Certificates, declarations and test reports, and what each is worth

Sort the folder into three piles first

Every document in a compliance folder was produced by one of three acts. Somebody measured something. Somebody certified something against a scheme. Or somebody asserted something. The three are worth different amounts and most folders are dominated by the third.

The reason this matters before anything else is that the piles behave differently under pressure. A measurement can be wrong about your goods but it is about goods. A certification is about a scheme, a scope and a period. It says nothing outside those three. An assertion is worth what the party making it is worth, which is sometimes a great deal. It does not become a measurement by being filed next to one.

There is a second thing to hold in mind while you sort, because it decides what the sorting is for. A folder is documents. A passport is data: named values, reached through an identifier, that a system can read without a person opening anything. Nothing inside a folder converts one into the other, and most of this page is about what survives that crossing.

That correction matters more than it looks. The statutory route is a duty that pushes information towards the recipient of an article and a right a consumer can exercise, not an instrument a brand can point at its own supplier. What the route does give you, and what an unanswered request is worth, is set out in getting evidence out of suppliers.

The two certificates that get confused, and the asymmetry between them

A scope certificate says that a site holds a certification for named product categories, across a period. A transaction certificate says that a stated volume of certified material moved between certified parties on a stated shipment. People treat the first as the important one, because it is the one that looks like a credential and the one with a date on it. It is the second that is about goods.

Two of the six schemes whose published rules we read print materially the same disclaimer on the face of the scope certificate itself: that it provides no proof that any goods delivered by its holder are certified. Two independent scheme owners, writing their own rules, chose to put that sentence on the document rather than in a manual nobody opens. The proof of the goods is the transaction certificate, and the schemes say so themselves.

Now put the clocks beside the disclaimers. A scope certificate runs a one year term, so it announces its own end and any calendar can watch it. A transaction certificate carries no expiry field at all, because it is a record of a shipment that has already happened. The document that expires proves nothing about the goods. The document that proves something about the goods never expires, and has nothing for a calendar to watch.

One more limit belongs here rather than further down, because it governs both documents. The chain of custody standard's own published abstract states that a chain of custody document is not, on its own, able to support claims about an organisation's materials or products. We read the abstract and not the standard, which is behind a purchase, so that single sentence is all this page takes from it. Where this bites hardest on a real field, and what a volume reconciliation does and does not let you say, is on recycled content.

What the paper does not carry with it

A certification proves what it certifies and nothing adjacent. That single sentence, applied honestly, would prevent most of the trouble people get into with a full folder, and it is harder to apply than it sounds because the adjacent thing is usually the thing you wanted.

Three omissions recur. The document rarely names the items it covers, so the scope is supplied by whoever files it. It rarely says at what granularity it was produced, so a batch result is read as a product fact. And it usually carries no relationship to the claim it is being asked to support, so the comparison that decides everything happens in somebody's head and is never written down.

The granularity omission is the expensive one. Evidence is produced at batch or lot level because that is how goods are made, and claims are made at item or model level because that is how goods are sold. The join between the two is lossy in the direction that matters, and no document in the tables below carries the aggregation rule that would make it safe. What that means for a passport, which has to pick a level and stay on it, is set out on what a passport field can prove.

What each document establishes, and how far we have read the instrument behind it

The final two columns are the interesting ones. The jurisdiction column says which body of law the row was read against, and the read status applies to the instrument behind the row rather than to the document in your folder. Where a row says not read at source, it describes usage that practitioners share rather than a legal weight we can support, and it says nothing about what any particular jurisdiction would make of the document.

That ratio is the honest state of this page rather than an editorial choice, and it is the reason the middle columns of the unread rows describe what the document is for rather than what it proves in law. A partial table with real read statuses is worth more than a complete one that is guessing, and the missing rows are a job of work rather than a mystery: sectoral conformity law, per document type, per jurisdiction.

It is worth saying how large the public gap is, because it explains why this page is built the way it is. The best public answer we found to the oldest question in this vocabulary, which is what separates a certificate of conformance from a certificate of compliance, is a practitioner forum thread first posted in 2005. Its accepted answer is that the documents are in general legally worthless. That conclusion is not one this page repeats: instruments governing the conformity declaration and the technical file retention period are confirmed and registered here, which is enough to know the blanket answer is unsafe and not enough to state what replaces it. The thread reached us through a review rather than at the thread itself.

