The White House Office of Trade and Manufacturing Policy identifies more than 40 jurisdictions associated with elevated transshipment risk and cites five estimates ranging from $40 billion to $303 billion annually. These figures measure different definitions and are not directly comparable. Their common signal is that tariff evasion has scaled beyond physical inspection.
The customs officer with a clipboard is no longer the unit of enforcement. The unit is a data pipeline.
Why Physical Inspection Fails
Three constraints make box-opening an inadequate enforcement model.
Scale. CBP processes too many shipments for physical inspection to cover more than a small fraction without obstructing trade. Fraud distributed across ports, products, and importers defeats spot checks.
Paperwork moves faster than cargo. Certificates, invoices, labels, and routing records can be rewritten instantly. Goods remain physically slow; their declared identities do not.
Tariff arbitrage keeps changing. Every country-product differential creates an incentive to reroute goods or manufacture a new origin story. The tariff map is legally unstable and operationally dynamic. Static watchlists decay quickly.
Enforcement must ingest tariff changes and continuously rescore shipment risk.
Testing Transformation Versus Relabeling
“Substantial transformation” separates lawful production shifts from origin laundering. It is a legal judgment, but the evidence must be generated at shipment scale.
| Data layer | Core question | Enforcement signal |
|---|---|---|
| Shipment and routing data | Where did the goods move? | Matching product codes, timing, value, region, and dwell time |
| Ownership networks | Who controls the entities? | Beneficial ownership, shell companies, financiers, and related parties |
| Capacity validation | Could the facility produce the declared output? | Footprint, equipment, inputs, labor, satellite imagery, and shipment volume |
| Port computer vision | Does the physical cargo match its declaration? | Container markings, packaging, and non-intrusive inspection imagery |
| Behavioral analytics | Does the transaction resemble genuine manufacturing? | Abnormal routes, ownership changes, dwell times, and value patterns |
Commerce’s stricter methodology flags flows when the same HS8 product enters a country from China and leaves the same subnational region for the United States within the same quarter. Exiger uses shipment attributes, facility capabilities, satellite imagery, components, and a 90-day dwell-time threshold to test whether claimed production is plausible.
No layer proves fraud alone. A clean manifest can fail a capacity check. A real factory can still serve a shell-controlled pass-through operation. Enforcement value comes from correlation across layers.
The Vendor Stack
The market divides into three functional tiers.
| Tier | Function | Examples | Defensible moat |
|---|---|---|---|
| 1 | Enforcement intelligence | Exiger, Altana | Risk scoring, capacity validation, and workflow integration |
| 2 | Entity resolution | Sayari | Cross-border ownership and corporate-registration graphs |
| 3 | Shipment-data aggregation | Panjiva, Descartes, ImportGenius, TradeInt | Coverage, update frequency, and record cleaning |
Raw manifest data is increasingly commoditized. The durable advantages sit above it:
- Entity-resolution accuracy
- Facility and production-capacity evidence
- Integration into targeting and adjudication workflows
The architecture therefore depends less on exclusive access to shipment records than on converting fragmented records into defensible enforcement evidence.
Concentration at the integration layer creates a separate risk. A critical public enforcement system resting on a small number of private platforms introduces dependency, auditability, and model-governance problems.
Better Detection Relocates Arbitrage
The deterrence equation is simple:
Evasion remains profitable when the tariff savings exceed the expected cost of detection, penalties, delay, and operational concealment.
Large tariff differentials can support expensive evasion infrastructure. Better detection therefore changes behavior along three margins.
1. Weaker-data jurisdictions
As major hubs become easier to score, flows migrate toward jurisdictions with poor export records, limited customs capacity, free zones, bonded warehouses, or low volumes that remain below aggregate detection thresholds.
2. More credible transformation
If short dwell times and exact product matches trigger enforcement, operators add real processing, hold inventory longer, alter product codes, or diversify inputs. This raises local value added, but it also makes origin laundering harder to distinguish from legitimate production.
3. Deeper ownership obfuscation
Name screening pushes bad actors toward clean intermediaries. Ownership tracing then pushes them toward additional shell layers, nominee structures, and fragmented control.
This is an arms race, not a one-time software deployment.
What the Estimates Actually Mean
The report’s five headline figures are alternative measurements, not a confidence interval around one hidden value.
| Source | Estimate | Measurement scope |
|---|---|---|
| Goldman Sachs | $40B | Narrow rerouting estimate |
| CEA | $60B | Rounded midpoint of a country-product screen |
| Exiger | $75B | Shipment-level analysis and systemic upper-bound midpoint |
| Commerce OTEA | $109B | Broad trade-transfer benchmark |
| Altana | $303B | Facility-level supply-chain exposure |
The range reflects different datasets, definitions, and screens. It should not be averaged or added together.
AI outputs also remain risk signals, not legal findings. A routing anomaly, ownership link, or capacity mismatch should trigger document demands and investigation. It does not independently establish fraud.
Strategic Implication
The enforcement bottleneck has moved from inspection labor to data fusion.
A functioning customs-intelligence stack must connect declarations to routing history, ownership, physical capacity, product composition, and observed cargo. Its output must then enter an enforcement workflow capable of collecting duties, imposing penalties, or excluding shipments.
Better infrastructure will raise the minimum cost of evasion and reduce the number of jurisdictions where crude relabeling remains profitable. It will not eliminate the underlying incentive. As long as large tariff differentials exist, the arbitrage survives by relocating, becoming more operationally credible, and hiding behind deeper corporate structures.
The winning enforcement system is therefore not the one with the longest blacklist. It is the one that continuously maps economic origin against physical reality.