What is FTA origin determination, and why is it harder than HS classification?

By Saurabh Goyal, Founder & CEO of Phlo Systems. Published 7 October 2026.

Ask most import teams how confident they are in their HS classifications today, and the answer has improved. AI classification tools read a commercial invoice or product description, propose a commodity code, and show the reasoning behind it, and that has taken a slow, specialist task and made it fast and checkable. Ask the same team how confident they are that their products actually qualify for the free trade agreement (FTA) preference rate they are claiming, and the answer gets quieter. That second question is a different problem, and it is the one that is still mostly unsolved.

The 30-second answer:

• HS classification answers "what is this product," and assigns it a commodity code. It is increasingly automatable, because the input (a product description) maps to a fixed, published code list.
• Origin determination answers "does this product qualify for a preferential rate under a specific FTA," and the answer depends on where every material input came from, how much value was added where, and the exact rule text of the agreement being claimed. It cannot be solved by looking at the finished product alone.
• A product can be correctly classified and still fail to qualify for preference, because qualification is a separate test applied after classification, not a consequence of it.
• A product can qualify under one FTA and fail under another, for the same shipment to the same country, because different agreements set different origin rules for the same commodity code.
• Getting this wrong does not just mean a rejected declaration. A denied preference claim can mean duty assessed at the full rate retroactively, and in some regimes, penalties for an unsupported claim.

Why classification got easier and origin did not

HS classification is, at its core, a matching problem. There is a finite, published nomenclature, a product has physical and functional characteristics, and the task is to find where in that nomenclature the product belongs. A large language model that has read enough tariff schedules, explanatory notes, and classification rulings can do a credible first pass at this matching problem, and increasingly does. That is why AI-driven classification has moved quickly from novelty to standard tooling across customs software.

Origin determination is not a matching problem. It is a tracing and arithmetic problem, applied against rule text that differs by agreement, and it starts only after classification has already happened. To determine whether a finished product qualifies for preference under a given FTA, you need to know, for every material and component that went into it, where that input came from, what it cost, and in some cases what it was classified as before assembly. None of that is visible from looking at the finished good. It lives in the bill of materials, the supplier invoices, and the production record, and it changes every time a supplier or a sourcing mix changes, even if the finished product's classification does not.

What origin determination actually involves

Qualifying a product for an FTA preference generally comes down to applying one or more of three kinds of rule, set out in the specific agreement being claimed:

1. Wholly obtained or produced. The simplest case: the product is entirely grown, mined, or manufactured in the exporting country with no foreign inputs at all. Rare for manufactured goods, common for some agricultural and raw-material trades.
2. Value-added (regional value content) thresholds. The agreement sets a minimum percentage of the product's value that must originate in the exporting country or trading bloc, calculated under a defined method (often based on the ex-works price minus the value of non-originating materials, though the exact formula varies by agreement). Get the calculation method wrong, or miss a non-originating input, and a product that looks comfortably over the threshold can actually fall short.
3. Tariff shift (change in tariff classification) rules. Instead of, or alongside, a value test, the agreement requires that non-originating inputs be classified under a different tariff heading or chapter than the finished product, reflecting that genuine manufacturing, not just assembly or relabelling, took place in the exporting country. Applying this rule means classifying the inputs as well as the finished good, and checking the shift against the specific rule for that product's tariff line, not a general rule of thumb.

Many agreements also allow cumulation: treating inputs sourced from certain other countries (often other members of the same trade bloc, or countries with their own agreement with the exporting country) as if they originated in the exporting country for the purposes of the value or tariff-shift test. Cumulation can turn a product that fails the origin test on a standalone basis into one that passes, but only if the specific agreement permits cumulation with the specific country the input came from, and the paperwork proves it.

Every one of these rules is set at the level of the individual FTA, often down to the specific tariff line within it. There is no single, generic "rules of origin" standard that applies across agreements. The rule for a textile product under one agreement can look nothing like the rule for the same tariff heading under another.

Why origin determination is commonly got wrong

A handful of mistakes account for most of the bad preference claims we see described across the industry:

• Treating "shipped from" as "originates in." The country a shipment departs from, or is invoiced from, is not the same thing as its country of origin for preference purposes. A product assembled in one country from components sourced elsewhere can easily be shipped via a third country, and none of that changes what the origin rule actually requires: tracing the inputs, not the shipping route.
• Applying the wrong FTA's rules, or the right FTA's rules to the wrong tariff line. An importer who knows "we have an agreement with this country" sometimes applies a generic sense of what that agreement requires rather than the specific rule text for that product's own tariff heading, which can set a materially different value threshold or tariff-shift requirement than a neighbouring line in the same chapter.
• Assuming a product that qualifies under one FTA qualifies under all of them. It does not. The same finished product, exported to two different markets under two different agreements, can pass the origin test for one and fail it for the other, because the two agreements set different thresholds, different cumulation rules, or different qualifying processes for the same commodity. Each claim has to be tested against its own agreement, every time.
• Missing a change in sourcing. An origin determination made once, at product launch, goes stale the moment a sourcing mix changes. A supplier switch for a single component can move a product from qualifying to not qualifying under a value-added or tariff-shift test, and that change is invisible unless someone re-runs the test.
• Treating the classification as the origin answer. Classification tells you which tariff line's origin rule applies. It is the input to the origin test, not the output. A correct commodity code and an unqualified preference claim can coexist on the same declaration.

