The passport figure you can honestly publish is capped by where your material flow commingles, so the first exercise is mapping the commingling points rather than choosing a schema. Granularity is set by the delegated act and not by you, and the step from model level to item level multiplies the record count by three or four thousand and turns a data project into a serialisation and labelling project. Rolling recycled content up a bill of materials gives a ten point range on the same product depending on how unknown mass is treated, which means the roll-up has to carry an explicit unattributed bucket. Approved alternates make the attribute set time-varying, so it belongs on the source of supply record with an effective date rather than as one field on the item.
The product is a moulded housing with three qualified resin suppliers. Planning alternates between them on availability and price, which is exactly what qualifying three suppliers was for. In February the compliance team asks what recycled content figure goes on the passport for the units that shipped in March.
Purchasing knows which orders were received in February. Production knows which run made which pallet. Neither of them knows which resin went into which unit, because all three suppliers deliver into the same silo, and once the material is in the silo the distinction stopped existing in the world rather than in the database.
That is the shape of most digital product passport programmes once they get past the schema discussion. The physical process sets a ceiling on the resolution of any claim, and no amount of software raises it. What software can do is compute a defensible figure at the resolution the process supports, and record honestly what that resolution is.
For the silo, that figure is a mass balance: opening stock and its blended attribute, receipts by supplier with their attributes, draw over the period, closing stock. Planning teams already run this calculation shape for cost, for allocation and for shelf life. Running it for a material attribute needs no new method, and it needs the receipts tagged with a supplier attribute at goods receipt, which is usually the missing piece.
What the regulation actually obliges you to do
The Ecodesign for Sustainable Products Regulation, (EU) 2024/1781, entered into force in July 2024 and sets the framework. On its own it obliges nothing about any specific product. The obligations arrive through delegated acts adopted per product group, and each one names the data carrier, the identifiers, the fields, the granularity and who is allowed to see what. The Commission's first working plan under the regulation, adopted in April 2025, set out the priority groups, textiles and apparel and iron and steel among them.
One passport already has a fixed date. The Battery Regulation, (EU) 2023/1542, requires a battery passport from 18 February 2027 for electric vehicle batteries, light means of transport batteries and industrial batteries above two kilowatt hours. That one is per individual battery, which settles the granularity question for anybody in scope and gives everyone else a preview of what item level costs.
Underneath, CEN and CENELEC's joint technical committee on the passport framework, set up in 2024, has been producing the standards the delegated acts will point at. Which means the carrier format, the registry interface and the exact schema were all still moving while firms were being told to prepare.
Granularity is the decision that costs the money
Three levels are possible and the delegated act picks, so treat this as a question about your floor rather than your preference.
Take a business with 900 finished goods, 4.2 million units a year, and an average production batch of 3,000 units, so roughly 1,400 batches a year.
At model level the passport is a property of the specification. That is 900 records, revised when the spec changes, and it is a documentation exercise with a maintenance routine attached.
At batch level it is 1,400 records a year, each one needing the actual sourcing for that batch rather than the nominal specification. That requires the batch record to carry supplier and lot for the attributes you are claiming, which most batch records do for regulated ingredients and not for anything else.
At item level it is 4.2 million records a year, each keyed to a unique identifier created at the point of pack. That ratio, roughly four thousand seven hundred to one against the model level, is the part a database can absorb. The rest lands on the warehouse floor, where you need serialisation at pack, aggregation to case and pallet so that receiving does not have to scan every unit, and a defined behaviour for repacks, splits and returns. Anybody who ran pharmaceutical serialisation under the US Drug Supply Chain Security Act knows the shape of that project. Anybody who has not tends to price a passport as a database.
The planning consequence is to build the identifier scheme to reach batch level even where the first delegated act only asks for model. Going from model to batch later means retrofitting production records that have already been written, and the records you need cannot be reconstructed after the fact.
Rolling an attribute up a bill of materials
Most passport fields at finished goods level are mass-weighted roll-ups over the bill of materials, and the roll-up breaks wherever a component has no attributed value.
Take a product weighing 1,000 grams. The housing is 600 grams at thirty percent recycled content, verified with the supplier. Packaging is 150 grams at eighty percent, also verified. The fasteners, internals and adhesive are 250 grams with nothing attributed.
Treat the unknown as zero and the answer is 180 plus 120 over 1,000, which is thirty percent. Treat the unknown as the average of what you do know, where 750 attributed grams carry 300 grams of recycled material for forty percent, and the answer is forty percent. Same product, same evidence, ten points apart, and the passport field has one box.
So carry an unattributed mass bucket explicitly through every roll-up and publish the coverage alongside the figure. Seventy five percent of mass attributed, thirty percent recycled content on the attributed portion, is a statement somebody can act on and defend.