Where an absence like that is typed, and why not found by us never becomes there is none, is set out in how we know. The short version is that we looked in a stated place and did not find it, which is a fact about our search rather than about the world.

What each scheme says it does not certify

Scheme owners are usually more careful about their own limits than the people quoting them. Every scheme whose published rules we read states, somewhere in its own documents, what its certificate does not cover. Those statements are the cheapest evidence on this page, because the scheme is the authority on its own rules and nobody can argue with a limit the scheme itself wrote.

Schemes are described here by what they certify rather than by name, for the same reason they are on the neighbouring pages. Naming a scheme is a claim about that scheme, and a claim about a scheme needs its address registered and its documentation read to the depth the claim requires.

Two of those six rows carry the limit as the scheme states it on the document itself, because that disclaimer is printed on the face of a certificate we read. The other four are recorded from the schemes' published rules and their exact wording is still being transcribed against the document code and version it appears in, which is why the third column says so rather than offering a paraphrase as though it were a quotation. On the chemical rows the read is thinner still: seventeen of the twenty two observed sources behind that layer have not been read, so no row there states what a scheme fails to publish.

The clock map, document by document

This is the same question the previous tables answer from the other end. Not what a document proves but for how long, and whether anybody finds out when it stops. Read the fifth column first.

The fifth column is the finding. It gives the same answer in all nine rows, for the same reason each time: these documents are issued to a holder, and nothing in any of the schemes whose rules we read carries a change in one of them to the party who relied on it three tiers downstream. A column with one answer in it looks like unfinished research. It is the result.

The second observation is about the shape of the rest. Four of the nine rows carry a date a calendar could watch, and a fifth does so only if the certifier chose to print one. Everything else in the table is a dependency problem rather than a scheduling problem, which is a much harder thing to build, and it is why most systems in this area quietly implement the dated rows and call the result expiry management. A calendar is the right tool for a term and the wrong tool for an alteration nobody logged.

One row runs on a clock nobody in your chain controls. A substances declaration goes stale when the underlying list is revised, and a reference recorded once and never revisited is the standard failure. That behaviour, and what to do with a reference you cannot look up, is on the SCIP reference field.

Nothing propagates

When one of these documents stops being good, nothing moves. Of the six schemes whose published rules we read, not one operates automatic invalidation propagation, so there is no route by which the withdrawal of a certificate reaches the party who relied on it.

The clearest public test of that is not an assertion and does not need to be. After an accreditation body suspended a certification body, the scheme's published answer was that previously issued certificates were still considered valid unless otherwise notified. That is a defensible position for a scheme to take, and it is also a plain description of a system in which nothing downstream is told anything by default.

There is one near counter example and it is worth stating exactly, because it is the closest the market gets. One of those six schemes issues transaction certificates that cite the input certificates behind them and carry an invalidation reason code, so the data structure to express an invalidation exists. Even there the act is a certification body's discretionary decision, taken one node at a time, by somebody who has to notice first. A field for a reason is not a mechanism.

The consequence lands on whoever published a value that a document was holding up, and that is a different problem from filing: what to do about a field you have already published when the evidence under it changes is set out in owning a published field.

The collapse from a record to a PDF, and the gap that follows it

The largest single drop in this whole system happens before your folder ever receives anything. A certification body runs an audit and produces a structured record: the scope by product category, the sites, the dates, the auditor, the findings and how they were closed, the volumes reconciled. What leaves the building is a PDF with a certificate number on it. From that PDF a downstream system recovers the holder, the scheme, the number and a date, if a date was printed. Everything that made the record evidence stays behind.

Downstream of that collapse sits the format gap, and it is the shape of the whole problem in one sentence. Retailers collect compliance as PDFs in a vault, indexed by supplier and by date. A passport is structured data: named values, reached through an identifier, that a machine reads without opening anything. Nothing in the market is bridging the two, and the honest half of that observation is the description of the gap rather than a proposal to fill it.

