The product data you already have, and which of it a passport can use

The data is scattered because nobody had a reason to keep it together

A passport wants values created by different functions, at different moments, for different reasons. No business keeps them together because until recently no business needed to.

Two observations follow and the second is the useful one.

The richest material sits in development and in supplier correspondence, and both are the systems least likely to be in scope when somebody scopes a data project around the commerce catalogue.

And a fact does not become more true by being in more places. The same composition and origin are commonly written on a sample form, again on a specification sheet, again on a certificate with its own issuer and its own validity, and again onto the physical label sewn into the garment. Four artefacts have to agree before a passport is even in the picture. Where they already disagree, a passport does not create the disagreement. It publishes it.

Four questions that decide whether a value is usable

Not a maturity model, and there is no score. Four questions asked of one value at a time. A value that cannot answer all four is not yet publishable, whatever field it is sitting in.

One: what object does it describe?

Almost every unusable value in a business is a true statement about the wrong object.

A value can describe a site, a process, a formulation, a material, a shipment, a style, a production run or one physical garment. Those are eight different things, and a passport attribute is a statement about one of them. A supplier's audit result describes a facility. A restricted substance list describes purchasing. A wet processing conformance describes a process over a period. A transaction document describes a movement of goods. None of them is a statement about a garment, and none becomes one by being filed next to a product record.

This is the most common failure and it is not a data quality problem. The document is genuine, the programme behind it is well run, and the object is different. Where a whole class of evidence sits on the wrong side of that join, the reasoning is worked through class by class on chemical data, and whether it reaches the garment.

Two: at what level is it true?

A composition is true of a style. A dye lot is true of a run. A wear mark is true of one garment. A value filed at the wrong level is either overstated or lost.

The expensive failure is the first. A claim true of the black and not the ecru, published at style level, asserts of a whole range something that is true of part of it. The cheaper and more common failure is the second: a fact captured at run level with no record of which products that run became, which cannot be joined back afterwards because the join was never written down.

The practical rule is short. Record every fact at the level it is actually true, and keep the unit you sell separate from the unit you record. Facts filed where they belong can be rendered at whichever level a delegated act picks. Facts flattened onto the wrong unit have to be collected again. Which level a passport eventually sits at is not settled for textiles, and what turns on it is on how many passports a range needs.

Three: whose statement is it, and can you produce the document?

A value without provenance may well be true. What it cannot do is answer for itself. The test is not whether somebody remembers where a number came from. It is whether the source, the document and the date can be produced today, because a value you cannot support is one you cannot defend, cannot correct with confidence and cannot hand to anybody else.

Three states are worth keeping apart and most systems merge them. A value somebody measured. A value somebody stated. A value a model produced. Two of the three are assertions and one is a measurement, and a field called composition holds all three without saying which it holds.

That is not a subtlety. Two declared compositions disagreeing is a records problem, and somebody has to work out which document is newer or better. A declared composition and an analysed composition disagreeing is a different problem with its own arithmetic and its own tolerances. If the record does not say which kind of value it holds, nobody downstream can ask the right question about it.

Four: what does it cover?

The last question decides cases and it cannot be answered by looking at the value. It is a comparison between two things: the scope of the evidence and the wording of what you intend to publish.

Scope fails on several dimensions independently, and a document can be perfect on all but one. Which market. Which variants. Which production runs. Which chain of custody model, where a certificate rests on one. Which period. And whether a facet does not arise at all, which is a finding in itself rather than a blank. The vocabulary for all of that, and the reason an absence is typed before it is written down, belongs to how we know, and what a blank means, which is the page the rest of this estate defers to on the subject.

The denominators nobody records

Three of the most common values in a catalogue carry a denominator that nobody writes down, and each of them propagates.

Weight. A catalogue's weight field usually exists to price a shipment, so it includes whatever the parcel includes. A passport wants the product's own mass without packaging, because several calculated values divide by it. The two agree closely enough to pass any plausibility check and differ by the weight of a mailing bag, which makes this a quiet error rather than a loud one, and every figure calculated downstream inherits it.

