Rules of Origin: Why the Same Product Qualifies Under One Agreement and Fails Under the Next

Updated: August 5, 2026 • Audience: exporters, manufacturers, trade compliance teams, customs brokers

Most trade compliance teams spend their tooling budget on HS classification, while the duty exposure actually sits in origin. An HS code answers a single question about what the good is. Origin depends on where every input came from, what you did to those inputs, what you paid for them, and which agreement you're claiming under. A wrong answer means repaying duty you claimed relief on, potentially across every consignment for the last three to five years, with interest, on a product line you've been shipping in good faith the whole time.

Every free trade agreement publishes its origin rules in an annex, so the source material is freely available to anyone who wants it. The difficulty is that each agreement writes its own rule for the same product, drafted in a different HS edition, offering a different menu of qualifying routes, and measured against a different valuation basis.

Let's make that concrete with one product.

1. What a single rule actually contains

Take HS 2801.10, chlorine. The EU-Japan Economic Partnership Agreement gives it the rule below. Click through the routes and the process names to see how it's built, reproduced exactly as the annex sets it out.

The punctuation here carries legal weight. Semicolons separate the conditions and a final "or" closes the list, which makes this an ANY OF rule. You qualify by satisfying any single one of the four conditions, and you choose which one you rely on.

In practice that turns origin determination into a routing problem. For a given production run you work out which of the four doors your available facts can open, then make sure you can keep it open under verification. Each door draws on a different part of your business:

Those four routes sit with four different teams, which is the main reason origin work stalls inside an organisation. No single person holds all the facts, and until somebody decodes the rule, nobody knows which facts to go and collect.

2. What MaxNOM actually means

MaxNOM stands for Maximum value of Non-Originating Materials, and it sets a ceiling on your foreign content:

Value of non-originating materials ÷ Ex-works price ≤ 50%

Three details in that formula cause most of the failures we see.

The denominator is the ex-works price, meaning the price of the good at your factory gate, before inland haulage, port charges, freight and insurance. Teams routinely divide by the sales invoice they issued to the customer instead, which is a larger number and produces a ratio that looks comfortably compliant when the legal calculation would have failed.

Originating status is a legal test rather than a question of geography. A material bought inside the partner country counts as originating only if it satisfies the agreement's rules in its own right and you hold a supplier's declaration confirming it. Without that declaration it goes into the numerator alongside everything else you imported.

Anything whose origin you can't evidence counts as non-originating, so incomplete supplier records translate directly into a worse ratio.

The value routes assume you actually produced something

Every route in the annex sits underneath a general provision that most teams read once and then forget about. In the EU-Japan agreement it's Article 3.4, and it opens with the word "notwithstanding", which is the drafting signal that it overrides whatever came before it. A good doesn't originate if the only things done to the non-originating materials come off a listed set of operations: changes of packaging, breaking up or assembling packages, washing and cleaning, sorting and grading, affixing marks and labels, simple mixing, simple dilution, simple assembly of parts, and around ten more. The agreement then defines "simple" as any operation needing neither special skills nor machines or equipment specially produced or installed for the purpose.

That override lands on the value routes exactly as hard as it lands on the tariff shift, and MaxNOM is where it catches people out, because a comfortable ratio can come entirely from your margin rather than from anything you did to the goods.

The same logic reaches the tariff shift route as well, because for some goods repacking from bulk into retail presentation genuinely does move the product into a different subheading. Meeting the letter of the annex entry through packaging alone gets you nowhere under either route. The value tests and the shift tests both assume real production sits behind them, and the insufficient operations article is what enforces that assumption.

3. What RVC actually means

RVC stands for Regional Value Content, and it approaches the same question by setting a floor on regional content instead of a ceiling on foreign content. The EU-Japan rule uses the build-down method:

(FOB value − Value of non-originating materials) ÷ FOB value ≥ 55%

Build-down starts from the finished good's full value and subtracts the foreign content. The build-up method starts at zero and accumulates qualifying originating material costs, and the two generally produce different numbers from the same set of books. USMCA adds transaction value and net cost as further bases again.

The two value routes are easy to treat as interchangeable, though their denominators differ in a way that matters. MaxNOM measures against the ex-works price while RVC measures against FOB value, and FOB is always the larger figure since it covers moving the goods to the port and loading them. The calculator below lets you move the figures around and watch the two tests separate.