The mechanics of a real submission route show what a vault costs once it is large. One large platform we looked at accepts compliance data against the article rather than against the supplier, and its own rules state that data for articles onboarded before a stated date cannot be updated retroactively. We are not naming it, and the documentation behind that observation sits in the research record rather than among this page's declared sources, so treat it as a shape to check for in your own channels rather than as a rule we are citing. The consequence is worth carrying either way: a vault acquires an oldest layer that no correction can reach, and it is the layer with the most product in it.

What to do with the folder before anything else

Classification comes before collection. It is cheaper, it can be done in an afternoon with what you already hold and it needs nobody else to reply. It also tends to show that the missing documents are fewer and more specific than they felt, which changes what you ask for.

1. Sort by what produced it. Measurement, certification or assertion. Three piles, no subcategories yet.
2. Separate the site documents from the goods documents. Anything that certifies a site or a scope goes in one pile and anything that names a volume, a lot or a shipment goes in the other. If the second pile is empty, you hold no document about your goods.
3. Write the scope on the row. Which products, which batches, which markets, which stage of the life cycle. If the document does not say, record that it does not say rather than inferring it.
4. Write the expiry, or write that there is none. A blank expiry field and an absent expiry are different states and only one of them is safe. Record the rule document, its version and its effective date while you have the certificate open.
5. Name the claim each document is being asked to support. Then compare the words of the claim with the scope you just wrote down. This step is where the value is.
6. Where nothing supports the claim, record the gap. Not the nearest document. A gap with a reason is a usable artefact and a nearly relevant certificate is a liability with a filename.

The last thing to do is the one nobody schedules. Decide, for each document, who would find out if it stopped being true, and by what route. Where the answer is nobody, that is not a process failure on your side. It is the fifth column of the clock map, and it is the state of the market.

Sources

• REACH Regulation (EC) No 1907/2006, Articles 3(33) and 33
  Art. 33(1), Art. 33(2)CELEX 02006R1907-20250422In forceRelevant provisions reviewed

  Read in full at the consolidated text. It carries two rows of the weight table and the correction in the first section: one paragraph is a push duty owed to the recipient of an article, the other is a consumer request right with a stated period, and neither creates a right for a brand to require anything of its own supplier.

  View official source

• Regulation (EU) No 1007/2011 on textile fibre names and related labelling, consolidated text
  In forceRelevant provisions reviewed

  Reviewed in part rather than in full, and the composition row is marked accordingly. What is supported is that the labelling rules govern the fibre names, their order and their percentages. What is not supported, and is not written, is anything about who checks a composition statement against the goods.

  View official source

• Chain of custody scheme rules, scope and transaction certificate specifications
  Scheme rulesRelevant provisions reviewed

  The spine of the two certificates section, the limits table and most of the clock map. Published rules read at the schemes' own documentation, which is what allows this page to state that two schemes independently print materially the same disclaimer on the face of a scope certificate. Three of these documents carry effective dates in late 2026 and early 2027 and both changes alter what this page says, so every row is written to carry its document code, its version and its effective date. Those identifiers sit in the research record and are marked in the table as awaiting transcription rather than being written from memory.

• Published scheme documentation for the textile chemical management data layer
  Scheme documentationOfficial source confirmed, detailed review pending

  Behind the chemical rows of the limits table and the clock map only. Partly read: seventeen of the twenty two observed sources behind the wider proposition have not been read, so each row carries its own read status and no row states what a scheme fails to publish.

• ISO 22095, chain of custody, published abstract
  StandardPublished abstract read, full text behind a purchase

  One sentence is taken from the published abstract and no more: a chain of custody document is not, on its own, able to support claims about an organisation's materials or products. The governed model names are behind the purchase and this page does not use them.

• Conformity documentation requirements, declarations and technical file retention
  Registered, not read in fullOfficial source confirmed, detailed review pending

  Registered so that the weight table can carry a read status per row rather than a blank. The instruments behind the conformity declaration and the technical file retention period are confirmed and are not read in full, so those rows say what the document is used for and say nothing about what a court or a market surveillance authority would make of it.

• ECHA Candidate List of substances of very high concern
  Regulator dataReviewed in full

  Here for the clock map rather than for a count. The list is revised on a cycle, so a substances declaration can stop being current while sitting untouched in a folder, which is the quietest of the expiry behaviours on this page.

  View official source