Why Supplier Relationship Management Matters – and Why Half of Procurement Teams Still Cannot Say What It Earns
Supplier relationship management delivers value nobody measures. What the 2026 State of Flux research shows,...
READ ARTICLE →
A new refinery, LNG train or concentrator is assembled from tens of thousands of purchased items whose fitness for service nobody on the project has personally established. What the project has instead is documents: a material test report, a certificate of conformance, a welder qualification record, a factory acceptance test report, a hazardous-area certificate, a calibration certificate.
Those documents are produced by the party with the strongest commercial interest in them reading “pass”. They are reviewed, if at all, by an inspector working for the EPC contractor rather than the owner. They arrive as PDFs attached to a vendor data submittal. And when the project hands over, almost none of them are linked to the specific heat, lot or serial number of the item that was actually installed.
This is not a hypothetical weakness. It is the mechanism in nearly every documented case of fraudulent certification in industrial supply chains – and the regulators who have investigated those cases say so in plain language.
In February 2022 the former Director of Metallurgy at Bradken Inc.’s Tacoma foundry was sentenced to 30 months in prison and a $50,000 fine for falsifying test results on more than 240 productions of steel for US Navy submarine castings – about half the castings the foundry produced for the Navy, destined for 30 submarines, over 32 years. Bradken paid $10,896,924 to resolve the civil allegations. Management did not learn of the fraud until May 2017, when an employee found altered test cards.
What makes the case instructive is not the fraud. It is what the buyer then could not do.
The US Department of Energy’s follow-on notice, OE-2 2022-01 (April 2022), directed every DOE and NNSA site to determine whether it had procured Bradken items between 2008 and 2020, prioritising safety-class components, and to report within 90 days. The notice explains the difficulty: the falsified tests were destructive Charpy V-notch impact tests, and out-of-tolerance results of that kind “would likely have gone undetected in routine inspections.” Affected components might carry three different company names – Bradken, Atlas or AmeriCast – because of the ownership history. And, in the sentence that should be pinned above every vendor master: “Products are often sold through distributors, so direct purchase records may not exist.“
A government agency, knowing the supplier’s name and the exact test that was falsified, could not reliably tell its own sites what they had bought.
The pattern repeats. Two men were convicted in 2023–2024 of creating false calibration certificates for acoustic emissions testing equipment used at US nuclear plants: 15 false certificates, sent to plant owners 29 times inside NRC-required test reports, over eleven years from 2010 to 2021. They were found in 2021 – during an external audit, not by a customer and not at receipt.
Kobe Steel disclosed in October 2017 that non-conforming product had gone to 525 customers; by its own report of 6 March 2018 the figure was 688, across 19 locations. The company’s language is precise and worth copying staff “falsified inspection data as if it met customer specifications” and fabricated test data for products never measured. The US NRC’s Information Notice 2018-11 (24 September 2018) records that a Japanese government investigation pushed the timeline back from about 2007 to the 1970s – five decades – and Supplement 1 (1 December 2020) concludes that affected material “may still be lingering in the nuclear supply chain.”
And in the detail that ought to end any confidence in document review as a control: the NRC examined a certified material test report for a reactor coolant pump casing supplied to V.C. Summer Units 2 and 3. The review yielded no findings – on material from a supplier that had been falsifying certificates for fifty years.
Japan Steel Works disclosed in May 2022 that its M&E subsidiary had falsified, fabricated or misstated inspection results in 449 cases between 1998 and 2022, including 12 cases on valve and steam turbine casing materials and 20 on nuclear disc and head materials.
The industry’s answer to supplier risk is risk-based conformity assessment: inspection and test plans, hold and witness points, intervention levels scaled to criticality. IOGP’s JIP33 programme has standardized it properly. Each of the roughly 60 published specifications ships as four documents – the technical requirements specification (S-xxx), the procurement data sheet (S-xxx-D), the information requirements specification (S-xxx-L) and the quality requirements specification (S-xxx-Q), which defines inspection, testing and levels of intervention.
