How many passports a range needs, and why it is not settled

The short answer

The level has not been decided for textiles. The framework sets the choice out and hands it to the product act covering a group of goods, and no such act has been adopted for textile products. Anybody quoting you a number of passports for your range is quoting an assumption, including anybody quoting you a small one.

The three candidates are the model, the batch and the item. They are not three ways of describing the same thing, and on an ordinary range they differ by three orders of magnitude. Two things have hardened since this page was first written, and they pull in different directions. The registry that has been built assumes the same three levels and requires a business to link them where more than one exists, so the ladder now has a legal shape with the textile rung still empty. And the number of identifiers sitting underneath that ladder turns out not to be a modelling choice at all.

What is established, proposition by proposition

It is easy to write a page like this that sounds settled by accident, so the table below separates the propositions and gives each one its own state. The states are not opinions about likelihood. They record what kind of document each proposition rests on.

Two of those rows use unresolved language on purpose. Where the estate cannot settle a proposition it says so and records where it looked, and the method behind that is set out on how we know.

The registry already assumes the ladder

The strongest thing that has changed here is not an opinion about which level will win. It is that a working registry has been built on the assumption that all three exist. A registration is made at model level, at batch level or at item level, and where more than one of those levels exists for the same goods the regulation requires the registrations to be linked to each other.

That does two useful things and one unhelpful one. It confirms that the three levels are a real structure rather than a drafting convenience, because a system has been built that accepts records against each of them. It also confirms that they are expected to coexist, which is a different proposition from choosing one: a business may end up holding a model record and item records underneath it, joined. What it does not do is pick a level for textiles, and reading it as though it had is the commonest error currently in circulation about this instrument.

The practical consequence lands before any product act does. If you hold batch information with no reliable way to say which model it belongs to, you are not merely undecided about a level, you would fail a linking requirement that is already in force for anybody registering. What a registration actually contains, and what a business has to be before it can make one at all, is set out on registering a passport.

The arithmetic, and the half of it that is not yours to choose

The gap between the levels is easiest to see on a single style, and the arithmetic below is an illustration rather than a measurement. Take one style sold in five sizes and six colours, produced in two runs, with 1,200 garments made in total. Those numbers are assumed because they divide cleanly. They are not drawn from anybody's catalogue and nothing in the table is a finding.

The outer spread is between one record and 1,200. The more awkward spread is the one inside the first two rows, where the same word produces either one passport or thirty depending on what a model is taken to mean. That ambiguity deserves more attention than it gets, because it sits inside the answer most people assume is the safe one.

The thirty is not a modelling choice. It is imposed by the numbering scheme's own management standard, whose worked example makes each style, colour and size variation a distinct identifier. A business does not decide whether a size and a colour need separate identifiers any more than it decides its own postcode. That settles half of the argument this page is having: the identifier count for a given catalogue is a scheme rule, and a vendor arguing about it with you is arguing about the wrong document. What that standard asks, what allocation costs and why the answer depends on which country allocated the numbers is set out under identifiers.

The two questions still have two different answers. Identifier allocation follows the scheme and what the selling channels enforce. Passport granularity follows the product act. It may land on the same unit as the identifier and it may not, and treating the identifier count as the passport count builds an assumption into the plan at exactly the point where the plan becomes expensive to change, because carriers get printed onto goods that then go into circulation.

None of this changes the amount of information a garment needs behind it. A composition is one fact whether it is published once or twelve hundred times, and a worked passport shows what one of those records actually holds. What the level changes is how many records exist, how many carriers get printed, how long the records have to be kept reachable and what happens when one of them turns out to be wrong.

The level that is missing, and what it costs

Model, batch and item contain no variant level. A model is a style. A batch is a production run. An item is one physical garment. A colourway is none of those three, and that is where the model breaks in ordinary use rather than in a hard case.

The failure looks like this. One style is made in six colours, and a claim is true of some of them and not the others: recycled content in the black but not the ecru, a certificate that covers one dyehouse, a substance restriction met by four finishes out of six. At model level the record has two options and both are wrong. It can assert of the whole style what is true of part of it, which overstates. Or it can drop the claim entirely, which loses a fact the business paid to establish and can prove.

The batch level does not rescue it, because the join is lossy in the direction that matters. Reading upward is easy: from a batch you can always say which model it belongs to. Reading downward is the hard direction, and it is the one a passport needs. From a model record, or from a garment in a customer's hand, you cannot say which run it came out of unless something was captured at the item at the time it was made. So a claim that is true of one run cannot be attached after the fact to the units that came out of it.

