Most Customs Penalties Don't Start With Fraud. They Start With a Product Description.

Most customs penalties don't start with fraud.

They start with a product description that made sense to whoever wrote it, and means something different to the system that reads it.

The invoice says "industrial cleaner." The HS code reflects one assumption. The actual formulation triggers another. By the time that gap surfaces, the shipment is held, the compliance record has a flag, and someone is arguing with a customs officer about a document their supplier wrote six months ago.

No AI resolves that. No OCR tool knows what's actually in the box.

Classification accuracy starts before the shipment, at the point where someone who knows the product can describe it precisely, in structured form, once.

The problem isn't slow document processing.

It's that the data was never clean to begin with.

Where do you see classification errors originate most often - on the supplier side or internal?