// FIELD BRIEFING // TARIFF ENGINEERING

Tariff Engineering Is Legal. Faking It Costs $365 Million.

Reclassifying a product you never changed is fraud. Changing the product before it crosses the line is a right the Supreme Court handed importers in 1881. Here is exactly where that line sits, and how you prove which side you are standing on.

// The Rule That Makes This Legal

Every few months a CFO asks me some version of the same question, usually in a lowered voice: “Can we just classify it differently?”

No. But you can almost always ask a better question, which is: can we build it differently?

That distinction is the whole discipline. In Merritt v. Welsh (1881), duty on imported sugar turned on color measured against the Dutch standard. The government believed the sugar had been artificially darkened to test at a lower grade and draw a lower duty, but that evidence was excluded at trial. The case reached the Supreme Court on a record where the sugar was simply dark. So read Merritt for its reasoning, not for a story about tinted sugar: if goods are truly invoiced and honestly presented for examination, the government takes them as it finds them, and the commercial motive behind their condition does not rewrite the classification.

Thirty-one years later, in United States v. Citroen (1912), the Court sharpened the point that still governs today: classification turns on the condition of the article as imported. Not its history. Not its destiny. Its condition at the moment it hits the line.

That is the entire legal foundation of tariff engineering, and it is roughly 145 years old. It is not a loophole. It is not aggressive. It is settled law that CBP applies every single day.

// The $365 Million Counterexample

Ford is the cautionary tale everyone half-remembers and most people get backwards, so let me lay it out cleanly.

Ford imported Transit Connect vans from Turkey fitted with second-row passenger seats. Passenger vehicles carry a 2.5 percent duty. Cargo vans carry 25 percent, the old “chicken tax.” The seats came out shortly after the vans arrived in the United States, and a good number of them were reportedly headed straight for a shredder.

Ford won at the Court of International Trade in 2017. The Federal Circuit reversed in 2019, and the reasoning is worth reading closely, because it is the whole test in one paragraph. The court did not object to the seats existing. It objected to what they were: designed to be removed. No headrests. No wiring in the seatbacks. No tumble-lock mechanism. Cost-reduced fabric. Assemble that list and you are not looking at a passenger vehicle that happened to get converted. You are looking at a cargo van that was built to pass an exam. The Supreme Court declined to take the case. Then, in March 2024, Ford agreed to pay $365 million to settle separate government allegations that it had misclassified and undervalued those vans.

Here is the lesson people miss. Ford did not lose because it engineered a tariff outcome. Ford lost because the engineering was cosmetic and temporary. The seats were not a real feature of a real product. They were a costume the vehicle wore for the length of a customs exam.

Compare that to Columbia Sportswear stitching a functional zippered pocket below the waist on a shirt, or Converse laying felt across the outsole of a sneaker. Those changes survive because the product actually leaves the factory that way, gets sold that way, and gets used that way. The engineering is real. The classification simply follows it.

// The Three Levers

Duty is a function of three inputs. That is it. Every legitimate duty-reduction strategy pulls at least one of them, and every one of them lives or dies on documentation.

Notice what is missing from that table. There is no lever labeled “pick a different code.” That is not a lever. That is a false statement to the government.

// Origin Is the Hot Lever Right Now

Classification engineering is mature. Origin is where the pressure is in 2026, for two reasons.

First, the tariffs that hurt most are origin-based, and the legal authority behind them has been torn down and rebuilt twice this year. The Supreme Court struck down the IEEPA tariffs, including the entire reciprocal regime, in Learning Resources v. Trump on February 20, 2026. A Section 122 surcharge of 10 percent filled the gap. The Court of International Trade struck that down too, in May, but relief ran only to the plaintiffs and the ruling went up on appeal, so collection continued until the surcharge hit its statutory 150-day wall and expired on July 24, 2026. USTR replaced it the same day with new Section 301 “forced labor” tariffs running 10 to 12.5 percent across roughly 60 economies. Add the long-running China Section 301 actions on top of that, and Section 232 alongside it, which keys off the article rather than the country.

The statutes keep changing. The mechanic does not: the rate still keys off where the good is from. Move the origin legitimately and you move a far bigger number than a two-point rate difference in the schedule.

Second, and this is the part that gets people in trouble, everyone else figured that out too. CBP has poured enforcement resources into transshipment. If a supplier in Southeast Asia offers to “re-export” Chinese goods under their own origin paperwork, that is not tariff engineering. That is a felony with your company’s name as importer of record on the entry.

