Origin verification: the audit arrives two years after the shipment

A preference claim is unusual among export obligations in that nothing checks it when it matters. The goods move, the reduced rate is applied, the shipment is delivered and paid for, and the question of whether the claim was sound remains open for years.

It is closed by a verification, and a verification is a conversation about a shipment nobody involved remembers.

Why the delay is structural

Under CPTPP the certification of origin is made by the exporter, producer or importer, with no prescribed form and no third-party authentication. The agreement pairs that with streamlined procedures establishing obligations for importers, exporters and producers, and those obligations are the other half of the bargain: you certify without pre-approval, and in exchange you can be asked to prove it afterwards.

So the checking has not been removed. It has been moved, from a moment when everybody has the facts to hand to a moment when the people who had them have moved on.

That is a reasonable design and it has one practical implication that dominates everything else in this article: **the file has to be built when the claim is made, because it cannot be rebuilt later**. A company that knew exactly why a product qualified in 2026 and cannot show it in 2028 is in the same position as one that never checked.

Who gets asked, and who pays

The money lands on the importer. They claimed the preferential rate on their entry; if the claim fails, they are assessed for the duty foregone, usually with interest and often with a penalty.

The information sits with the exporter or producer. The importer cannot answer a question about your bill of materials, and a verification that starts with them frequently arrives at your door within a week.

The commercial consequence follows from that split and is worth stating plainly. Your buyer pays first and then looks at you, having relied on a certification you made or on assurances you gave. Whether they have a legal claim depends on the contract; whether the relationship survives depends on how the conversation goes, and it goes badly when the answer is that the records no longer exist.

A verification can also be addressed to the exporter or producer directly, depending on the procedure the destination follows. In that case you are answering a foreign customs authority about your own production, in a correspondence with deadlines, and the quality of your file is the whole of your position.

What is actually asked for

Verifications vary, but the questions do not, because there are only so many ways a claim can be tested.

**What is the product, and how is it classified.** The rule of origin is written against the classification, so the first thing tested is whether the right rule was applied. A classification that turns out to be wrong invalidates a claim even where the goods would have qualified under the correct rule.

**Which origin criterion was relied on.** You stated one on the certification. The verification asks you to demonstrate it.

**Where did the inputs come from.** For anything other than a wholly obtained good, this means the bill of materials with origins, and supporting declarations from suppliers for the material claimed as originating.

**Show the calculation.** Where a value rule applies, the calculation in the prescribed method, reproducible from records rather than reconstructed.

**Prove this shipment was that product.** The gap that catches sound files. Demonstrating that a product qualifies is not the same as demonstrating that the consignment on a particular entry, on a particular date, was made to that formulation. Lot traceability closes it, and it usually lives in a different system from everything above.

How claims fail that were originally sound

Very few verifications find a company that lied. Most find a file that decayed.

**The formulation changed and the claim did not.** A supplier substitution in year two, made for cost or continuity, changes the inputs the qualification rested on. Nothing prompts a review, because procurement is not looking at the origin file and the origin file is not watching procurement.

**Supplier declarations expired.** A declaration covers material supplied in a period. A folder of historical declarations does not support this year's shipments, and the gap is invisible until somebody checks dates.

**A blanket certification outlived its accuracy.** A single certification covering repeat shipments over a period is efficient and removes the prompt that a new document would have provided. Every shipment after a change is covered by a statement that is no longer true.

**The person left.** The analysis lived in an individual's understanding and their email, not in a company record with a location and an owner.

**Records were kept but not linked.** Everything exists somewhere and nothing ties it together, so answering takes weeks and produces something that looks assembled after the fact, because it was.

What a defensible file looks like

Built at the time, held at company level, and reviewed on a trigger, not on a date:

• The classification of the finished product, and the basis for it
• The specific rule of origin applied, quoted rather than summarised
• The bill of materials with the origin of each input
• Classification of every non-originating input, where a classification-change rule applies
• Current supplier declarations for material claimed as originating
• The value calculation in the prescribed method, where one applies
• The certification itself, with its blanket period if it has one
• Production and lot records tying shipments to the formulation
• A change log: what changed, when, and what was re-checked as a result

The last item is what turns a folder into a file. It is also the cheapest thing on the list and the one most often absent.

Triggers that should prompt a re-check

Rather than reviewing everything annually, which becomes a ritual, tie the review to the events that actually invalidate a claim:

• Any formulation or recipe change, however small
• Any change of supplier for an input, including a second source added alongside a first
• Any change in the classification of the finished product
• Renewal or expiry of a blanket certification period
• A supplier declaration reaching the end of its period
• A change to the rules themselves, or to the agreement's schedules

Each of those is a moment when somebody outside the origin file does something that changes the answer inside it. Building the prompt into the process is what stops the claim drifting away from the product.

The file is cheaper than the alternative, and it compounds

It is worth sizing this honestly instead of urging diligence in the abstract.

Building an origin file for a product is a few days of work: classify the finished good, classify the imported inputs, collect supplier declarations, run the calculation if one is needed, write down the reasoning. For a company with a handful of exported lines that is a project, not a programme.

The alternative is an assessment covering every shipment since the defect began, plus interest, plus a penalty, landing on a buyer who then looks at you. The exposure grows with volume and with time, which is the opposite of how most risks behave: the longer a sound-looking claim runs unexamined, the larger the consequence of it having been wrong.

The file also compounds in the useful direction. Most of what it contains, the classification, the bill of materials with origins, the supplier declarations, is the same material a destination's registration file wants, the same material a halal certifier wants, and the same material a buyer's technical team asks for. A company that has built it once for origin has most of what the rest of the market-entry work needs.

That is the honest argument for doing it properly, not a warning about audits. The audit is the reason it has to be true. The reuse is the reason it is worth the few days.

What to do if a verification arrives

• Answer it, promptly, within the deadline given, because a non-response is generally treated as a failed claim
• Tell your buyer immediately if it came to you, and ask to be told immediately if it came to them
• Produce the file as it stands, and not an improved version of it, since a file assembled after the question is visibly a file assembled after the question
• If the claim was wrong, establish how far back the error goes, because the same defect usually affects every shipment since the change that caused it
• Fix the underlying process before fixing the correspondence, or the same verification happens again on the next audit cycle

The last point is the one worth internalising. A verification is expensive once and survivable. The version that damages a business is the one that reveals a process which never had a mechanism for noticing it had gone wrong, because that finding applies to everything shipped since.