Where inland transport to port makes up a meaningful share of value, which is common for bulk chemicals and heavy industrial goods travelling a long way to a seaport, the spread between EXW and FOB gets wide enough to change the outcome. Since the rule reads any of, one working route is enough to claim preference, but establishing which route works means running both calculations.

4. Processes carry legal definitions

Route [2] attracts the most over-claiming, because it feels intuitive. Material went through your plant, so something must have happened to it. Each named process in the annex carries its own definition though, and those definitions run narrower than the ordinary English words suggest.

• Purification carries a threshold. You need to eliminate at least 80% of existing impurities, and you need input and output assays that demonstrate it.
• Chemical reaction explicitly excludes dissolving in water or other solvents, eliminating solvents, and adding or removing water of crystallisation, which happen to be the operations that look most like chemistry to anyone outside the lab.
• Production of standard materials requires a manufacturer's certificate of purity or proportion.
• Isomer separation requires analytical data such as chromatography, showing that isomers were genuinely isolated from a mixture.

The gap this opens up is rarely deliberate. A plant manager confirms that the site purifies the material, meaning the ordinary English word, while the annex means the 80% test, and both readings sit unreconciled in the file until a verification request arrives eighteen months later.

5. Now change the agreement

Everything above describes one rule in one agreement. The same chlorine, physically identical in every respect, meets a different legal test under every other agreement you might claim preference under.

Working through them shows the scale of the problem:

• The number of routes ranges from two to eight. The PEM and EU-South Korea rules offer no specified-process route at all, so your chemistry is legally irrelevant there and only codes and values count. At the other end, USMCA opens five process routes to this product through its Section VI notes, on top of a tariff shift and two RVC methods.
• The processes that qualify vary widely. UK-EU TCA lists seven, including mixing and blending, change in particle size, and biotechnological processing. EU-Japan lists four. EU-Mercosur lists exactly one. CETA tucks its two into notes at the head of the chapter and tightens them, since a chemical reaction only counts there if the CAS number changes. A purification step that confers origin on a UK-EU shipment does nothing for you under Mercosur.
• The tariff shift level changes. Japan, Mercosur and CETA want a subheading shift (CTSH, 6-digit), while PEM and the standard tier of the UK's DCTS want a heading shift (CTH, 4-digit). The 4-digit version is materially harder, since more of your inputs will share a 4-digit heading with the finished product.
• The valuation basis changes four times. Ex-works, FOB, transaction value and net cost all appear across these ten rulebooks, and a percentage means nothing until you know which base carries it. A 50% cap on ex-works price and a 50% cap on transaction value are different tests on the same shipment.
• The HS edition changes. These rules are drafted on HS2007, HS2012, HS2017 and HS2022, while your product is classified in HS2022. Applying the EU-South Korea rule means first working out which entry in an HS2007 list governs a code you hold in the current nomenclature. The next section covers what that involves.
• Sometimes the code alone won't get you there. Under the UK's Developing Countries Trading Scheme the rule that applies depends on the beneficiary country's tier. The Comprehensive Preferences tier gets a 6-digit shift, a 75% ceiling, or the same seven-process list the TCA uses, while the Standard and Enhanced tiers get a 4-digit shift and a 50% ceiling with no process route, so the product-specific rule can't be resolved from the HS code by itself. The EU's GSP is a separate scheme carrying its own thresholds, and the two shouldn't be read as one.

Currency matters here as much as content. The PEM Convention was revised with effect from 1 January 2025, the old rules ran alongside the revised ones through 2025, and since 1 January 2026 only the revised rules apply wherever an agreement carries a dynamic reference to the Convention, which EU-Switzerland does. An origin answer built on the superseded PEM list would have been right in 2024 and wrong today, without anything about the product having changed.

6. The rules are frozen in older HS editions

The World Customs Organization revises the Harmonized System every five or six years. The editions in circulation are HS2002, HS2007, HS2012, HS2017 and HS2022, and the next revision lands in 2028, with the WCO correlation tables for it already published. Your declaration uses whichever edition is in force at the time, so today that's HS2022. The annex you have to satisfy uses whichever edition was current when it was drafted, and it stays on that edition until the parties formally agree to move it.

For the ten agreements above, that leaves you working across four editions at once:

Agreements can be re-based, and the UK-EU TCA is the example here. Its product-specific rules annex was transposed onto HS2022 by a Partnership Council decision adopted in November 2024 and in force from early 2025, which is why it sits in the current edition while agreements concluded around the same time are still on HS2017. Re-basing happens by negotiation rather than automatically, so the spread of editions widens over time until somebody does the work to close it.

