What Customs Data Actually Reveals About the Counterfeit Watch Supply Chain

Every year, customs authorities publish tables of what they caught. Journalists read those tables as a story about crime. Supply chain people should read them as a story about parcels.

I spent a week pulling apart the two most authoritative public datasets on this trade — the OECD’s 2025 sector study of counterfeit Swiss watches, and US Customs and Border Protection’s annual intellectual property seizure reporting — and the picture that emerges is not the one the headlines describe. The counterfeit watch trade is not primarily a manufacturing achievement. It is a distribution architecture that happens to move watches. And the architecture is visible in the data if you know which columns to look at.

The scale, and the caveat that has to come first

Start with the top line. In fiscal year 2025, US Customs and Border Protection seized nearly 79 million counterfeit items. Had those goods been genuine, CBP estimates their combined manufacturer’s suggested retail value would have exceeded USD 7.3 billion. For context, the equivalent value was around USD 5.4 billion in FY2024 and roughly USD 2.8 billion in FY2023 — the trend line is steep.

Now the caveat, because it changes how you should read everything that follows: seizure data measures enforcement, not trade. Every table in every report on this subject is a record of what inspectors found and chose to act on, at the ports where they had capacity, using the targeting criteria they were running that year. It is a biased sample of a population nobody can observe directly. The OECD says as much in its own methodology notes, and treats its global estimate as a lower bound for exactly this reason.

Keep that in mind, because the numbers below are still extraordinarily useful — just not as a measure of market size.

One product category absorbs almost everything

The most striking figure in the OECD’s 2025 report on counterfeit goods infringing Swiss intellectual property is a concentration that borders on implausible.

Watches accounted for 87 percent of all seizures of counterfeit goods infringing Swiss IP rights in 2020–21. By value, they accounted for 94 percent.

Read that again. Of every counterfeit item seized worldwide that infringed a Swiss brand, roughly seven in eight were watches, and their share of seized value was even higher — because watches are dense with value per unit. Jewellery, leather goods, ready-to-wear, electronics and packaging make up the remaining six percent of value between them.

The dollar figure attached to this: fake watches infringing Swiss trademarks were worth an estimated USD 1.88 billion in 2021, more than 40 percent of the total value of all counterfeit goods infringing Swiss IP that year. That is equivalent to about 7.7 percent of Switzerland’s legitimate watch export value for the same period.

And it is not a rounding error in the industry’s economics. The same study estimates Swiss watchmakers lost approximately USD 1.08 billion in sales, equivalent to 4.2 percent of sector revenue, with around 2,537 jobs affected — 4.5 percent of industry employment.

If you work in any other consumer category, this concentration should be the first thing that interests you. One product class, one country’s brand portfolio, absorbing nearly the entire enforcement surface. That is not normal. It is a signal that something structural is happening in this market specifically.

The mechanism is postal, and it is small

Here is where the data stops being about crime and starts being about logistics.

The OECD breaks down how counterfeit goods infringing Swiss IP actually moved in 2020–21:

Transport mode	Share of seizures
Post / mail	75%
Air	16%
Road	5%

Three quarters of everything moved through the postal system. Not container freight, not air cargo with a manifest and a broker and a bill of lading. Mail.

Now the shipment-size distribution, which is the more revealing table:

Items per consignment	Share of seizures
Single item	~75%
Fewer than six items	90%

Nearly three quarters of all seizures contained exactly one item. Nine in ten contained fewer than six.

This is the whole architecture in two rows. A single-item consignment sits below most de minimis thresholds, requires no formal customs entry, generates no broker record, and cannot be economically inspected at the volumes involved. At roughly 1.3 billion de minimis shipments processed by CBP in FY2024 alone, no agency on earth has the inspection capacity to sample that population meaningfully.

The counterfeit watch trade did not defeat border enforcement by being sophisticated. It defeated border enforcement by being small enough that inspecting it is a losing proposition on a per-unit basis. The supply chain was fragmented precisely because fragmentation is the evasion strategy, and it worked.

This is also the reason the de minimis reform debate currently running through US and EU trade policy is not an abstract customs question. It is a direct response to this data.

Where it comes from, and why the map misleads

Origin data is where most reporting goes wrong, so it is worth being precise.

By count of seizures in 2020–21, the provenance economies for fake Swiss watches were:

• China: approximately 54%
• Hong Kong (China): approximately 11%
• Netherlands: approximately 11%

By value, China and Hong Kong together accounted for 84% of the total estimated value seized.