What survives that is item-level credentials, which is the most expensive answer on the page and the one nothing has yet required. The honest position is that a per-variant claim has no clean representation in a structure built from models, batches and items. Three routes lead out of it: hold the claim at the item, split the model so that each colourway is its own model, or stop making the claim. Most catalogue systems have already pushed businesses part of the way down the second route by putting the identifier on the variant and the descriptive fields on the product, which is why the split feels natural in a shop and awkward in a passport. Where those fields sit in one common system is worked through under Shopify.

None of that is an argument for buying item-level identity now. It is an argument for knowing, before anybody quotes you, which of your claims are true of a whole style and which are true of part of one. That inventory takes an afternoon and costs nothing. It is also the thing that decides whether the cheapest level is available to you at all.

What a price per passport is actually pricing

A price expressed per passport is a price for a quantity nobody has fixed. That is worth saying plainly, because the quantity moves by three orders of magnitude between the answers on this page and the reader is the party carrying that risk.

The nearest thing to a published price in this territory is identifier allocation, and its shape is instructive. Across the national member organisations of the numbering scheme whose fee pages we read, one prices by turnover band with an allocation capacity attached, one prices by capacity band combined with turnover and one states its fee as independent of the number of codes altogether. Not one of the three prices an identifier individually. So even in the part of this market that has a real published price list, the shapes on offer are the size of the business or a capacity band, and in one country the count does not enter the calculation at all.

We went looking for a published self-serve price per unit for passport preparation from a European provider and did not find one. That is a statement about where we looked rather than a claim that none exists anywhere. It is consistent with what the published cost material does show, which is pricing by subscription and by band, and the figures that do exist are set out with their publishers on what the work costs.

The gradient across the three levels

Cost does not rise evenly as you go down the ladder, and it is more useful to know the shape than a number nobody has measured. Published here as a shape, because the material we read carries fee structures rather than prices per product, and because no cost per product for this work appears anywhere on this site.

• The evidence half is close to flat. A composition, an origin and a supplier document are the same facts at model, batch and item level. Establishing them once is most of the work and it does not multiply.
• The identifier and carrier half rises with the count. Allocation, artwork, printing and the checking that goes with them all move with how many distinct things exist, which is exactly the dimension the scheme rule above has already settled for identifiers.
• Keeping and correcting rises fastest. A wrong value in one model record is one correction. The same wrong value at item level is a correction exercise across everything already in circulation, and carriers already printed cannot be recalled.

One consequence of that gradient is worth stating without hedging. If the model level holds for textiles, an item-level commercial argument has no basis in the regulation. It may still have a basis in something else, in anti-counterfeit work, in warranty, in a service a business wants to offer, and those are real reasons that should be argued on their own merits. What they are not is a compliance requirement, and a proposal that presents them as one is describing a decision that has not been taken.

What has been proposed, and what a proposal is

A study prepared to inform the textile decision is the document everybody quotes here, and it is worth reading. What we read in it does not fix a level. It sets out the same three levels the framework names, and it says that where possible the granularity required should be aligned with the traceability practice already established in the value chain. That is a proposal about method rather than a proposal of an answer. Either way a preparatory study exists to be argued with, and the act it informs may take it up, vary it or set it aside.

Two things circulating as findings from that study are held back here. We could not obtain the sections that would carry a proposed level for apparel, so this page attributes no level to it. The other is a claim that the study's identification block names a serialised trade item number, an identifier that distinguishes one physical article from another. In the part we read it does not. The identification concepts there are a unique product identifier, a unique operator identifier and a unique facility identifier, with no serialised form specified.

The sector's trade association has published a position on granularity as well. An industry position is evidence about what part of an industry is asking for, and it is never evidence about what an act will require. Both of those documents are useful for understanding the argument. Neither of them ends it.

Timing gets merged into this argument constantly, so it is worth separating. An act being adopted is not an act applying: the framework sets a floor of no less than eighteen months between a product act entering into force and applying, except in duly justified cases. That is arithmetic rather than a date, and this page resolves it into no year. What has actually happened, with sources and verification dates, is maintained on the status tracker, which is the destination here that carries dates.

What to do while the unit is open

An open question is not a reason to wait, because almost all of the preparatory work is the same under all three answers. The exceptions are the decisions that are hard to reverse, and those are the ones worth holding.