Read the quality specification and three things become visible.
First, the default is the weakest setting. IOGP S-705Q Version 2.0 (November 2025), covering welding of pressure equipment and piping, states at clause 4.2.1: “CAS level D is specified unless risk assessment indicates that a more stringent CAS level is required.” Level A is the highest extent of verification; level D is the lowest. At level D, the specification’s own intervention table shows no intervention at all for post-weld heat treatment inspection or production weld testing. That is the standard’s own stated default for an industrially proven solution – not an inference about cost-cutting, but the published starting point from which every negotiation begins.
Second, the control is deliberately relaxed exactly where sub-tier risk concentrates. Clause 3.1.1 notes that conformity assessment “may be undertaken at supplier/sub-supplier premises” – permissive, not a right. Clause 4.2.4 provides that for work outside the primary supplier location, hold points are discouraged in favor of surveillance.
Third, the control belongs to the contractor. JIP33’s own guidance states that the request for quotation carrying these quality requirements “is prepared by the EPC contractor“, and every intervention in the specification is assigned to “the purchaser or purchaser’s representative” – which on an EPC project is the contractor, not the owner who will operate the plant for thirty years.
The standard form reinforces this. Under the FIDIC Silver Book (2017 Second Edition), Sub-Clause 7.3 gives inspection rights, 7.4 testing and 7.5 defects and rejection, and the new Sub-Clause 4.9 requires the Contractor to run a quality management system and a compliance verification system – submitting the QMS within 28 days of commencement, auditing internally at least every six months and reporting within seven days. But the express access right attaches to the Site and to the Contractor. A right to stand in a third-tier sub-supplier’s shop is not granted by the standard form; it has to be created by flow-down, and flow-down is precisely what gets priced, negotiated and trimmed.
The Bureau of Safety and Environmental Enforcement’s QC-FIT Report #2014-01 (August 2014) on subsea connector and bolt failures is the clearest documented case of that trimming producing a consequence.
All 36 bolts fastening an H4 connector fractured on the Discoverer India on 18 December 2012, releasing around 432 barrels of synthetic-based mud. The root cause was hydrogen embrittlement, and the mechanism was a missing process step: a third-tier coating subcontractor had been working to the 1998 edition of ASTM B633 rather than the required 2007 edition, omitting the post-bake step. BSEE’s findings state that the manufacturer “qualified and audited only first tier suppliers”, that the nonconformance ran “over a four-year period“, and that it “was also not detected by (1) third party quality management certification groups such as API” – nor by Transocean or Chevron in their assessment of contractor qualifications.
The consequences ran across the industry: 10,982 replacement bolts supplied for 361 connectors worldwide, and Petrobras reporting 56 drilling rigs and 27 wells with subsea BOPs requiring bolt repair in Brazil.
Note what the paperwork said. A 2009 material test report for the bolts recorded hardness and tensile values within specification. The MTR was correct. The part still failed, because the defect was introduced three tiers down by a process the certificate never described. BSEE found no evidence the bolts were counterfeit — and recommended component-level tracking of bolts and fasteners through their service life.
A buyer checking a supplier’s credentials can establish that a licence exists. Establishing that one was taken away is another matter.
API’s Composite List is a real-time directory with eight license statuses, including Suspended, Canceled and Withdrawn, and the Monogram programme requirements (FM-002 Rev. 20, dated 17 June 2026) give API the right to conduct unannounced surveillance audits at least annually and to audit subcontractors at the licensee’s expense. But API publishes no register, count or bulletin of suspensions, withdrawals or revocations. Its “Latest Updates” page lists specifications withdrawn from the programme – a scope change, not enforcement against a company, and an easy thing to misread.
What API does publish is a Fraud Alerts page naming entities circulating forged or altered API certificates – as listed in October 2026, 24 entities under “False Certificates” and four under “False Claims.” The specifications involved are core capital-project hardware: API Spec 600 and 602 gate and globe valves, 6A and 6D wellhead and pipeline valves, 609 butterfly valves, 6L line pipe, and Spec Q1 quality systems. Three of the named entities are US-registered, in Houston, San Jose and Walnut, California. The alerts carry no individual dates, which limits how they can be cited — but the list exists, and most buyers have never looked at it.