Why getting this wrong matters beyond a rejected declaration

A classification error on a declaration is usually caught and correctable: the authority queries it, or the declaration is rejected, and it gets refiled. A wrong origin claim often behaves differently, because the preference has already been granted at the point of import, based on the importer's or exporter's assertion that the goods qualify.

If that assertion is later found to be unsupported, typically on a post-clearance audit or a verification request from the importing country's customs authority, the consequence is not a bounced declaration. It is duty reassessed at the full, non-preferential rate, applied retroactively across however many shipments the unsupported claim covered, and in many regimes, the possibility of penalties on top of the duty itself for making a preference claim without adequate support. That exposure can sit quietly for years before a verification request or an audit surfaces it, which is a very different risk profile to a declaration that simply gets rejected and refiled within the week.

What a serious origin determination process looks like

1. Treat origin as a per-agreement, per-product test, not a one-time company-level fact. "We qualify for preference with this country" is not a complete sentence; it needs a specific agreement and a specific product behind it.
2. Keep the bill of materials and supplier origin data current, and re-test when sourcing changes, not only when a customer or auditor asks.
3. Apply the actual rule text for the specific tariff line, not a general sense of the agreement, and document which rule (wholly obtained, value-added, tariff shift, or cumulation) the claim relies on.
4. Keep the proof of origin on file in the form the specific agreement requires, whether that is a supplier declaration, a certificate of origin, or an exporter's own records, ready for a verification request rather than reconstructed after one arrives.

Frequently Asked Questions

Is FTA origin determination the same thing as HS classification?

No. Classification assigns a product its commodity code based on what the product is. Origin determination is a separate test, applied after classification, that asks whether the product qualifies for a preferential duty rate under a specific trade agreement, based on where its inputs came from and how much value was added where. A product can be classified correctly and still fail to qualify for preference.

Does the country a product ships from determine its origin for FTA purposes?

No. The country of origin for preference purposes depends on where the product (and its qualifying inputs) were produced or sufficiently transformed, not on which country the shipment physically departs from or is invoiced through. A product can ship from a country that has nothing to do with its origin determination.

Can a product qualify for one FTA but not another?

Yes, and this is common. Different trade agreements set different origin rules, value-added thresholds, tariff-shift requirements, and cumulation allowances, for the same commodity code. A product exported to two markets under two different agreements has to pass each agreement's own origin test independently; qualifying under one says nothing about the other.

What happens if a business claims an FTA preference it does not actually qualify for?

The risk usually surfaces later than a classification error would. Because the preferential rate is typically granted at import based on the claimant's assertion, an unsupported claim found on a later audit or verification request can mean duty reassessed at the full rate retroactively across the shipments covered, plus potential penalties in many regimes, rather than a declaration simply being rejected and refiled.

What are the main types of rules of origin used in FTAs?

Three kinds, usually set out per product in the specific agreement: wholly obtained or produced (no foreign inputs at all), value-added or regional value content thresholds (a minimum percentage of value must originate in the exporting country or bloc, calculated under the agreement's defined method), and tariff shift rules (non-originating inputs must be classified under a different tariff heading or chapter than the finished product). Many agreements also allow cumulation, treating inputs from specified other countries as originating, subject to that agreement's own conditions.

How Phlo Systems helps

tradePhlo Intelligence checks duty and FTA eligibility as part of classifying the goods, tracing the specific agreement's rule, value-added threshold, or tariff-shift requirement against the product's own tariff line rather than applying a generic sense of what an agreement covers. tradePhlo Declarations then carries that classification and preference data straight into the CDS, NCTS6, or NL DMS declaration, so the origin basis for a preference claim is documented at the point of filing, not reconstructed after a verification request arrives. See tradePhlo Intelligence.

---

Related reading:

• What is the UK Customs Declaration Service (CDS), and how do you submit an import declaration in 2026?
• What is landed cost, and why is my customs broker's number always wrong?

Saurabh Goyal is the Founder & CEO of Phlo Systems. He spent 12 years building CTRM and ERP systems for global commodity trading houses before founding Phlo in 2016.