The direction of the bias is worth deciding deliberately, because the two uses pull opposite ways. Unknown treated as zero understates the claim, which is the safe direction for a published figure you carry legal exposure on. The same default makes internal progress invisible: a team that improves attribution coverage from sixty to eighty percent will see the published number rise without any material having changed, and a team that adds a component to the bill of materials will see it fall. Track the internal target on the attributed portion and publish on the conservative one, and keep both visible so the movement can be decomposed.
Approved alternates make the attribute set time-varying
This is the part that lands on planning rather than on compliance, and it is usually noticed late.
Qualifying a second source is meant to make switching free. Once the alternate is approved, the planner switches on availability, price or lead time, and nothing downstream cares. A passport regime ends that.
Say the two qualified resins carry thirty percent and twelve percent recycled content. Switch mid production order and you have two options. Split the order into two batches with distinct identifiers and issue two passports, which is a shop floor and labelling change made under time pressure. Or issue one passport at twelve percent, which throws away the claim on the better material for the whole run.
Neither is expensive on its own, and both are expensive when they happen forty times a year without anyone having decided they should.
Two design consequences follow. The attribute set belongs on the source of supply record, keyed on item, supplier, manufacturing site and a valid-from date, rather than as a single composition field on the item master. Most item masters hold one value per item and cannot express the sentence "this item, from this supplier, since March". Changing that data model before you have published passports is a schema change. Changing it afterwards means restating documents that are already in circulation.
The other consequence is a scheduling preference nobody has told the scheduler about. Where source switching carries a passport cost, the run should consume one source per batch wherever the supply position allows, and mid-batch switches should be an exception with a reason rather than a default the system reaches for. That preference has to be encoded, because a scheduler optimising on cost and availability alone will keep making the locally cheaper choice.
Not every field goes to every reader
The passport is read by consumers, repairers, recyclers and market surveillance authorities, and the framework contemplates different fields being visible to different parties. Deciding which parties see which fields is a commercial decision, and it usually reaches the people entitled to make it far too late.
A recycler needs composition to sort and process the product at end of life. Composition is your formulation. Somewhere between those two facts is a conversation between the sustainability lead who has committed to the disclosure and the commercial director who has never seen the field list, and it happens about a week before launch unless somebody schedules it earlier.
The same tension runs up the chain. A supplier giving you a composition figure for a passport is giving you something that will be published to a defined audience, which is a different request from giving you a figure for internal accounting. Some will decline, some will offer a range, and some will want it in the contract. Put the publication permission into the supply agreement at the same time as the data request, because going back a second time costs more than asking for both at once. A separate post covers how to make that ask land.
Run the field list past commercial and legal before the pipeline gets built. Record, per field, who may see it. It is a short meeting that prevents a rebuild.
What to build while the delegated acts settle
Building to a schema that is still being drafted is wasted work. Building the parts that no version of the schema will change is not, and there are four of them.
An identifier scheme that reaches at least batch level and extends to item without a redesign. An attribute store keyed on item, source of supply and effective date, with an unknown state that is distinct from zero. Mass and material composition as maintained fields on the item master with a named owner rather than as a project extract. A roll-up that carries coverage through every level of the bill of materials.
Defer the data carrier, the public rendering and the registry integration. Those are a few weeks of work once the underlying data exists, and they are the parts most likely to be redrawn by the delegated act. The passport itself is an export format over data you either hold or do not, and the export is never the hard half.
Where this stops
Anything built against a specific schema today will need rework, and a programme that promises otherwise has not read the regulation's structure. Budget for a revision after the delegated act covering your product group is published, and keep the presentation layer thin enough that the revision is cheap.
The retention obligation is longer than planning teams are used to. A product sold for eight years with a spare parts tail needs its passport available across that whole period and beyond it, which makes the attribute store a long-lived record system. The usual planning data lifecycle, where detail is archived after two years and aggregated after five, is the wrong policy for it, and the supplier who provided a composition figure may no longer exist when somebody queries it.
The honest limit is that a passport improves no decision on its own. It is a disclosure obligation, and it pays back only where the attribute store underneath it also feeds specification and sourcing choices. Where it does not, you have built a compliance artefact with a permanent maintenance cost and no operational return, which is a legitimate thing to do under a legal requirement and worth naming as such in the business case.
And the ceiling from the opening still applies. Where your process commingles material, the resolution of the claim is fixed by the physical flow, and the only way to raise it is segregation, which costs storage, changeovers and yield. A separate post works through what segregated flows cost to plan, and that trade should be decided on its own merits rather than as an afterthought to a reporting project.
Pick one finished good this week, roll its bill of materials, and count the share of its mass you can attribute a verified composition to today. Everything above is a consequence of that percentage.
More on this in sustainability on the DemandLabs platform.
