Independent validation of NDT reports you are being asked to rely on

Validation starts where the examination ended

A report validation begins after the examination is finished, the crew has demobilised, and the document is sitting in somebody's inbox waiting to be accepted, paid for, or argued about. No probe goes back on the weld. What gets examined is the record: the procedure it invokes, the people who signed it, the equipment it names, the technique it describes, the criteria it cites, and the disposition it reaches. The output is a written opinion from an ASNT Level III on whether that record supports the conclusion printed on its front page.

This is a different product from inspection, sold to different people for different reasons. An inspection company sells you a finding. A validation sells you an answer about somebody else's finding. That distinction is commercial as well as technical: the validating party has no stake in the original result and no scope of its own to defend. It also changes how the work is done. A validator reads for internal contradiction, missing linkage and code compliance — not for flaws in steel.

The service exists because the alternative is often unavailable. The vessel is closed, the line is buried, the coating is back on, the fabricator has been paid and demobilised, the ship has sailed. When re-examination is impossible, disproportionate, or too slow for the decision in front of you, the record is the only evidence that exists — and somebody has to be able to say whether it is worth anything. The full trade-off is set out in third-party review versus re-inspection.

Six triggers, and each one asks a different question

The asset owner's trigger is distrust plus impossibility. A contractor has handed over a report that clears a weld, a shell course or a run of supports, something about it reads thin, and the asset is back in service. The question is not whether there was a flaw — it is whether this document justifies releasing retention and closing the work order. A validation answers that from the paper, and does it in days rather than waiting for an outage window.

The EPC or fabricator's trigger runs the other way. Their client's reviewer has rejected a report, and the fabricator believes the rejection is wrong: the criteria applied were never in the contract, or the reviewer misread a technique that the code permits. Here the buyer wants the report defended. An honest validator will sometimes tell them it cannot be, and that is exactly why the opinion carries weight when it goes back across the table.

Four more triggers behave differently again. An insurer or underwriter assessing a claim whose outcome turns on an inspection record needs to know whether that record is evidence or noise. A purchaser in due diligence on a plant, vessel or fleet is buying the inspection history along with the steel. A party in litigation or arbitration needs a record tested before it is relied on. And a Level III who has inherited a programme needs to know what the previous regime actually produced, which is closer to a programme audit run backwards through the files.

The seven checks a validation runs

Procedure compliance comes first. The report must invoke a procedure by number and revision, that procedure must cover the configuration and thickness actually examined, and it must have been qualified or demonstrated where the referencing code requires it. Personnel qualification comes second, and it is checked against the date the work was done. A certificate that is valid today proves nothing about a scan performed eighteen months ago under a certification that had already lapsed at the time.

Third is equipment and calibration traceability: the instrument identified by serial number, the calibration status current at the time of examination, and a chain back to a recognised standard rather than a sticker with a date on it. Fourth is technique adequacy for the flaw type expected. A shear-wave setup optimised for planar weld defects will not reliably characterise volumetric porosity, and a technique that cannot orient to the expected flaw is inadequate regardless of how well it was executed on the day.

Fifth, the acceptance criteria must be the right ones, stated with code and edition, and taken from the construction or in-service code rather than from Section V, which prescribes methods rather than accept or reject limits. Sixth, the results must have been correctly interpreted against those criteria. Seventh, the report must be complete against what the governing code requires it to contain — the element list is set out in full on what makes an NDT report defensible.

Qualification is tested on the date of the work, not the date of the review

In the ASME world, NDT personnel are certified by their employer, not by an outside body. ASME BPVC Section V requires qualification in accordance with the employer's written practice, and that written practice is built on ASNT SNT-TC-1A or ANSI/ASNT CP-189. In the 2023 edition of Section V, the personnel requirements that had lived in Mandatory Appendices III and IV were moved into T-120(e). A validation therefore starts by asking to see the written practice, not just the certificate.

The 2020 edition of SNT-TC-1A sets recertification for all three levels at five-year intervals, near-vision acuity examination annually — Jaeger Number 2 or equivalent at a distance of not less than 12 in. (30.5 cm) — and colour contrast differentiation at initial certification and at five-year intervals after. Three separate clocks run on every technician. The finding that turns up repeatedly is a certification file that is complete on the day it was scanned and short a vision record on the day of the scan.

Outside ASME, the model changes and so does the check. Marine work under IACS UR Z17 expects the responsible supervisor for thickness measurement to hold a recognised national or international qualification such as ISO 9712 Level II, with operators certified to Level I and holding adequate knowledge of ship structures. ISO 9712 is third-party certification, so the evidence is a certificate from a certification body rather than an employer's internal file — a different document, tested a different way.

Three verdicts, and the third is the one people misread

Verdict one: sound. The examination described was adequate for the flaw type expected, the criteria applied were the contracted ones, the interpretation follows from the data, and the record proves it. The disposition stands. This verdict is written to be usable — it names what was checked and what evidence supported each check, so the party who commissioned it can put it straight in front of the party who doubted the report.

