What a textile passport programme costs, and who can tell you

Why there is no published figure

Preparatory work funded by the European Union went looking for exactly the number this page is about. It interviewed four small businesses and four passport service providers. It reported that it could not obtain quantified cost data, because the providers were unwilling to give figures for commercial reasons and because some of them did not yet know.

The people who couldn't answer were the suppliers of the service. The study records a supply side declining to price its own work at a point when the work isn't fully defined. That's a finding about a market, not about the businesses that would buy from it.

It also sets the standard for everything below. Where a figure exists, it is an estimate published by a party with something to sell, and this page says so each time. Where no figure was found by us, the page says that instead of rounding the absence up into a number. How the estate decides what counts as established, and what it does with a figure it cannot trace to a method, is set out on how we know.

What is published, and by whom

Three kinds of figure circulate, plus one published absence. None of the three is a measurement of what this work costs a business like yours, and each is useful for something different.

Read together, they say the software is cheap and its price is settled. The work around the software is where the money goes, and that is the half nobody has priced.

The published prices also agree with one another more than they disagree. Across the independent European anchors we collected, the annual asking price for a compliance oriented programme clusters into one band rather than scattering across the map. We are not printing that band here. A price anchor is only worth anything when every point in it carries the address it came from and the date it was read, and that transcription has not been done, so the figures stay in our research record until it is. What publishes today is the shape: several parties who did not consult each other have landed in the same region, which tells you the category has a price. It does not tell you that the price is yours.

The second calibration comes from next door. Tidying product data, deduplicating it and getting it consistently attributed is not a new market. It has been sold under other names for years, and the published asking prices in that adjacent category sit at a fraction of the figures quoted for passport programmes. That ratio is in the research record on the same condition, and it is the shape that matters rather than the multiple. A reader being quoted for a passport programme is partly being quoted for data work that already has a market price, so asking a seller which half of the quotation is which is a fair and answerable question. Put it to us as readily as to anybody else.

The one item on this page with a real price list behind it

Almost everything above is somebody's estimate. Identifier allocation is not. If you want machine readable identity on your goods you join a national issuing organisation, and that organisation publishes a fee table you can read before you commit to anything. It is among the smallest costs here and it is the only one you can look up.

The awkward part is that the fee tables do not agree across borders. Of the three national member organisations we read, one prices by turnover band with an allocation capacity attached to the band, one prices on capacity and turnover together, and one states that its fee is independent of how many codes you use. So a brand whose range runs to a wide size and colour matrix pays for that width in some countries and does not pay for it in others, on nothing more than where the business happens to be incorporated. We looked for a vendor proposal or an industry model that prices that asymmetry and did not find one, which is a statement about where we looked rather than a claim that none exists.

The amounts are not on this page. The fee pages were read as published pages, the figures sit in our research record, and each will be transcribed with the date it was read before any of it appears here. Two disciplines travel with them when they do. Any total for a catalogue of a given size is derived from a fee shape rather than quoted from a price list, and it carries the word derived wherever it appears. And the cost of being allocated identifiers is kept apart from the cost of a code whose destination can be changed after printing. Those are two products with two bills. They are quoted together often enough to be worth prising apart in writing.

Estimate the work that genuinely needs a person

Start by separating the passport work ActivateDigital can do from the confirmations that need the merchant or supplier. A field list multiplied by every size is not a useful estimate of human effort.

ActivateDigital resolves the existing product data first. In the current controlled textile test flow, the initial queue then averaged 5.1 product-specific questions per product, or 6.1 including the shared company-identity question. These were starting questions before answers, so follow-ups can arise. The median and range are in how the test was counted.

The planning question is who can answer each genuine gap. A reusable business detail and a supplier evidence request can each be one question but have very different dependencies. These counts do not measure elapsed time, completed journeys or money saved.

See the passport work ActivateDigital does before asking you, then scope the confirmations and evidence requests that remain. The textile readiness field guide explains the underlying information. Independently sourced cost evidence elsewhere on this page remains separate from the test.

How to estimate the work in your own catalogue

You do not need a vendor to get within range of your own number. You need three counts and one mark, and every one of them comes out of your existing catalogue.

1. Reuse facts at the level where they apply. Several sizes can share one composition when the same declaration covers them; a different colourway, component or material may need its own evidence. Count that actual difference, not another full questionnaire.
2. Let the existing records answer what they can before chasing a supplier. Then count the specific outstanding facts and evidence requests. A genuine supplier dependency needs attention; information already in your files does not need collecting twice.
3. Count the attributes nobody can answer. Write each one down with the reason it cannot be answered. A recorded absence with a reason is usable evidence. A blank cell is a problem you will meet again later.
4. Mark the answers that will not hold still. Go back over the same style and mark every answer that moves when a season turns, a supplier changes or a certificate runs out of validity. That mark is your recurring cost, and it is the line the published estimates leave out.

