How Documentation Discipline Prevents Nuclear Project Delays

In December 2025, the Finnish radiation and nuclear safety authority, STUK, reported that its review of the operating licence for the Posiva spent-fuel repository had been delayed for the third time. The reason it gave, in its own quarterly oversight statement, was that the hold-up came from deficiencies in the documentation, updates caused by plant modifications, and the scope of those updates (World Nuclear News, December 2025). Finland's ministry has since pushed the deadline to the end of June 2026.

Read that back. The repository is largely built. The engineering is done. What is not finished is a complete and consistent documentary record of how it was built and how it will keep that waste safely isolated for the next hundred thousand years. A facility designed to outlast every institution that created it is being held up by paperwork.

That is the uncomfortable shape of technical documentation in nuclear work. The record is supposed to be the straightforward part. It is routinely the part that slips.

The Posiva repository in Finland (Source: https://www.posiva.fi/)

Three regulators, one kind of failure

Posiva is not an isolated case, and it is worth being precise about what kind of case it is. This is a licensing-review delay, not a construction delay. The reactor-scale building work is largely complete; the hold-up is regulatory approval of the safety documentation. That distinction matters because the same category of problem also shows up at the construction end.

At Olkiluoto-3, the Finnish EPR that finally entered commercial operation in 2022 after a delay of around thirteen years, STUK named the verification of documentation for the plant's all-digital instrumentation and control system as one of the reasons for the hold-up (NucNet, April 2014). Documentation was one of several causes there, alongside first-of-a-kind design issues and supplier disputes, and an honest account says so. But it was named, by the regulator, on the record.

Nuclear Power Plant Olkiluoto 3 in Finland

In the United States, the pattern repeated in a different form. In February 2022 the chief executive of Southern Company told investors that incomplete and missing inspection records, a backlog of more than ten thousand of them, had added three to six months and close to a billion dollars to the Vogtle Unit 3 project in Georgia (World Nuclear News, February 2022). Those records were not a clerical afterthought. They feed the inspections, tests, analyses and acceptance criteria submissions that the US Nuclear Regulatory Commission requires before fuel load. The gap could not be closed by filing paperwork faster. It required physically re-inspecting work that had already been done.

Three projects, three regulators, three jurisdictions. The common thread is not that documents went missing entirely. It is that the documentary record could not be shown to be complete, traceable and authorised to the standard the work's safety significance demanded. In nuclear projects, that is enough to stop a schedule.

Vogtle Units 3 and 4, Georgia, U.S.

What 'important to nuclear safety' actually means

This is the category of problem that ISO 19443:2022 exists to prevent. The standard adapts ISO 9001 quality management to the nuclear supply chain, and its first organising idea is a classification: items and services important to nuclear safety, or ITNS.

ISO 19443 defines an ITNS item as any product, service or activity whose failure could lead to unjustified radiation exposure of people or the environment. The classification is set by the licensee, the organisation holding the operating licence, on the basis of the safety classification of the systems, structures and components involved. Suppliers do not decide this for themselves; they inherit it and act on it.

Why does a classification matter for documentation? Because the obligations that follow are demanding, and they apply only to what is marked ITNS. The standard requires that ITNS documents are traceable and authenticated, with superseded versions actively kept out of use. It requires review by someone who is both competent and authorised, which it treats as two separate tests. It requires that acceptance and inspection are carried out by people other than those who did the work, and that any non-conformance is segregated, notified to the customer, and held until the customer authorises what happens next. None of that is unusual in nuclear terms. But all of it is effort, and the classification is what decides which documents have to carry it. Until you have drawn that line, you cannot know which parts of the record must meet this standard and which need not. 

The graded approach: rigour matched to risk

Classifying everything to the highest level would be safe and unaffordable. Classifying everything to the lowest level would be cheap and indefensible. The standard's answer is the graded approach, and it is the second idea that makes ISO 19443 work in practice.