Substantial transformation is still a case-by-case name, character, or use test, and CBP is scrutinizing it hard on assembled goods: electronics, adapters, automotive components, anything with a multi-country bill of materials. Simple assembly does not transform. Screwing a Chinese board into a Vietnamese housing does not make it Vietnamese.

Worth watching: CBP has a long-pending proposal to replace that case-by-case analysis with objective tariff-shift rules for non-preferential origin of goods from Canada and Mexico. The one you want is 86 FR 35422, published July 6, 2021, RIN 1515-AE63. Do not confuse it with the 2008 uniform-rules proposal, which CBP largely abandoned in 2011 after the comment file came back hostile. Note also that Part 102 already governs Canadian and Mexican goods for country-of-origin marking; the pending rule would extend the same tariff-shift methodology to the non-marking questions, meaning antidumping, quota, and government procurement. It is a proposal, not law. Confirm its current status before you build a program around it.

// Why the Math Got Louder

Tariff engineering used to be a rounding-error discipline. Stacking changed that.

A single entry today can carry an MFN base rate, a Section 301 rate, and a Section 232 rate, all applied against the same customs value.

On Section 232, stop thinking in terms of the word “derivative” and start thinking in terms of annexes, because that is how the duty is actually assessed. Annex I-A carries 50 percent and covers all aluminum and steel articles, most copper articles, and certain derivative articles. Annex I-B carries 25 percent and covers the remaining copper articles plus the aluminum and steel derivatives that did not land in I-A. So a derivative that is almost entirely metal can absolutely sit at 50, and the only way to know is to look your subheading up in the annex. Copper has one more wrinkle: the action reaches semi-finished forms and copper-intensive derivatives, but refined cathode and ore sit outside it.

Now the piece most importers still have backwards. The April 6, 2026 restructuring abandoned metal-content valuation. Those rates now apply to the full customs value of the article, not to the value of the metal inside it. That was the headline of the change, and it repriced entire product lines overnight for companies who were still modeling the old way.

The country-keyed layer, currently the Section 301 forced-labor tariffs, exempts articles already subject to Section 232, so a steel article does not pay both. The China Section 301 actions do stack. A Chinese steel article can land at 50 percent plus 25 percent.

Run that against a real annual volume. A classification or origin position that used to be worth a few basis points is now worth seven figures. Which is also exactly why CBP is looking harder at the ones that changed recently.

// How to Do This Without Losing

1. Design first, classify second. Bring trade compliance into the product development meeting, not the entry summary review. Once the tooling is cut, your options are gone.
2. Get a binding ruling. Search CROSS for how CBP has treated products like yours, then file through eRulings. A ruling is legally binding on CBP for future entries of that product, and it is the single best piece of paper you can own. If your position is too weak to submit for a ruling, that is your answer.
3. Make the change real and permanent. The feature has to exist in the product that gets sold to the end customer. If it comes off in your warehouse, you are Ford in 2019.
4. Build the condition-as-imported file before you need it. Photographs, drawings, lab reports, production records, dated. Reconstructing this two years later during an audit is where good positions go to die.
5. Check who is importer of record. Engineering a beautiful duty position means nothing if your Incoterm quietly made you the importer on a shipment you never controlled, or made you the party on the hook for a supplier’s origin claim you never verified.
6. Re-run it annually. The HTS moves. The 2026 Basic Edition published December 31, 2025, and revisions have followed through the year. A position that was correct in 2024 may not survive a heading change you never noticed.

// What Getting It Wrong Costs

Under 19 U.S.C. Section 1592, the multiplier scales with your state of mind, and the exposure does not disappear just because the mistake was cheap:

Pay attention to that right-hand column, because it is the one that catches classification people. Plenty of misclassifications lose the government nothing. You declared the wrong heading and the rate happened to be identical, or you overpaid. No revenue loss, no harm, no case? No. You are still exposed to 20 or 40 percent of the dutiable value of the merchandise. The materially false statement is the violation. The lost duty is only a multiplier.

And the bottom row is not a typo. Fraud exposure is the value of the goods themselves, not a multiple of the duty. On a container of anything worth having, that is a company-ending number.

A prior disclosure cuts all of this dramatically. But understand when the window actually shuts. It is not when a notice lands in your mailbox. It closes once CBP has commenced a formal investigation of the violation and you have knowledge, actual or constructive, that the investigation exists. By the time formal paper arrives, you are usually well past that point and arguing about what you knew and when. That is not an argument you want to be having.

// The Short Version

You are allowed to build your product so it costs less to import. You are not allowed to describe it as something it is not. The distance between those two sentences is measured in engineering drawings, production records, and binding rulings, and if you cannot produce those on demand, you do not have a tariff strategy. You have a bet.