Verdict two: deficient in a way that changes the disposition. The accept or reject call is not supported. This covers the wrong acceptance criteria, a technique incapable of detecting the flaw type the component is prone to, an indication recorded but evaluated against a limit that does not apply, or coverage that never reached the volume the code required. This verdict has commercial consequences and is written to survive being contested by a competent opponent.

Verdict three: deficient only in documentation. The examination was plausibly adequate, but the record does not prove it — a missing revision number, an unrecorded couplant, a calibration block never identified, a lighting check absent from a fluorescent examination. Nothing about the steel changed. What changed is that the report can no longer defend itself, and a client, auditor or opposing expert can reject it without ever disputing the technician's work. Most of what a validation finds sits here.

A validation that confirms the report was right is a result, not a failure

Report validation is not a defect-finding exercise pointed at contractors. A significant share of validations conclude that the original report was correct, adequately documented and properly disposed. That outcome is worth paying for, because the buyer did not purchase a criticism — they purchased a defensible answer, and "this report holds" is a defensible answer that ends the argument, closes the claim, or lets the purchase proceed on schedule.

This shapes how the work is run. A reviewer paid, structurally or psychologically, to find problems will find them. The discipline is to test each element against the code and record the outcome either way, including the elements that passed. A validation report that lists only failures is a complaint; one that lists what was tested, what passed and what did not is evidence, and only the second kind is any use in front of a tribunal or an underwriter.

It matters to the contractor on the other side too. A fabricator whose report has been rejected by a client's reviewer commissions validation precisely hoping for verdict one, and the value of getting it is that the same document would have said otherwise had the evidence pointed that way. An opinion that can only come back one way is worth nothing to either party, and everyone reading it knows that.

The deliverable, and what you can do with it

The deliverable is a signed technical opinion, element by element. Each of the seven checks appears as a heading with the evidence examined, the finding, and the code paragraph or contract clause the finding rests on. The verdict is stated in one line at the top. Where a deficiency is documentary rather than technical, it says so explicitly, because the remedy is different, the timescale is different, and the commercial consequence is different.

Attached to it is the evidence pack: the procedure revision reviewed, the certification records with their dates, the calibration certificates and their traceability chain, the technique sheets and calibration records, and the raw data, radiographs or scan files where they exist. The pack matters as much as the opinion, because it lets a third party — an auditor, an underwriter, a tribunal — re-run the reasoning without having to re-run the review from scratch.

Where the finding is that documentation is short, the deliverable states what would close each gap and whether it can still be closed after the fact. Some gaps can: a certification file can be produced, a calibration certificate retrieved, a technique sheet located in an archive. Some cannot: a lighting verification never performed cannot be reconstructed a year later. Saying which is which is the practical half of the report, and the half buyers act on first.

Independence: who can validate, and who cannot

The validator must not have produced the thing being validated. A firm that wrote the procedure, trained the technicians, or performed the original examination cannot review its own work and call the result independent. Where Atlantis has previously supported a client with procedure development or programme work, the validation is ring-fenced from that engagement or declined outright — the answer is only worth as much as the independence standing behind it.

Independence is not the only limit. An outside Level III operates inside a defined authority: reviewing, evaluating and rendering opinion, but not assuming the employer's certifying responsibility unless that has been formally delegated in the written practice. The boundaries are set out in what an outside NDT Level III can and cannot do, and they apply to validation work exactly as they apply to any other outside engagement.

Method scope is the third limit. A Level III validates in the methods they hold, and a validation touching UT, RT, MT and PT across the same weld package needs qualification across all four or a named reviewer per method. A report that quietly reviews a method outside the signer's scope has the same defect it is accusing the original report of having, and an opposing expert will find it in the first hour.

United States first, and worldwide because the codes are

The primary market is the United States, where ASME BPVC construction and in-service codes, API in-service inspection practice, and employer-based certification under SNT-TC-1A create the exact conditions that make report validation buyable: a document-driven regime where the paper is the proof. Most of the disputes that reach a validator are about whether the paper does its job, not about whether the metal underneath it is sound.

The service travels because the codes travel. A Section V examination performed in Jubail, Rotterdam or Singapore is examined against the same paragraphs it would be in Texas, and an ISO 9712 certification is verifiable from anywhere. Reviews are conducted on documents, so time zone and mobilisation stop being cost drivers — which is why validation reaches assets that no inspector would realistically be flown out to look at.

What changes across jurisdictions is the referencing code and the contract. A European fabrication under EN and ISO invokes different acceptance criteria and a different certification model from a US pressure vessel, and a marine survey under class rules invokes another again. The seven checks stay constant; the documents they are tested against are read out of the contract. Send the report and the governing code, and the scope defines itself — start here.

Request a consultation