The graded approach scales the quality, documentation and monitoring requirements to the nuclear safety significance of the item or activity. ISO 19443 makes it a requirement, not a suggestion. It is a shall, not a should. The organisation has to apply it, on the basis of three factors: the ITNS requirements coming from the licensee, the complexity of the item or activity, and organisational considerations.

What the standard deliberately does not give you is a lookup table. There is no clause that says class A means two reviewers and class B means one. The organisation decides how deep the review goes, how tight the document control is, how many levels of authorisation apply. But it has to document that reasoning and be able to defend it at audit, and the document recording the reasoning is itself subject to the same traceability and authentication rules. You have freedom in the form, not in the proportionality. The rigour has to match the real safety significance of what you supply, and you have to be able to show your working.

In my experience, this is exactly where documentation packages come apart. I have seen records that were complete by every internal measure and still could not show where a single requirement had come from, who was authorised to approve it, or why it had been controlled at the level it was. That package is not wrong, exactly. It is undefendable, and in a nuclear licensing review, undefendable is hard to tell apart from incomplete.

Why getting classification right early protects the schedule

The reason all of this connects back to Posiva, Olkiluoto and Vogtle is timing. The graded approach is cheap to apply at the start of a project and expensive to apply retroactively. Decide the classification early and the documentation is built at the right level the first time, with traceability and authorisation captured as the work happens. Leave it until the licensing review, or until a customer audit, and you are reconstructing provenance for work that is already finished, sometimes for thousands of records at once. Vogtle had to re-inspect physical work to rebuild a documentary trail. That is the most expensive way to discover that a record was never properly controlled.

This is what compliance-by-design means rather than compliance-by-retrofit. Documentation structured to the applicable standard from the start is not slower. It is what stops the record from becoming the thing that delays the licence.

ITNS is necessary, not sufficient

ITNS classification and the graded approach decide how tightly the documentation process is controlled. They do not decide whether the document itself is any good. A manual can satisfy every control requirement in ISO 19443, with a clean authorisation trail and an immaculate version history, and still be incomplete, ambiguous, or unusable to the person who has to follow it. The standard that governs that, the content, structure and completeness of the instructions themselves, is EN IEC/IEEE 82079-1 and ISO 19443 does not replace it.

That is the distinction worth holding on to. ISO 19443 sits on top of an ISO 9001 quality management system and adds the nuclear requirements. EN IEC/IEEE 82079-1 sets the bar for the instructions as a product. ISO 19443 itself reaches beyond classification: it requires a nuclear safety culture in which raising a doubt about a document is expected rather than discouraged. Above all of it sits IAEA GSR Part 2, the governance requirement that makes documentation a controlled system rather than a one-off deliverable. ITNS gets the most attention because it is the most visibly nuclear. It is not the layer a reader of the manual ever sees.

Treat ITNS as the entry requirement, not the finish line. The projects that stay on schedule control the process and get the content right, and know these are two different obligations.

Where this leaves suppliers in the nuclear chain

For an organisation supplying products or services into a nuclear programme, ISO 19443 is increasingly the price of entry, and the documentation requirements are where most of the effort lands. Getting the ITNS classification and the graded approach right is less about writing more and more about writing at the right level, with the traceability, authorisation and version control the classification demands, captured as the work is done rather than assembled afterwards. For suppliers whose customers also require structured documentation standards such as S1000D, the same discipline carries across.

Documentation can itself become a non-conforming item. ISO 19443 treats counterfeit, fraudulent and suspect items as a category in their own right, and the same logic applies to a document with a false authorisation stamp or an untracked change. That is a separate subject, and one I will return to in a later article.

This is the discipline we are building with organisations operating in the nuclear sector: documentation that is classified correctly, controlled proportionately, and defensible when a regulator asks how it was made.

A nuclear project is not finished when it is built. It is finished when the record proves it.

CONTACT US