When the lab and the supplier disagree

Below the tolerance there is no conflict

An accredited result that differs from a supplier's composition statement is not automatically a contradiction. The textile fibre labelling regulation sets tolerances, and inside them a declared value and an analysed value are both correct at the same time. So the first question is arithmetic rather than legal: how far apart are the two numbers, and were they produced by the same method.

Above the tolerance the position changes, and it changes on the date the result was issued rather than on the date somebody acts on it. The same instrument places accuracy on the party placing the product on the market, which means an inaccurate statement is inaccurate whether or not anybody has noticed. Almost nothing written about supplier evidence says this part out loud.

Neither of those sentences settles which of the two numbers is true. That is a separate question with a worse answer, and it is the subject of the last section on this page.

A declaration and a test result are not the same kind of object

A composition statement is an assertion about what a style is made of. A test report is a measurement of a sample, by a named method, on a named day. Setting the two side by side and calling the difference an error assumes they were trying to measure the same quantity, and often they were not.

The regulation does not leave the method open. It prescribes how a composition is determined for the purposes of the instrument, on a stated mass basis, after specified non fibrous matter has been removed and with defined allowances applied. A result produced by a different preparation, or by a method the report does not name, is a number about a different quantity. It can still be useful. It is not a like for like contradiction of the declaration.

Three separations do most of the work here, and only the third of them is arithmetic.

• What was measured. A sample of a fabric, a garment as constructed or a single component. These give different answers on the same goods and none of them is wrong.
• By what method. The prescribed one, an in house one or one the report leaves unstated.
• Against what statement. The composition as stated on the label, or a value in a specification that was never on a label at all.

Sorting a report into the right one of those before comparing anything is the step that dissolves most apparent conflicts. What each kind of document was produced by, and therefore what it establishes at all, is set out on what each document is worth.

Two tolerances, doing two different jobs

Readers who know one figure usually know the wrong one for the situation in front of them. There are two, they measure different gaps and they carry different conditions.

The carded figure is the one that goes missing most often. It is higher because carding is a process in which stray fibre is genuinely hard to exclude, and the higher figure is fenced by the double condition rather than being available on request. A product that is not carded does not get it.

Drawn out, the manufacturing tolerance is a band around the declared value rather than a threshold under it. That shape is the reason a difference in either direction can sit inside it, and it is the picture this page exists to produce.

Everything above is about one attribute, and that is not an accident. Fibre composition is the claim family we have found in EU law that carries both a prescribed test method and a prescribed tolerance, which is why a disagreement about it has a procedure and a disagreement about most other attributes does not. The field itself, including how a declaration is read before any of this arises, is on fibre composition.

The two may be added together only in one circumstance

The regulation says when the tolerances may be applied cumulatively, and the gate is narrower than most summaries of it. They may be applied together where the extraneous fibres found are of the same chemical type as fibres already named on the label. Outside that circumstance the two allowances are separate and stay separate.

This is the expensive error, because getting it wrong is always comfortable. Adding two per cent to three per cent produces a five per cent cushion that feels generous and defensible, and a carded product treated the same way produces an eight per cent one. A reader who has quietly granted themselves those points will believe they are inside a tolerance when the instrument does not put them there.

The condition is also a test about chemistry rather than about labels. Fibres of the same chemical type as ones already declared behave differently in an analysis from fibres of a type nobody declared, which is why the gate is written that way. Nothing about the gate turns on how the extraneous fibre got there or on whether anybody meant it to.

What changes on the day the result is issued

This is the part of the page that matters most and the part that is least often written down. Where the difference sits outside the tolerances, the accuracy of the composition statement is a duty of the party placing the product on the market, and that duty does not wait for an enforcement authority to arrive. The declaration a supplier sent does not carry the duty across. It is evidence about what the supplier said.

So the state of the goods changes at the moment the result is issued rather than at the moment somebody opens the report. A brand that has the result and has not read it is in the same legal position as one that has read it, which is an uncomfortable thing to know and a useful one.

Two things are not established and this page will not pretend otherwise. What becomes of stock already shipped when a result lands is not settled by anything read for this page, and the answer probably differs by market and by channel. Whether any rule exists for turning a batch level measurement into a fact about every item in a style is a question we looked for and did not answer; no instrument read here supplies one and no schema examined here carries one. How the estate types absences of that kind, and why not found by us never becomes there is none, is on how we know.

How often the two disagree, and what that measurement cannot tell you

One public measurement is worth putting on the record, with the caveat that comes attached to it.

The same campaign reports a gradient rather than a flat rate. Across three bands of declaration complexity, recorded non compliance rose with the complexity of the declaration. That is a mechanism, and a mechanism travels further than a category: it suggests the failures cluster where a declaration has more fibres to get wrong rather than in one product type.