1. Record every fact at the level it is actually true. A composition belongs to the style. A dye lot belongs to a run. A wear mark belongs to one garment. Facts filed where they belong can be rendered at whichever level the act picks, and facts flattened onto the wrong unit have to be collected again.
2. Write down which of your claims are true of part of a style. That list is the variant gap in your own catalogue, and it is the input to every later decision about level. It costs nothing to make now and it cannot be reconstructed later from records that never carried it.
3. Keep the unit you sell apart from the unit you record. They are separate columns in the same spreadsheet and they answer to separate rules. Merging them is the cheapest mistake to make now and one of the more expensive to unpick later.
4. Make sure a batch record can always name its model. Linking upward is already required of anybody registering at more than one level, and it is far easier to capture at the time than to reconstruct.
5. Do not print anything that assumes an answer. A carrier on goods in circulation cannot be recalled, and a serialised carrier commits you to a level before anybody has chosen one.

None of that is wasted if the act picks a level nobody expected. That is the test worth applying to any preparation you are asked to pay for now: whether it survives all three answers, or whether it quietly bets on one.

Sources

No article number appears below for either of the two instruments, and that is deliberate rather than an omission. Two independent reads of the framework disagreed about which of its articles carries the passport, the identifier and the registry. Separately, the internal numbering of the registry regulation has not been checked against the Official Journal PDF. So every provision on this page is named by what it does, and the numbers can be added later without a sentence changing. Two of the sources are published by the organisation that runs the numbering scheme this page keeps describing. They appear registered with a read depth, which is the accepted alternative to citing such a document as an authority: the page states how far each was read rather than resting on the fact of citation.

• Ecodesign for Sustainable Products Regulation (EU) 2024/1781, consolidated text
  CELEX 02024R1781-20240628In forceRelevant provisions reviewed

  The framework, read at the consolidated text. One sentence in it carries this whole page: the provision listing what a product act specifies includes whether the passport is established at model, batch or item level. The same provisions were read for the application floor between a product act entering into force and applying. Every route to the document stopped inside the later articles, so the annexes were not read and nothing here rests on them. No article of it is cited on this page, because two independent reads of the same text attribute its passport provisions differently.

  View official source

• Commission Implementing Regulation (EU) 2026/1778 establishing the Digital Product Passport registry
  CELEX 32026R1778In forceRelevant provisions reviewed

  The implementing regulation establishing the passport registry, read at article level by one research programme in August 2026. It is here for one provision: registration is made at model, batch or item level, with a requirement to link those registrations to each other where more than one of the levels exists for the same goods. It selects no level for textiles and it sets no compliance deadline of its own. Article numbers are withheld pending a check of the instrument's internal numbering against the Official Journal PDF.

  View official source

• Joint Research Centre, Study on DPP content for textile apparel products under ESPR, 13 May 2026
  ProposedOfficial source confirmed, detailed review pending

  The study prepared to inform the textile decision, dated 13 May 2026. It is pre-decisional by design and it is never cited on this page as law. Its address was confirmed in this pass on the Product Bureau document list and again by opening the file, and the section on granularity was read there. The retrievable text stopped before the sections that would carry a proposed level for apparel, so this page attributes no level to it.

  View official source

• EURATEX position on Digital Product Passport granularity, March 2026
  Industry positionOfficial source confirmed, detailed review pending

  The sector trade association's published position on granularity, dated March 2026. It is evidence of what part of the industry is arguing for. It is not evidence of what a product act will require, and the page says which of the two it is.

  View official source

• GS1 GTIN management standard
  StandardRelevant provisions reviewed

  The allocation standard for the numbering scheme, read for the rule whose worked example makes each style, colour and size variation a distinct identifier. It is what turns the middle row of the arithmetic below from an opinion into a scheme rule. Reuse of an identifier already printed onto goods is governed by the general specifications rather than by this document, and that release is unread, so nothing here turns on it.

• Identifier issuer published fee, allocation and continuity material
  Issuer termsRelevant provisions reviewed

  The published fee, allocation and continuity pages of national member organisations, read for the shapes the fees take rather than for their amounts. Three shapes were read: a turnover band with an allocation capacity attached, a capacity band combined with turnover and a fee stated as independent of the number of codes. Fee variation beyond the organisations read is unknown and no amounts are transcribed on this page.