Those counts are your estimate. They are also the inputs any quotation you receive should turn on, so they are worth having before you read one. What the money buys is a separate question, and the honest answer is that nobody has measured the return.

To turn the counts into a budget line, put your own time and rate against each one. Count shared company setup once where the same business, role and scope apply, not once per size. Then add the outstanding product facts and evidence requests, the absences you record with a reason and, for every year you expect to sell the range, the answers you marked as moving. Supplier follow-ups and corrections belong in that last line, because that's where the time goes. The total is your figure, built from your catalogue, and it gives every quotation the same four lines to answer.

Before comparing quotes, run a representative-product software pilot. Record what the system does, what your team still does and which tasks recur when the product changes.

The quickest way to get the first count is to run those counts against one product from your catalogue and see what the records already answer.

What actually moves the number

Four things change the count. None of them is the software.

1. How much of your catalogue is already true. A field holding what its name says resolves for nothing. A field holding an internal code, a guess or a value somebody typed in a hurry costs a question and then a correction. Which of your existing store records already hold a usable answer, and which only look as though they do, is worked through on the platform guide this page sits under.
2. How many distinct facts your range rests on. A style made in six sizes and four colours is one composition question and twenty four sellable units. Counting units tells you how much printing you face. Counting facts tells you how much asking you face. Which of the two is also the number of passports has not been settled, and that is the subject of how many passports a range needs.
3. How reachable the people upstream are. The same question costs an email to a mill you have worked with for a decade and costs months to a supplier you buy from through an agent. Nothing on your side of the relationship changes that, which is covered on getting evidence out of a supplier.
4. How many of your attributes have no answer anywhere. These are the cheapest to identify and the most expensive to pretend about. Recording one as unanswerable, with the reason, costs an afternoon. Discovering it late costs a reprint.

None of the four is knowable from a price list, and all four are knowable from your own catalogue, so if you want the count for yours rather than for ours you can run the same pass against your own catalogue.

The cost gradient: where the money actually sits

The four above are about your catalogue. There is a second gradient underneath them, and it explains why published prices and real programmes disagree so widely: the raw digital core is the cheap end, and almost nothing else is.

Resolution itself is close to free. GS1 runs a shared resolver service and says organisations can connect barcodes to the web at no cost or run their own. Compute is the same shape: Cloudflare publishes a worked example of 100 million Workers requests a month at 45.40 dollars under its stated assumptions, and AWS Lambda publishes 20 cents per million requests before instance and management charges. Those are real prices and they are not the cost of a passport programme. They are the cost of answering a request.

Everything that makes the answer worth having sits above them: getting the data right, integrating the systems that hold it, putting a carrier on a physical object, governing who may change what, and measuring whether any of it worked. CIRPASS interviewed potential providers and found them reluctant or unable to give quantitative implementation costs at all, on a sample of eight interviews across four SMEs and four providers, before any obligation applied.

Published subscription prices, with what each one leaves out:

The gap between the third row and the fourth is the whole point. A list price is a price for a subscription. It is not a price for the programme, and no published figure we found is.

Who ends up paying for each layer, and on what commercial model, is a different question from what the work costs, and it is answered on who pays for a connected product.

What a real buyer asks for, and at which moment

We read a set of retailer supplier manuals, the documents a large buyer hands a supplier before the first order goes in, to see what product data a funded and contractually enforced request contains.

The structure is the finding rather than the contents. The questions are not one list handed over once. They arrive at four separate moments, they belong to different people inside the supplier, and only two of the four repeat.

Two things follow for anybody counting questions. The first is that the material a passport wants sits almost entirely in the first two moments, which are answered once per supplier and once per style, and not in the two that repeat with trading volume. Our own test catalogue reached the same finding by a different road, so counting styles rather than sellable units is the right instinct.

The second is that some of those items run on a clock. They are due a stated number of days before or after an event, and a late answer is an operational failure rather than a data quality problem. We are not reproducing the intervals, because each is stated per manual and per version, and the versions sit in our research record rather than in transcribed form.

Which of these questions are worth putting to your own supply chain, and how far up it is reasonable to go before the answers stop being obtainable, is a separate judgement from what they cost. It is worked through on how far up the chain you actually need to go.

The same answer, written down in four places

Those documents also show where an answer ends up. Composition and origin are not captured once. In the sets we read they are written on a sample form during development, again on a specification sheet, again on a certificate that has its own issuer and its own validity, then onto the physical label that goes into the garment.

That count is about documents rather than about behaviour. What it does say is that four separate artefacts have to agree with one another before a passport is even in the picture, and that a passport is a fifth. Where the four already disagree, the passport does not create the disagreement. It publishes it.

What it costs to keep, which is the half nobody prices

Every public costing page we read prices a first year. A first year is easy to sell because it has a shape: a subscription, an onboarding fee, a one off data load. Almost none of the real work has that shape.