ASME publishes no enforcement register, no revocation list and no counterfeit-stamp notices; the only route is inbound reporting, with the complainant required to name the certified supplier, cite the code paragraph and “provide objective evidence.” Accreditation bodies do publish sanctions – UKAS recorded 11 suspensions and 7 withdrawals in January–March 2026, 19 and 3 in April–June 6 and 2 in July–September – but these police certification bodies, not suppliers. There is no public record anywhere of a supplier losing an ISO 9001 certificate for cause.
IECEx is the structural counter-example and the model worth copying: every hazardous-area certificate of conformity is held centrally, published online free of charge, and carries a live status of Current, Suspended or Cancelled. A certificate number that cannot be found online is, in IECEx’s own words, not valid. That only helps a buyer whose supplier record stores the certificate number in a re-queryable field. A PDF in a submittal folder cannot be re-checked when the certificate is later suspended.
Where someone does test marked products systematically, the results are sobering. The EU’s EMC ADCO 12th joint cross-border market surveillance campaign (reported November 2020, fieldwork July 2019–March 2020, 97 products across 11 countries) found 60% overall non-compliance, with 45% failing on documentation against 25% failing emissions testing, and only 64% of declarations of conformity correct. That was CE-marked LED lighting, not process equipment – but the shape of the finding is the point: the paperwork failed more often than the product.
CFIHOS, IOGP’s JIP36, is the industry’s information handover specification; version 2.0 was announced as a major release completed in November 2025 (the specification document itself is dated October 2024). It is the right place to look for whether conformance evidence survives the EPC-to-operations transfer.
The vocabulary is there. The CORE reference data library defines 330 document types, including “material certificate” (synonyms: materials test report, mill test report), “material traceability report”, “nondestructive test report”, “heat treatment records”, “welder qualification certificate”, “calibration certificate” and “hazardous area certificate”.
The requirements are not. Across 1,050 class-to-document requirement rows covering 358 equipment and tag classes, a material certificate is required for seven classes – all electrical (motor control centres, transformers, control panels, voltage regulators). A non-destructive test report is required for none. A material traceability report is required for none. Heat treatment records: one class. A welder qualification certificate: one. An inspection and test plan: nine. A pressure test report: ten.
A spare part list is required for 321 classes.
No piping, valve, vessel, flange or fastener class requires a material certificate at handover under the CORE library. There is no “manufacturing record book” document type at all. The specification’s five handover events – detailed engineering, construction, commissioning, start-up, operations – contain no quality or conformance gate. And its scope clause expressly excludes “processes that govern how the contractor or the supplier/manufacturer creates and/or quality assures the engineering information.“
The capability has been demonstrated: version 2.0 added a bridge to JIP33 information requirements, 533 rows that are certificate-rich and carry handover flags – covering six tag classes, drawn from seven specifications, out of 875 tag classes and roughly 60 published specifications.
The traceability ambition is bounded in the same way. S-705Q defines three levels: Level I, unique identification tracked to a specific position on the equipment, applies only to skids and packaged equipment; Level II is traceability to a material certificate; Level III is a supplier declaration. For most of what a plant is built from, the best the standard contemplates is “traceable to a certificate” – not “traceable to the item in the pipe rack.”
None of this is technically hard, and one scheme proves it.
Since 1 July 1921 the National Board of Boiler and Pressure Vessel Inspectors has registered pressure-retaining items by requiring the manufacturer to submit the Manufacturer’s Data Report for permanent retention, within 30 days of certification, and to stamp the National Board registration number – the manufacturer’s sequential identifier for that specific item – on the nameplate above the ASME markings (NB-264 Rev. 19, approved 10 May 2025). The National Board reported in early 2022 that items had been registered more than 63 million times since 1921, with more than two million registrations in fiscal 2020 alone. Records are searchable by National Board number, manufacturer’s serial number, Canadian registration number or drawing number, and anyone can order a copy for a flat fee.