Chlorine is an easy case in this respect. Subheading 2801.10 has carried the same scope through every edition in that table, so it maps one-to-one in all four directions and the correlation step is invisible. That's what makes it a clean example for showing the rest of the machinery, and it also understates how much work correlation is for products that moved.

Correlation gets difficult in three situations:

• A subheading splits. One legacy code becomes several current ones. Your HS2022 code then maps back to a single older entry that also covers products you no longer ship, and the rule you inherit was written with that wider scope in mind.
• Subheadings merge. Several legacy codes collapse into one current code, so a single HS2022 code corresponds to more than one entry in the old list, and those entries can carry different rules. You have to establish which one governs your specific good rather than picking either.
• The code is genuinely new. HS2022 created subheadings that had no predecessor, including heading 88.06 for unmanned aircraft and 85.49 for electrical and electronic waste and scrap. An FTA list drafted on HS2017 contains no entry naming them at all, so you have to work out which older entry the product would have fallen under and apply that rule instead.

None of this is a one-off lookup you do at onboarding and then forget. It's a legal step in the determination, it has to be reproducible, and a verifying authority can ask you to show your working on it years later. When HS2028 lands, every agreement still sitting on HS2017 or HS2012 moves one revision further away.

7. The product-specific rule is only one condition

Satisfying the annex entry doesn't make your good originating on its own, and this is the part most origin tooling passes over. The agreement's general provisions, known as the horizontal rules, sit above the product-specific rule and can override it in either direction.

Alongside these sit rules on sets, packaging, accessories and spare parts, fungible materials and accounting segregation, neutral elements, returned goods, and the unit of qualification. Fifteen separate horizontal provisions govern the EU-Japan entry above, and eight introductory notes govern how its annex is read, so an origin answer that cites only the annex row has covered a fraction of what decides the claim.

8. Why this gets expensive by hand

The workload for one product under one agreement runs roughly like this. Find the governing entry in the right HS edition, correlate your code back if the edition differs, parse an ANY OF rule into its alternative routes, work out which routes your available facts can support, get costing to produce a value calculation on the correct basis, get the plant to produce process evidence against a legal definition, check every horizontal provision, then assemble a record that will still stand up to verification years later.

Multiply that by your catalogue and again by the number of agreements you trade under. A mid-sized exporter with 400 SKUs and eight agreements is carrying 3,200 determinations, each of which goes stale when an agreement is amended, a supplier changes, a cost moves, or the HS nomenclature gets revised.

The rational response, and the one most teams land on, is to determine origin once for the highest-volume lanes, document it thinly, and roll it forward until something forces a review. Verification is usually what forces it.

9. How Sail approaches it

Sail treats origin as a structured determination rather than a document for somebody to read.

Rules are decomposed into their criteria. Every product-specific rule is broken into its alternative routes, and each route into the criteria it imposes: the shift level, the named processes with their definitions and carve-outs, and the value limits with their basis and subject. MaxNOM 50% EXW and RVC 55% FOB are modelled as distinct objects with distinct denominators, so the system evaluates them separately.

Every rule is pinned to its source. Each entry carries the file, locator and checksum of the legal instrument it came from, together with the HS edition it was drafted on. When an HS2022 code has to be matched against an HS2007 list, the correlation is explicit and visible on the record.

Unresolved determinations come back as unresolved. Where a rule depends on facts outside the product, such as a beneficiary tier, a declared process, or a cost you haven't supplied, Sail returns the determination with the missing facts named. The classification and origin models are fine-tuned on your own data and built to surface that gap, since a confident wrong answer costs considerably more than a flagged one.

Horizontal provisions travel with the answer. Tolerance, insufficient operations, cumulation, non-alteration, drawback and the proof-of-origin obligations attach to the determination, flagged wherever they could block a claim that the annex row alone would have allowed.

The reasoning is stored with the result. Each determination records which route was taken, which facts supported it, which alternatives were evaluated and rejected, and which legal source governed the decision, so a verification request three years later gets answered by retrieval rather than reconstruction.

Where to start

To gauge your current exposure, take your three highest duty-saving product lines and ask two questions about each. Which route did we claim under, and can we produce the evidence that route requires today? In our experience the first question frequently can't be answered from existing records at all, and that on its own is the finding worth acting on.

If you'd like to walk through your own product mix and agreement coverage, Portmind's team is happy to go through it with you.