Conformance evidence attaches to an assembly as tested rather than to any component within it, which makes every value-engineering substitution a silent reopening of the evidence question, authorised several levels below where anyone would check.
A buyer is offered the same generic component from a different manufacturer at a better price and a shorter lead time. Same material family, same nominal dimensions, same declared function, an established brand with its own certification. The decision takes four minutes, sits comfortably inside a delegated authority, and is recorded as a saving. On any reasonable reading of the buyer's role, it is a good decision competently made.
It may also have destroyed the enterprise's ability to demonstrate a performance claim on which the asset's approval, insurance and warranty all depend — and nothing in the process will say so.
The reason is a property of how technical performance is established. High-consequence performance is not measured on a material in the abstract. It is measured by subjecting a complete assembly, built in a stated configuration, to a defined set of conditions, and observing what that assembly does. The result describes the specimen. It describes the components, their arrangement, their fixings and their interfaces, taken together. It does not describe any component in isolation, and it does not travel with a component into another arrangement.
Every element of the tested configuration is therefore load-bearing to the evidence, including the ones nobody regards as performance-critical. The fastener, the sealant, the adhesive, the backing board, the gasket, the spacing between fixings: these are not incidental to the certificate, they are part of what the certificate is about. Replace one, and the enterprise no longer holds evidence for what it has actually built. It holds evidence for something adjacent.
This is not an obscure technicality. It is the mechanism by which large, well-governed enterprises discover, years later and usually in front of someone else's lawyer, that the thing they installed is not the thing they can prove.
The Strategic Context
Substitution is not an aberration. It is a routine and largely necessary industrial act. Components go obsolete, suppliers consolidate, lead times move, a manufacturer reformulates a product without changing its part number, a contract requires cost to be taken out after award. Any delivery organisation of scale performs substitutions continuously, and a business that could not would be uncompetitive.
The problem is not the act. It is the authority attached to it. Delegated authority in almost every enterprise is calibrated to the price of what is being bought. The consequence of a substitution is calibrated to the value of the claim the assembly supports, and beyond that to the value of the asset the assembly exists to protect. Those two quantities are unrelated, and the gap between them is where the exposure sits. A component whose price justifies clearance by a buyer can carry a claim whose failure would be reported to a board.
There is a second structural feature. Conformance evidence frequently originates with a component manufacturer, who builds a configuration around their own product in order to demonstrate what it can do. That configuration reflects their commercial interest, not the project's design. A delivery team then adopts the certificate for a build that resembles it. The evidence and the asset were never the same object, and the divergence begins before any substitution occurs.
What Leaders Commonly Misread
The first misreading is that "like-for-like" means something evidential. It does not. It is a procurement concept meaning functionally and dimensionally interchangeable, and it is entirely sound within procurement. Functional equivalence carries no implication whatever about evidential equivalence. The phrase moves a decision from one category to another with no analysis at all, which is exactly what makes it dangerous: it sounds like a finding when it is a description.
The second is that better cannot be worse. A substitution that improves the component in every measurable respect still produces an untested configuration. Behaviour under extreme conditions is frequently non-monotonic — a stiffer, denser or thicker element can change how an assembly moves, seals or fails, and the direction is not reliably intuitable. The enterprise's position after an upgrade may be technically superior and evidentially empty.
The third is that change control already covers this. In most enterprises it does not, because change control fires on design change and a substitution is classified as a procurement event. The classification is made early, informally, and by someone with no reason to think it consequential. Once it is a procurement event it attracts a price review, not a verification review, and no one downstream will revisit it.
The fourth is that the substitution decision belongs to the enterprise at all. In several industries the final assembly is completed by the customer, the installer or the clinician, from components the enterprise supplied separately. Where that is true, the enterprise's evidence depends on a configuration decision made outside its walls by someone who has never seen the certificate.
Reframing the Issue
Stop treating a certificate as a property of a product. Treat it as a description of one specific object that once existed, under one specific set of conditions, and ask a single question of every claim: is the thing we built the thing that was tested?
That question has only three honest answers. Yes, and we hold the document that names our configuration. No, and we know it. Or we do not know — which, in any assurance regime worth the name, is the same answer as no.
Reframed this way, the governance failure is easy to see. The enterprise has distributed the authority to answer that question across hundreds of people who have never been told the question exists, and it has calibrated their authority to a variable that has nothing to do with the answer.
Where the Authority and the Consequence Separate
The configuration that forks quietly
Consider a hypothetical modular data centre programme, where the commercial logic rests on building the same module many times. The first module is fully evidenced: the assemblies are tested, the certificates name the configuration, the claim is clean. Over a multi-year rollout, a gasket supplier reformulates, a cable gland goes obsolete, a penetration seal is consolidated to a preferred vendor, and a fixing pattern is adjusted on the factory floor to suit a new jig.
Each change is small, sensible and separately approved. Cumulatively, the module in production is no longer the module in the certificate, and no single decision in the chain was large enough to attract attention. The programme's greatest asset — a repeatable, evidenced configuration — has been eroded by exactly the incremental improvement discipline the operation was designed to reward. How reward architecture produces that behaviour, and what it reveals about what an enterprise believes a team to be, is the argument in [Related article: The Reward System Tells You What You Think a Team Is]; the concern here is the evidence, not the incentive that erodes it.