Composition percentages. The denominator is fibre mass rather than product mass, and a declaration that does not add to one hundred is telling you something. Scaling it silently to one hundred throws that away.

Any percentage from a certificate. Per cent of what, across what period. A proportion taken off a document that records a volume relationship across a supply chain, and printed next to a garment, has changed subject without saying so.

One sentence covers all three. Store the strongest form of the value and record its denominator in the same place. Everything weaker can be derived from the strong form, and nothing can be derived from a weak one.

A value can pass all four questions and still be wrong now

The four questions establish what a value is. They do not establish that it is still true, and a surprising amount of what a business holds was right when it was recorded.

Three clocks run under an ordinary product record and none of them announces itself. A certificate expires. A commodity code ages, because the nomenclature it comes from is republished every year and a stored code silently stops matching. And a fact changes because the world moved: a factory changes between orders and the origin is now wrong with nothing having been done to the product.

The thing that makes this hard is not the clocks. It is that nothing carries a change back to the value. When a document supporting a published field stops being good, no mechanism tells the field. What ends each kind of document's useful life, and whether anything announces it, is on certificates, declarations and test reports.

For an audit of your own data the consequence is one extra mark rather than a fifth question. Beside every value, note whether it can go stale and what would make it. That mark is the difference between a data project and a maintained record, and it is the line every published cost estimate leaves out.

What "we already have the data" usually means

Run the four questions across a real catalogue and the answers fall into a pattern consistent enough to plan around.

A large share of what a passport wants is already present and correct and needs structuring rather than obtaining. Identifiers, variant structure, weight, origin and the commodity code usually sit in the catalogue already, although not always where a merchant expects. On one common platform the commodity code and the country of origin live on the inventory item rather than on the product and are absent from the export most merchants use to audit themselves, so a business looking at its own spreadsheet concludes it does not hold values that it does hold. What that platform does and does not hold is worked through on passports from a Shopify catalogue.

A second share exists as prose rather than as data. A composition sitting inside a product description is a real fact in an unusable form, and it stops being extractable the moment somebody edits the description for tone.

A third share exists as documents rather than as values, in a folder indexed by supplier and by date. Nothing inside a folder converts a document into a value. The conversion is manual, one field at a time, and it is work most businesses end up doing twice.

And a fourth share does not exist anywhere, in any form, for anybody.

The gap that tidying does not close

On the test catalogue this business runs, several of the attributes we track resolve for nobody at all, and the reasons are different in kind rather than in degree.

One is structural. A reference to a substance notification can be recorded by whoever notified and cannot be discovered from a product identifier by anybody, because of how the database is built rather than because of how hard anyone looked.

One is access. The records exist and confirming them anonymously and free of charge is not possible.

One is that the evidence does not exist until somebody upstream writes it down, and the single statutory route compels disclosure of what a supplier already holds and never compels a supplier to go and find out.

One is that no method with authority behind it exists yet, so the correct behaviour is to wait and record why rather than to publish an opinion with a number attached.

That distinction decides where effort goes, and it matters more than any completeness percentage. A field empty because nobody tidied is cheap. A field empty because the evidence has never been created is not a data project at all, and no software moves it. Which attributes sit in which group, and how far the evidence behind each has to reach, is set out row by row on the field guide.

The uncomfortable version is worth saying plainly. The limit on what a passport can carry today is not the quality of anybody's product information management. Some of the facts a passport wants were never created by anybody, and a business that has done everything right will still have empty rows.

How to run this on your own catalogue in a day

Take one representative style rather than the whole range. All of it is done against your own records and none of it requires a reply from anybody.