What the campaign cannot support is a published rate of wrongness, because it does not say which tolerance thresholds it applied to reach its count. Under the estate's own rules a figure whose threshold is unstated is a count of something rather than a rate of something, and publishing the absence is more useful than publishing the number as though it meant more than it does.

One attribute where this conflict cannot arise

Recycled content does not produce this problem, and the reason is not that the paperwork is better. It is that nothing can contradict the paperwork. A preparatory study informing the textile decision reports that no laboratory test at scale determines whether a fibre is recycled or virgin, and we reached that study through a testing house summary rather than at the document itself.

If that holds, there is no second value for a declaration to disagree with. Verification rests on chain of custody records, which are documents about volumes and about parties rather than measurements of a garment. A disagreement between two such records is a disagreement about bookkeeping, and none of the machinery on this page applies to it.

That is a different problem with a different shape, and it belongs to the field rather than to this page. What a chain of custody record supports, and the gap between a mass balance position and an identity preserved one, is set out on recycled content.

The procedure, with its states named

Five gates in order, and two terminal states. The order matters because each gate can end the question before the next one is reached, and because the arithmetic gate is fourth rather than first. A great many disagreements never get as far as the numbers.

The right hand terminal state is the one that distinguishes this model from the ones that circulate. A procedure that always ends in a single value has to pick a winner somewhere, and the picking is done silently by whoever built it. Here a disagreement that survives all five gates is recorded as a conflict, with both values, both sources and both dates kept, and it is published in that state rather than tidied into a number.

What this page does not say

Three things, and the first of them is the one most likely to be read into the page anyway.

There is no universal source precedence rule

Nothing says a test beats a declaration, and nothing says a declaration beats a test. No provision read for this page establishes an order between sources. No published passport data model we examined carries a precedence concept, a source rank or a conflicted state at all, which is a gap in those models rather than evidence that the question does not arise. We publish that as a finding rather than leaving it to be inferred from silence, and where we looked is on the source list below.

What the regulation supplies instead is narrower and more useful: a prescribed method, two tolerances, a gate on combining them and a duty of accuracy that sits with one named party. Those settle whether there is a conflict. They do not settle which value is true, and the honest procedure has a state for that.

Nothing here is advice about a particular product

This page sets out what the instrument says and what the states are called. It does not tell anybody whether to ship, relabel, retest or stop on goods sitting in a warehouse, because that turns on facts this page does not have and on legal exposure it is not qualified to weigh. Whether a value that survives the gates will support the claim being made on it is a further test again, and it is on whether the claim survives.

No rule for turning batch evidence into a product fact

Evidence is produced at batch or lot level because that is how goods are made. Claims are made at item or style level because that is how goods are sold. We looked for an aggregation rule that would make the join safe and did not find one, in the instruments read here or in the schemas examined here. Where the procedure reaches that boundary it names the limit and stops, which is a worse answer than a rule and a better one than a rule somebody invented.

The neighbouring case, where there is no second source to disagree with because nobody ever answered, has its own procedure and its own permitted outputs. It is on what you may publish when you do not know.

Sources

• Regulation (EU) No 1007/2011 on textile fibre names and related labelling, consolidated text
  CELEX 02011R1007-20180215In forceRelevant provisions reviewed

  The spine of this page. The provisions read at the consolidated text are the extraneous fibre tolerance with its carded variant and both of its conditions, the manufacturing tolerance between stated and analysed composition, the provision governing when the two may be applied together, the analysis provisions with the prescribed method annex, and the provision placing accuracy on the party placing the product on the market. Each is named on this page by what it does rather than by an article number, because the internal numbering has not been eyeballed against the Official Journal and inventing a citation would be worse than describing the provision. That check is open.

  View official source

• Commission joint enforcement campaign on textile labelling, 2026
  Regulator publicationReached through a secondary reproduction, primary text not read

  Here for one figure and for one absence. The figure is published with its denominator. The absence is that the publication does not state the tolerance thresholds it applied, which is why its headline is not republished here as a rate of wrongness. One sub figure in the same campaign rests on five garments and is not published by us in any form.

• Joint Research Centre, Study on DPP content for textile apparel products under ESPR, 13 May 2026
  Pre-decisional researchReached through a secondary reproduction, primary text not read

  Reached through a testing house summary rather than at the study itself, and the sentence on this page that rests on it says so. It carries one proposition here, that no laboratory test at scale determines whether a fibre is recycled or virgin, which is the reason one attribute cannot produce the conflict this page is about. It is pre-decisional and is never cited as law.

  View official source

• Battery Pass Consortium data model
  Published data modelRelevant provisions reviewed

  One of the public passport data models the estate positions against, and the only one of the four whose address is registered here. It is on this page for a negative finding rather than a positive one. It models neither a precedence order between sources nor a conflicted state, and the finding that no universal precedence rule exists rests partly on having looked at models like this one and not found it.

  View official source