A permanent, centrally held, item-level record keyed to a number physically stamped on the installed item, retrievable decades later by anyone who can read the nameplate. The question is not whether it can be done. It is why nothing equivalent exists for valves, piping, flanges, fasteners, rotating equipment or electrical gear – the items that make up most of a plant and most of the API fraud alert list.
Three caveats belong in any serious treatment of this subject, and leaving them out invites contradiction from the same sources.
This has rarely caused failure. The NRC’s own position, in Regulatory Issue Summary 2015-08, is that “most CFSI-related events have not risen to the safety thresholds established by 10 CFR Part 21 and 10 CFR 50.9.” EPRI has stated that no known counterfeit item has failed in a safety-related application at an operating US plant. When the US GAO audited this properly in 1990 – finding nonconforming products at or suspected at least 72 of 113 US nuclear plants, and 8% of installed safety-related fasteners out of specification – it also found that only 2% were sufficiently out of specification to raise a performance concern. This is an assurance, cost and rework problem, not a body count.
Supplier error is not the dominant rework driver. In the standard peer-reviewed dataset (Hwang, Thomas, Haas and Caldas, Journal of Construction Engineering and Management, 2009, using 359 projects from the Construction Industry Institute database), the total field rework factor averages 4.9% of construction cost, of which vendor error or omission accounts for 0.3 points and vendor change 0.1 – against 1.5 points for design error and omission. The defensible claim is narrower and sharper than “suppliers cause rework”: a non-conforming item that is installed and never detected does not appear in a field rework statistic at all.
Nobody can tell you how common this is on projects like yours. There is no published incidence rate for counterfeit or fraudulently certified items on EPC capital projects – not in oil and gas, not in LNG, not in mining. The only denominator-backed figures anywhere come from the South Korean nuclear forgery scandal of 2012–2013 (277 of 22,000 certification documents forged across 20 operating reactors, and 2,010 of 218,000 for eight further units) and the US GAO’s 1990 audit. The EU’s flagship counterfeiting dataset contains no figures for machinery, electrical equipment, tools or spare parts; even the OECD and EUIPO’s dedicated study of dangerous counterfeits never quantifies electrical machinery. No credible source publishes supplier quality audit failure rates, and no benchmark exists for vendor documentation rejection or late delivery.
That absence is itself the finding. The one industry that spends the most money on this problem has produced two denominator-backed datasets in thirty-six years.
The practical conclusion is unglamorous: the conformance evidence is supplier data, and it has to be managed like supplier data rather than like correspondence.
That means deciding the conformity assessment level package by package rather than inheriting level D by default and knowing – before the purchase order – which interventions that choice removes. It means specifying item-level traceability where it matters and accepting certificate-level traceability where it does not, consciously rather than silently. It means recording certificate numbers in fields that can be re-queried against a live register, not filing PDFs. It means asking a prospective supplier the question the API fraud alert list makes unavoidable: does the license number on this certificate resolve, today, in the issuer’s own directory. And it means writing the conformance records into the information handover requirement, because the industry’s own handover standard does not ask for most of them.
That is the gap Dharas addresses: holding the supplier record, its certificates, their issuers, their numbers and their expiry dates as structured, query able data tied to the purchase order and the package – so that when a notice like DOE’s lands on a Tuesday morning naming a supplier and a twelve-year window, the answer to “what did we buy, and where is it installed” takes an afternoon rather than a year.
The alternative is the position the US Department of Energy found itself in: knowing exactly what had gone wrong, and unable to say where it had gone.
Supplier relationship management delivers value nobody measures. What the 2026 State of Flux research shows,...
READ ARTICLE →
Local content reporting is now third-party certified, physically verified and criminally sanctioned. Why the number...
READ ARTICLE →
Brazil's regulator has assessed R$376,968,025.02 in local content fines and collected R$256,586,930.28 of it across...
READ ARTICLE →