One narrowing before you start, because it saves most of the day. Annex III of the framework sets out the elements a delegated act may draw on for a passport, and it is a closed list of twelve. Identity and codes, compliance documentation, instructions and warnings, and the operators and facilities behind the product, plus a pointer to whatever information requirements are set elsewhere. That is the outer edge. So the diagnosis below is worth running first on identifiers, on the documents you hold and on the operators and sites behind the style, because those are the parts of the list that exist independently of any act. The substantive product claims sit behind the pointer, and nobody knows yet which of them a textile act will select.

List every place the style's facts are written down. Catalogue, development system, finance system, compliance folder, supplier email, buyer pack, physical label. The list is usually the first surprise.

For each value, write down the object it describes. A site, a process, a material, a shipment, a style, a run or a garment. Do not write the field name. Write the object.

Mark the level each value is true at, and mark separately every value that is true of part of the style rather than all of it. That second list is your variant gap. It is the input to every later decision about level and it cannot be reconstructed from records that never carried it.

Mark whether each value was measured, stated or modelled, and whether the document behind it can be produced today rather than found later.

Write the scope onto the row for every document you hold, in the document's own words rather than in yours. Where a document says on its own face what it does not cover, that sentence is part of the scope.

Mark the values that will not hold still, and what would move each one.

Write down what you cannot answer, and which kind of absence it is. A recorded absence with a reason is usable by the next person. A blank cell is a question nobody knows has been asked.

At the end of that day you have three lists no software gives you: values you could already publish, values that need somebody else, and values nobody can produce. Every quotation you receive afterwards should turn on those three, and one that does not is being priced against somebody else's catalogue.

What this page does not settle

Three things, and each is somebody's decision rather than a fact anybody can look up.

Which copy is the true one. The same attribute is routinely held in the catalogue, the channel feed and the compliance record, and they disagree. No instrument read for this estate sets a precedence order between a business's own systems. A universal rule would be an invention and a plausible invention would be worse than none, because somebody would follow it. What can be said is the shape of a real answer: it is specific to your systems and to who maintains each of them, it is a decision somebody takes and writes down rather than a fact anybody discovers, and what makes it work afterwards is that the decision travels with the value.

How much of your data is usable, as a number. This estate publishes no completeness score, here or anywhere. A figure rating a business against its own gaps would turn a method into a product and would rest on a denominator nobody agrees on.

Whether a value is true. Everything above establishes what a value is and what it covers. None of it establishes whether the person who stated it was right, and that is a different question with its own instruments.

Sources

• Ecodesign for Sustainable Products Regulation (EU) 2024/1781, consolidated text
  In forceRelevant provisions reviewed

  CELEX 32024R1781. In force. Read in full at the Official Journal HTML rendering, 28 August 2026, including Annex III. Read for one proposition, which opens this page: that Article 9 requires the data in a passport to be accurate, complete and up to date, without a materiality threshold and without a reasonable efforts qualifier. The article number is cited on the strength of the direct read, corroborated by a Commission implementing decision published in the Official Journal which names the framework's passport requirement articles by number. Annex III is now read and is used for one further proposition, in the section on which fields are worth diagnosing first: that the elements a delegated act may draw on are a closed list, and that identity, codes and documentation dominate it.

  View official source

• Shopify Admin API product and inventory item documentation
  Platform documentationRelevant provisions reviewed

  Platform documentation. Relevant provisions reviewed. Checked 28 August 2026. A market participant rather than an authority, and evidence of platform behaviour rather than of what the law asks. One checkable claim rests on it: that the commodity code and the country of origin are held on the inventory item and are absent from the product export most merchants use.

  View official source

• Retailer supplier manuals and onboarding requirement sets
  Buyer requirement setRelevant provisions reviewed

  Buyer requirement sets. Relevant provisions reviewed. Read for where one product fact is captured across a trading relationship. One proposition rests on them: that a composition and an origin are commonly written onto a sample form, a specification sheet, a certificate and the physical label before any passport exists. These are enterprise documents written by large buyers for suppliers who have already signed a trading agreement, and a brand selling through its own store never receives one. None of the manuals read asks for a passport.