Classification Rules

The General Rules of Interpretation, in Plain English

The General Rules of Interpretation are six legally binding rules, printed at the front of the HTSUS, that govern how every imported product gets classified. They are not a menu — they are a mandatory sequence. GRI 1 is evaluated first, always, and every later rule may only be invoked once the classifier has established that every prior rule genuinely failed to resolve the classification. Most successful CBP challenges to a classification don't turn on which heading was chosen. They turn on whether the classifier actually earned the right to be at the rule they used.

---

Why Order Matters More Than Content

The GRI were written by the World Customs Organization and adopted into the HTSUS as binding law, and U.S. courts have repeatedly enforced them as exactly that — not guidance, not best practice, law. The six rules address progressively harder classification problems: GRI 1 resolves the large majority of products through heading text and notes alone. GRI 2 extends that coverage to unfinished goods and mixtures. GRI 3 resolves genuine competitions between two or more headings. GRI 4 is the residual rule for goods nothing else describes. GRI 5 handles cases and packaging. GRI 6 repeats the whole exercise one level down, at the subheading.

The part that trips people up is the constraint between the rules, not inside them: a classifier cannot invoke GRI 3 because a product feels complicated. They have to first show that GRI 1 and GRI 2 were genuinely insufficient. That sequencing requirement isn't a formality — it's enforced in CBP rulings, protest proceedings, and Court of International Trade decisions. A classification that jumps straight to the rule that produces the desired answer, without documenting why every earlier rule failed, is not a defensible classification. It's a guess wearing a citation.

GRI 1 — Start With the Words That Are Actually Binding

GRI 1 says something specific and easy to misread: the chapter and section titles you see printed above the headings are for reference only and carry no legal weight. What actually governs is the heading text itself, read together with whatever section or chapter notes apply to it — and those notes carry the same legal force as the heading text, not some lesser, secondary status. A note that excludes a product from a section is just as binding as the heading language itself.

This is where most classification disputes actually live. Not in some exotic GRI 3 or GRI 4 argument — in whether the right notes were even consulted at GRI 1, whether an exclusion buried in a chapter note was missed, whether heading language got read more broadly or narrowly than it should have. If GRI 1, read against every applicable note, already answers the question unambiguously, the analysis is over. Nothing later gets reached.

GRI 2 — Unfinished Goods and Mixtures Don't Get a Pass

GRI 2(a) closes an obvious loophole: an unfinished or unassembled article classifies as the finished article, as long as it already has the finished article's essential character. An automobile body shell without an engine still classifies as a motor vehicle, not as a steel structure, because anyone looking at it can tell what it's going to be. Flat-pack furniture shipped disassembled classifies as furniture, not as a box of boards and screws.

GRI 2(b) closes a different one: a heading that names a material — "articles of plastics," say — covers that material mixed or combined with other materials, not only the pure form. That sounds narrow until you notice the consequence: a mixed-material product can now be prima facie described by two or more material headings at once, one per component. GRI 2 doesn't resolve that competition. It just proves the competition is real and hands it to GRI 3.

GRI 3 — Competing Headings, and Where the Real Fight Is

GRI 3 only gets reached when a product is genuinely, prima facie classifiable under two or more headings — usually because GRI 2(b) mixture coverage or a genuinely ambiguous product produced real competition. The first move, GRI 3(a), is to prefer the more specific heading. That resolves most of these cases outright. It doesn't resolve the case where both competing headings are equally specific relative to the product, and that's where GRI 3(b) — essential character — takes over.

Essential character is the GRI 3(b) test for deciding which component gives a composite good, mixture, or retail set its classification identity, and it deliberately has no formula: nature, bulk, quantity, weight, value, and functional role are all listed as relevant, and none of them is automatically decisive. That refusal to reduce judgment to arithmetic is exactly why GRI 3 is the most litigated of the six rules — and it's substantial enough that it gets its own full treatment, worked through against a real CBP ruling where the higher-value component still lost, in Why "Essential Character" Is the Most Litigated Phrase in the Tariff Schedule. If neither 3(a) nor 3(b) resolves it, GRI 3(c) is a purely mechanical fallback: whichever heading comes last in numerical order wins. Reaching 3(c) means both of the real tests genuinely failed.

GRI 4, 5, and 6 — The Rules Everyone Forgets Exist

GRI 4 is the rule for the product nothing else describes — classify it under the heading for goods most akin to it, by nature, use, composition, or commercial category. It exists because no tariff schedule can anticipate every product that will ever be traded. But it's invoked rarely, and for good reason: CBP and the CIT treat a GRI 4 argument with heightened skepticism, because in the overwhelming majority of cases where someone reaches for GRI 4, a more rigorous GRI 1 through 3 analysis would have found the actual answer without it.

GRI 5 is narrower still: it decides whether the case or container a product ships in classifies with the product or separately. A camera case shaped specifically for one camera model, presented with the camera, classifies with the camera. A reusable, purpose-built shipping drum with real independent commercial value can classify on its own. Everyday disposable packing — cardboard, foam, plastic wrap — always classifies with what it's packing, no analysis required.

GRI 6 doesn't introduce new logic — it re-runs GRI 1 through 5 by analogy one level down, at the subheading, once the heading itself is settled. The rule that catches people here is structural: subheadings are organized in tiers (1-dash, then 2-dash within it), and you cannot compare a 1-dash subheading against a 2-dash subheading as if they were peers. Skipping straight to a 2-dash subheading without first confirming the correct 1-dash parent produces a wrong 6-digit code even when the heading itself was right — and the 6-digit level is exactly where antidumping and countervailing duty scope and most USTR exclusions actually operate.

The Additional U.S. Rules — GRI's U.S.-Only Cousins

Four more rules sit alongside the GRI, specific to the U.S. tariff schedule rather than the international Harmonized System: the Additional U.S. Rules of Interpretation. Rule 1(a) governs "principal use" headings — language like "of a kind used for" — and directs classification by the general commercial use of that class of goods, not any one importer's specific intended use. Rule 1(b) governs the narrower "actual use" provisions, which require the goods actually be put to a stated use, with proof furnished to CBP within three years of entry.

Rule 1(c) makes parts-and-accessories provisions residual: they apply only when a more specific heading doesn't already name that part directly. Rule 1(d) resolves textile blends by predominant weight, absent a more specific rule for that mixture. These matter most in exactly the chapters you'd expect — agricultural equipment, sporting goods, medical devices, textiles — where getting the use-provision analysis wrong is a documented, recurring source of audit exposure.

Why This Is a Software Problem, Not Just a Legal One

Most classification tools don't actually run this sequence — they pattern-match a product description to a plausible-looking heading and stop. That produces a code, but not a defensible one, because the code isn't backed by a record of which rule resolved it and why every prior rule was insufficient. If a classification can't answer "why didn't GRI 1 resolve this" and "why didn't GRI 2" before it gets to whatever rule it actually used, it isn't a determination CBP has to respect on protest — it's an assertion. It's also exactly how two brokers land on two different codes for the same product and neither can prove they're right — see Why the Same Product Gets Two Different HTS Codes from Two Different Brokers.

Kanon's classification engine runs the GRI in the sequence the law actually requires: GRI 1 evaluated first against the product's attributes and every applicable note, advancing to GRI 2 only when GRI 1 is genuinely insufficient, and so on through GRI 6. Each rule's outcome — resolved, or insufficient and why — is recorded in the Classification Support Package, so the traversal that produced the code is auditable, not just the code itself.

Frequently Asked Questions