Trade-based money laundering moves value across a border by misstating the price, quantity, quality, or description of goods in a trade that is otherwise real. The shipment may genuinely sail. The invoice may genuinely be paid. What differs is one documented fact, and the difference between the documented trade and the physical one is the transfer.
This is worth stating carefully, because the usual framing, that criminals “hide money in trade”, makes it sound like concealment. Nothing is hidden inside the container. What is being exploited is that two institutions record the same shipment differently, both correctly by their own standards, and that no process compares the two records. The value is not smuggled past a control. It is created by the disagreement between records that each control accepts.
The mechanism, in one transaction
An exporter in country A ships goods to an importer in country B and invoices $184,200. The importer pays. Everything about the payment is ordinary: correct beneficiary, plausible amount, commercial reference, a documentary credit behind it.
Now suppose the goods were worth $133,440 at any defensible market price. The exporter has received $50,760 more than the shipment justifies, and the importer has moved that amount abroad without a single transaction that looks like a transfer of value. Reverse the direction and you get the mirror case: under-invoice an import, and value stays with the buyer in the destination country.
The variations are all the same move applied to a different field.
- Price. The count is honest, the unit price is not.
- Quantity. The price is defensible, and fewer goods ship than are billed.
- Description. Cheap goods are declared under a heading that carries a much higher value, or a controlled good is declared as something unremarkable.
- Repetition. One genuine shipment supports two or more invoices, usually presented to different banks.
- Absence. A phantom shipment: a complete, orderly set of documents for a container that was never loaded.
Each of these leaves a trace somewhere. The problem is that the trace is never in the same place as the payment.
Why the controls pass it
The most useful description of the difficulty comes from the supervisor rather than from vendors. FinCEN’s 2010 advisory on suspicious activity reporting for TBML puts it in two sentences:
It can be difficult to identify these activities given that financial institutions see only the documents related to a transaction and not the goods themselves. Further, documents related to trade-based money laundering may be created by the money launderers themselves with no neutral third party to verify the validity of the documents.
Both halves matter. The first is a structural limit: a bank is not a customs inspector and cannot open the container. The second is worse, because it removes the assumption that quietly underpins document review. A commercial invoice is not independent evidence of value. It is an assertion by one of the two parties who benefit from the misstatement, and a packing list is an assertion by the same party.
Documentary credit practice does not close this, and does not claim to. UCP 600 article 5 states that banks deal with documents and not with the goods, services, or performance to which the documents may relate. That is a deliberate allocation of risk, and it is what makes documentary credits work as an instrument: the bank’s obligation turns on whether the presentation complies, not on whether the underlying trade was honest. Article 14(b) gives a bank five banking days to decide whether a presentation is complying. A complying presentation can describe a shipment that never happened, at a price with no market basis, and it is still complying.
Transaction monitoring does not close it either, for a simpler reason. Monitoring built on account behaviour looks for patterns in payments: velocity, structuring, unusual counterparties, high-risk jurisdictions. In a misinvoicing scheme the payment is not the anomaly. The payment is the honest part.
Where the evidence actually sits
Read as an inventory rather than as an argument, the situation is plain. The payment record holds amount, currency, value date, the institutions on each side, and the names of the remitter and beneficiary. It cannot tell you whether anything shipped. The document set holds a goods description, an HS heading, a quantity, a unit price, an incoterm, ports, a vessel, and a consignee. It cannot tell you whether the price has any market basis, or whether the same documents already supported a different payment.
Neither record is deficient. Each is complete for its own purpose. The transfer is visible only in the comparison, and no system produces the comparison as a by-product of normal operation. Someone has to go and build it, under a deadline, across systems owned by different teams.
That is also why the classification layer matters more than it looks. The Harmonized System, maintained by the World Customs Organization, gives six internationally agreed digits per commodity group, with further digits added nationally. Comparisons that ignore this are easy to break: a benchmark drawn at four digits will happily average across goods that differ in price by an order of magnitude, and a price that looks anomalous at the wrong level of aggregation will not survive a challenge from the relationship manager.
The scale question, honestly
Estimates of TBML volume should be treated with care. FinCEN’s advisory cites a State Department estimate that the annual amount laundered through trade “ranges into the hundreds of billions,” and notes that illicit activity is hard to isolate partly because of the sheer volume of legitimate trade. Global Financial Integrity’s work on trade misinvoicing derives value gaps by comparing what partner countries report about the same trade flows, which measures discrepancy rather than criminality: mismatches arise from misinvoicing, but also from valuation convention, timing, transshipment, and plain error.
The honest position is that the scale is large enough that nobody serious disputes the exposure, and imprecise enough that a specific dollar figure should not be load-bearing in a control design.
What follows for a review
FinCEN’s own caution is the most important line in its advisory for anyone building a process: no one activity by itself is a clear indication of trade-based money laundering, and indicators have to be read alongside the activity expected for that particular customer. A price variance is a question, not a finding. It becomes a finding when the quantity does not reconcile, or the consignee changed in transit, or the party paying has no role in the contract.
This has a practical consequence that runs against how most alerting works. Adding indicators to a rules engine produces more single-signal alerts, each individually weak, each needing the same manual reconstruction of context. What changes the outcome is putting the records that disagree into one place, and keeping every figure attached to the document it came from, so the analyst spends the review reasoning rather than gathering.
The decision stays with the investigator. FinCEN asks filers to mark “TBML” in the SAR narrative and to set out why they suspect it, which requires a claim about a specific customer and a specific trade that a variance figure cannot supply on its own.