The assembly completed by someone else
Now take a hypothetical orthopaedic implant distributor. Evidence in such systems commonly attaches to a matched set — the implant, its matching components and its instrumentation — because the performance being claimed is a property of the combination. Individual items from two different systems may each carry impeccable evidence, and the combination of them carry none at all.
The distributor does not build the combination. A procurement decision inside a hospital, made on price and availability, may put two evidenced systems together into an unevidenced one. The distributor's exposure is created by a substitution it did not authorise, cannot see, and will nonetheless be asked to answer for. Supplying components separately without supplying the combination rule is where the commercial model and the evidence model diverge.
The axes this article does not cover
A configuration can also fail the evidence test without anything being substituted at all, because the test method itself has moved and the obligation names a particular edition of it. That is a different axis of the same architecture, and it is worked through in [Related article: The Law Pins the Edition; the Standard Moves On]. Here the assumption is that the right method applies and the question is only what the resulting evidence describes.
A claim can also be missing at the root, where the estate carries no benchmark for the class of thing at all and the enterprise's own specification becomes the only stated performance. That case turns on who authored the benchmark rather than on what the evidence attaches to, and it is the subject of [Related article: Where the Standard Is Silent, the Benchmark Becomes Yours]. This article takes both benchmark and evidence as given.
Decision Framework
The instrument is the substitution evidence gate. It is a single mandatory checkpoint that every proposed component change passes through before commitment, and its defining rule is that its authority level is set by the value of the claim, never by the price of the part.
Define the trigger broadly. The gate fires on any change to a component's manufacturer, designation, formulation, dimensions, finish, fixing method or position within an assembly that supports a conformance claim — including a change made by the supplier without a change to the part number. Silent reformulation is the most common trigger and the least often caught, which means the gate must also fire on a supplier's product notification, not only on the enterprise's own request.
Three questions decide the outcome.
| Question | Test | Outcome | Authority to clear |
|---|---|---|---|
| Does a conformance claim depend on this assembly? | Trace the claim, not the component | If no, release | Delegated buyer |
| Was the incoming item named in the evidence we hold? | Read the document, not the datasheet | If yes, release with record | Technical reviewer |
| Do we hold evidence naming the new combination? | Evidence in hand, not obtainable in principle | If no, refuse or re-evidence | Named accountable engineer |
Four rules make the gate real rather than decorative.
No delegated clearance inside a high-consequence assembly. Where the claim touches personal safety, structural adequacy or a regulated performance outcome, no procurement authority may clear the substitution at any value, including zero.
Evidence in hand. A configuration is evidenced only if the enterprise holds the document naming that configuration. Evidence a supplier says exists, or that could be obtained, is not evidence.
Custody at the asset, not the project. The evidence pack follows the asset through handover, sale and refinancing. A programme that disbands with the certificates in a project drive has already lost them.
A retrospective sweep. Run the gate backwards across assemblies installed within the liability window. The purpose is not to apportion blame but to size the exposure, because the enterprise cannot price a risk it has not counted.
From Strategy to Execution
Immediate. Publish the trigger definition and remove substitution from delegated procurement authority wherever a conformance claim is in scope. Instruct the supply base to notify formulation and specification changes, and treat silence on that point in a supply agreement as a defect to be closed.
Medium term. Reclassify substitution from a procurement event to a verification event in the change-control taxonomy, so the default routing carries it to a reviewer rather than to a price adjustment. Build a register that maps each conformance claim to the assembly it describes and to the components named in the evidence, so the second gate question can be answered in minutes rather than weeks. Where the enterprise supplies components that a customer combines, publish the combination rule as a condition of supply.
Long term. Treat evidenced configurations as capital assets with a maintenance cost. They decay through supplier change whether or not the enterprise touches them, and an operation that never re-evidences anything is not holding its position; it is running down an asset it never put on the balance sheet.
Signals to Monitor
Watch for the phrase "equivalent" or "like-for-like" appearing in approval records without an attached evidence reference. Watch for certificates held for a configuration that differs from the installed build, particularly where the certificate came from a component manufacturer. Watch for supplier product notifications that are filed rather than assessed. Watch for value-engineering exercises reporting savings on packages that carry regulated claims. And watch for handovers where the evidence pack is a folder of brochures rather than documents naming what was actually built.
Questions for the Leadership Team
- For our five highest-consequence conformance claims, do we hold a document naming the configuration we actually built, and has anyone read it?
- What is the lowest authority level at which a component inside one of those assemblies can currently be changed, and what is the claim's value?
- How many substitutions were approved on our largest programme in the last year, and how many were routed to a verification review rather than a price review?
- Which of our suppliers are contractually obliged to notify a formulation or specification change, and when did one last do so?
- Where our customers complete the final assembly, what have we told them about which combinations our evidence covers?
- If a regulator or insurer asked today for the evidence covering an installed assembly, how long would it take to produce, and what would the gap look like?
Closing Perspective
Most enterprises believe their conformance evidence describes what they sell. It describes what they once tested. The distance between those two statements is created by thousands of small, competent, well-intentioned decisions, each of which was correct within the frame the decision-maker was given and none of which was ever assessed against the frame that matters.
That distance is not primarily a technical problem, and it will not be closed by better engineers. It is a question of who is permitted to reopen an evidence position, and at what level of the organisation that permission sits. At present, in most enterprises, the answer is that hundreds of people hold that permission and none of them know they do. The one decision leaders can make this quarter is to take it back — deliberately, in writing, and at a level where the person holding it also carries the consequence.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.