technical file, with a mandatory yes/no and a named reviewer. Route every yes to the file owner, and review the routing quarterly for missed changes. This is not bureaucracy; it is the difference between a file that describes your product and one that describes a product you used to make.

The file only defends the product if it still matches what the change board actually approved on the floor.
The file only defends the product if it still matches what the change board actually approved on the floor.

Post-market obligations: where the file meets reality

Post-market surveillance is no longer a polite phrase. The Market Surveillance Regulation obliges economic operators to cooperate with authorities, and product-safety legislation increasingly expects manufacturers to monitor field performance and act on it. Complaints, warranty returns, incident reports and service findings are all inputs, and they all flow back into the file's risk assessment.

The concrete mechanism is a feedback loop with teeth. Every field incident involving a safety function, whether a guard defeat, an emergency-stop failure or an unexpected start-up, triggers a review of whether the original hazard identification missed something. If the risk assessment assumed operators would never reach over a fence, and service reports show they do, the assumption is falsified and the assessment must be revised. Measure this by audit trail: for each significant field complaint, the file should show what was reviewed, what was concluded, and what changed.

Serious incidents carry reporting duties under the relevant directive. Aged files fail here in the cruellest way: the organisation wants to report accurately, but the file cannot tell them what the machine was supposed to do, so the incident analysis starts from guesswork. Post-market obligations punish the unmaintained file twice, once as non-conformity and once as incapacity.

File structure that survives ten years

A technical file only its author can navigate is worthless once that author leaves. I favour a structure that mirrors the conformity route: product description and variants, applicable legislation and standards register, hazard identification and risk assessment using ISO 12100's three-step risk-reduction process, calculations and test reports, electrical and pneumatic schematics as validated, software documentation including version and safety functions, instructions for use as shipped, and the declaration with its supporting traceability.

Two disciplines keep it honest. First, version control with dates: every drawing, report and assessment carries a revision, and the file carries a change log explaining what entered or left and why. Second, the shipped-configuration link: the file must answer what serial number 412 looked like when it left the factory four years ago. Where variants exist, a build-standard matrix connecting serial ranges to file revisions does the job. Without it, you are defending a product you cannot define.

Retention matters too. Most directives require the file to be kept for ten years after the last unit was placed on the market, not ten years from certification. Product lines that run long outlive their documentation. I have seen companies exit a product and shred the file within five years, discovering only later that machines sold in year nine still carry obligations.

The notified body relationship and declared scope

Where a conformity assessment module involves a notified body, such as machinery Annex IV categories, some pressure equipment or ATEX, the certificate, its annexes and its conditions form part of the file's spine. These certificates state a scope: product variants, standards applied, sometimes specific software versions. Drift outside that scope invalidates the certificate quietly, and the CE mark affixed on its strength becomes unsupported.

Check the annexes annually against the current product range. Compare the variant list in the certificate against the build-standard matrix. Verify the software version recorded at assessment against the version shipping today; a firmware update that touches a safety function can take the product outside the assessed configuration. Where the directive allows self-declaration for some variants and requires assessment for others, draw the boundary explicitly in the file, because sales teams will not draw it for you.

The declaration itself deserves a scheduled review. Authorised representatives change, addresses move, cited standards are superseded, and the signatory may have left years ago. A declaration signed by someone who departed in a reorganisation is a small embarrassment; a declaration citing a withdrawn standard is a finding; a declaration whose underlying file no longer matches the product is a market-access problem.

A file you cannot reconstruct within the authority's notice period is, in practice, a file you do not have.

Making upkeep someone's job

Everything above collapses into one governance question: who owns the file, with what authority and on what cadence? My answer is a named file owner, usually the technical responsible person, with a mandatory annual review that checks the standards register, the change log against engineering change records, field feedback disposition, and declaration accuracy. The review produces a dated record. An unreviewed file is presumed stale, and in my experience that presumption is correct more often than not.

The annual review should also rehearse assembly: can the complete file, or the sections a market surveillance authority may request, be produced within the notice period? If the answer requires a fortnight of archaeology, treat it as a non-conformity now rather than a crisis later. Authority requests, customer audits and incident investigations all arrive unannounced, and the file's only real protection is being ready before any of them do.

Annual technical file review

  1. 01Standards register checkEvery cited standard confirmed current in the Official Journal, with date and checker.
  2. 02Change log reconciliationEngineering change records compared against file revisions; missed changes closed out.
  3. 03Field feedback dispositionComplaints and incidents traced to risk-assessment conclusions and actions.
  4. 04Certificate scope checkNotified body annexes compared against the build-standard matrix and shipping software.
  5. 05Assembly rehearsalConfirm the authority-requestable sections can be produced within the notice period.
The review cadence that keeps a declaration defensible between certification and the end of the retention period.

Two files, one product

Assembled once

  • Declaration cites a superseded standard edition
  • Risk assessment describes a guard replaced years ago
  • No owner; the compiler left the company
  • Reconstruction takes weeks of archaeology

Maintained as evidence

  • Standards register checked by a named person on a cadence
  • Change requests route safety changes to the file owner
  • Field incidents show review, conclusion and action
  • Authority-requestable sections ready within the notice period
The practical difference between a file maintained as living evidence and one assembled once for assessment.