The work comes back. Suppliers change, fabrics change between seasons, a test report is about a sample tested on a date, and a carrier printed onto a swing tag is an object with a print run behind it. Below are the six recurring items we could name, against whether anybody publishes a price for them.

The last row is the only one with anything published against it, and what is published is a condition rather than a price. One identifier issuing organisation states in its own fee material that product data remains linked for as long as membership is renewed annually.

Read the conditional half rather than the promise. The guarantee is real and it lasts exactly as long as somebody keeps paying, which makes persistence a subscription with a very long tail rather than a property of the code. What happens at the point where nobody is paying is the subject of when the link dies.

So an honest budget has two lines rather than one. There is a first year, which the published estimates describe in their own way. And there is a standing cost that nothing we read has priced, which is not the same as its being small. If you are reading a quotation, the question to ask is not what the first year costs. It is which of those six items the quotation still covers in year three.

Six integration drivers, none of which has a published price

The recurring costs above are about the data. Connecting the systems that hold it has its own recurring bill, and it scales with six things rather than with the number of products:

• how many source systems hold a fact you have to publish, because each one is a connection to build and then to keep working;
• how complex the mapping is between what each system calls a thing and what the passport calls it;
• how often the data changes, because a nightly reconciliation and a change-on-write event stream are different pieces of engineering;
• whether you need event history or only current state, which decides whether you are storing a record or a timeline;
• reconciliation, meaning what happens when two systems disagree and somebody has to decide which is right;
• write-back governance, meaning who may change a published value, through which system, and what the audit trail looks like afterwards.

No published price exists for any of these, and the vendor return percentages that circulate in this territory are not evidence for them. They are cost categories to estimate against your own architecture, which is worked through on connecting a passport to ERP, PIM and PLM.

Four numbers this page does not carry

• A cost per product. No primary figure for one was found by us, the published estimates carry a catalogue size in the same breath as their total, and dividing one by the other invents a rate that neither party measured.
• A figure broken down by how big a business is. Nothing we read supports one. The obligations in this area are written in terms of what a business does rather than how large it is, and a cost figure sorted by headcount would imply a structure the evidence does not have.
• An average. One test catalogue is one catalogue. An average needs a population, and publishing a mean of a single measurement would be a presentation choice dressed as a result.
• Any promise about your effort. Our resolution figures describe what happened on our test catalogue when we ran the test. They are a shape you can reason with, not a floor, a ceiling or a guarantee.

Nothing added to this page since it was first written changes any of the four. The identifier fees, the anchor band and the recurring items above are observed prices and named cost items, and none of them is an estimate of what you will pay. Handled this way an absence is a position rather than an omission. The reasoning behind that, and what has to happen before any of these four becomes publishable, sits on how we know.

Sources

• CIRPASS, A study on DPP costs and benefits for SMEs, version 1.0, March 2024
  Official studyResearch findingRelevant provisions reviewed

  The work funded by the European Union that went looking for passport cost figures and reported that it could not obtain them, from eight interviews conducted in the second half of 2023. We read the report at its own address, and what we read is the interview base and the passage recording the refusal. The rest of it was not read and no wording from it is quoted here.

• PassportCraft published passport cost estimate for a small brand
  Secondary sourceVendor estimateOfficial source confirmed, detailed review pending

  A passport vendor's published year one estimate for a small brand, published 10 February 2026. It is an estimate by a party that sells into the work it is estimating, and the page names it as one. We reached it through our research record rather than a direct read, so the row is second hand.

• Renoon, published cost bands for a digital product passport solution
  Secondary sourceVendor estimateRelevant provisions reviewed

  A second vendor's published market benchmark, quoted here only for its lowest band. We read the page carrying the bands at its own address, which replaces the earlier second hand account. The bands are described by their own publisher as indicative, and they are an estimate by a party that sells into the work it is estimating.

• Published passport application listing prices and review counts in two commerce app markets
  Secondary sourceObserved pricesRelevant provisions reviewed

  Published listing prices and published review counts for the passport applications in two independent commerce app markets, as surveyed on 27 August 2026. Evidence of what is being asked and of how little has been bought. It is market behaviour and it is never authority on what the law requires.

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

  The published fee, allocation and continuity pages of national identifier issuing organisations, read as published pages for the shape of their fee tables and for the continuity statement. Three organisations were read and they price on three different bases. The amounts are held in our research record and are not published on this page, because each needs transcribing with the date it was read first. Fee variation across the organisations we did not read is unknown.

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

  Retailer supplier manuals, read for the stage and clock structure of what a large buyer asks a supplier for and at which moment. What was read is that structure and the places one product fact is captured. One current version of one manual is unread, the intervals are stated per manual and per version and are not reproduced here, and none of the manuals read asks for a passport.