Zoom out to all counterfeit goods infringing Swiss IP, and the pattern holds: China and Hong Kong together represented 82 percent of seized value and 63 percent of case count, with goods traced to 75 different provenance economies in total.

That Netherlands figure is the one to be careful with. The OECD explicitly flags it as a transit artefact — the country appears as both source and destination because provenance data records the last economy a shipment passed through, not where the goods were made. Rotterdam and Schiphol are European distribution nodes. Reading “11% Netherlands” as “11% manufactured in the Netherlands” would be a straightforward misreading of the column.

This distinction between count-share and value-share matters for a second reason: high concentration by value with much wider dispersion by count describes a business that is organised at the top and atomised at the bottom. A small number of coordinating entities, a very large number of individual parcels. Which is exactly what the shipment-size data already told us.

The consumer is the shock absorber

One structural detail deserves more attention than it gets.

When a counterfeit consignment is intercepted at a border, three parties are involved: the seller, the buyer, and the state. The seller has already been paid. The state has the goods. The buyer has nothing.

The OECD notes this explicitly in its analysis of why e-commerce is attractive to counterfeiters: with online sales, the financial burden has shifted — when goods are seized at the border, the consumer who has already paid is the one who bears the loss.

There is no recourse path that survives this. A chargeback requires demonstrating non-delivery by the merchant, and “customs seizure” is a defence the merchant will use. A warranty requires a legal manufacturer, and there isn’t one. Insurance is not available for an illegal purchase. The entire downside of a failed shipment is externalised onto the end customer, who has no information about interception rates at the time of purchase and no way to price that risk.

From a supply chain standpoint this is a remarkably efficient arrangement: the party with the least information carries the most risk. It is also the mechanism that keeps the model profitable despite what must be a meaningful interception rate.

What the data cannot tell you

This section exists because the honest limits of these datasets are more interesting than most of the claims built on top of them.

Seizure data reflects enforcement priority, not trade volume. A jump in watch seizures might mean more fake watches, or it might mean CBP retargeted its risk models that year. There is no way to separate these from the published tables.

Origin data records transit, not manufacture. Already covered, but it recurs in every secondary summary of these reports, which is how a mapping error becomes a widely repeated fact.

Category aggregation destroys precision. There is no reliable independent estimate for counterfeit jewellery specifically, because customs classification systems routinely group jewellery with watches, accessories, or leather goods depending on the jurisdiction. Any figure you see quoted for one of those categories in isolation should be treated as an estimate derived from an aggregate, not a measurement.

Year-over-year swings are dominated by individual events. A handful of large intercepts can move a category’s value share for an entire year. Single-year comparisons across categories are close to meaningless without the underlying case distribution.

And the headline global figure is a model output, not a count. The USD 467 billion estimate for global trade in counterfeit and pirated goods — roughly 2.3 percent of world imports — comes from the joint OECD–EUIPO Mapping Global Trade in Fakes (2025), which extrapolates from observed seizures to an estimated total using a statistical model. It is the best number available. It is not a measurement, and the OECD is careful to say so.

Where the constraint is actually moving

If you read the enforcement data as the primary risk to this trade, you are reading it wrong. Seizures are a cost of doing business — a percentage shaved off every shipment, priced into the retail markup.

The constraint that actually bites is financial. A seized parcel is a one-off loss. A terminated merchant account is an existential event, and acquiring payment processing is the hardest problem in any business that cannot pass a standard underwriting review. This is why the enforcement energy with the most leverage right now is not customs at all — it is platform liability rules, de minimis legislative reform, and payment network pressure.

None of those three show up in a seizure table. All three determine whether the businesses behind the parcels still exist in three years.

The takeaway

If you take one thing from this data, take the shipment-size distribution.

The counterfeit watch trade looks, from the outside, like a manufacturing story: factories producing convincing fakes, sold at a fraction of retail. The seizure data says it is a logistics story: a fragmented, postal-routed, single-unit distribution network that is profitable precisely because inspecting it at scale is economically irrational.

That distinction has practical consequences. It explains why enforcement keeps producing record seizure values without reducing the trade. It explains why the policy conversation moved to de minimis thresholds. And it explains why the retail side of this market — where buyers actually assess what they are getting and what it will be worth in three years — has become the place where the real information asymmetry sits. The people buying into this segment rarely have the data to price the risk they are taking on.

Which is, in the end, the same problem the customs tables describe from the other direction.