How it works

How Classifyr works, and the mistakes it's built to catch

Most guidance on commodity code classification tells you what can go wrong, often as a lead-in to why you need large, expensive enterprise compliance software. This page shares the same, genuine mistakes, but as part of the actual mechanics of how a conversation with Classifyr turns into a code you can be confident in.

The mechanics

The short version: the system works through the tariff's own rules step by step, proposing a candidate code and then checking it live against HMRC's own Trade Tariff data before you ever see it. Nothing gets shown as an answer unless HMRC's own data confirms it actually exists and actually matches. If it doesn't, the conversation asks another question rather than guessing again.

How a description becomes a confirmed commodity codeFour stages left to right: you describe the product in plain language; a candidate code is proposed based on that description; the candidate is checked live against HMRC's own data; you receive a confirmed answer, or an honest statement that more detail or a specialist is needed. The first two stages are interpretive; the last two are grounded in official data.Describe itThe more detail,the betterIdentify itUsing the tariff'sown rulesConfirm itChecked, not assumedResolve itYour code, orclear next stepsInterpreting what you've told usGrounded in HMRC's live data

That confirming step is doing more work than it looks like. A general-purpose AI chatbot like ChatGPT proposing a commodity code from its own training is inherently a guess, an educated one, but still a guess about a tariff schedule that runs to 25,000 codes and changes regularly. Checking that guess against the real, current data before it reaches you is what turns a plausible-sounding answer into a credible one you can be confident in.

The conversation itself is doing real work, not just collecting details. Each question is adaptive and contextual: once HMRC's data shows the real, declarable options under a heading, the next question is built from what actually separates those specific options, not a standard checklist, and everything you've already said stays in view rather than being re-asked round after round. The more you tell it, the more precisely it can narrow down to the right answer, which is why giving more detail up front, rather than a single word or a vague category, genuinely gets you there faster. And the reasoning follows the same legally recognised order customs authorities themselves use to resolve genuine ties between two plausible codes, known as the GRI (General Rules for Interpretation): start with the heading's own wording, only move to how complete or unassembled something is if that doesn't settle it, and only weigh which material gives a mixed product its character if that still doesn't. It's not a shortcut through that process. It's the same process, asked conversationally instead of requiring you to already know it exists.

The 5 biggest commodity code mistakes

None of these are unusual. They're the ordinary ways classification goes wrong for genuinely well-intentioned businesses, and each one maps to something specific in how Classifyr is built.

1. Starting from a vague description

"Parts", "packaging", "components": broad terms that feel like a reasonable starting point but can genuinely land in three or four different chapters depending on what they actually are. A conversation can't resolve on a vague answer here. If a material fact is missing, that's the next question, not a guess filled in on your behalf.

2. Reading the title instead of the rule

Chapter and heading titles exist for ease of reference only. Legally, it's the heading's actual wording and the notes attached to it that decide, and a product can look like an obvious fit for one heading while a specific exclusion or functional detail moves it somewhere else entirely. That's exactly the order of precedence built into how a candidate code gets reasoned about here: not the section heading, but the actual legal text.

3. Picking whichever code costs less

When two codes seem plausible and one carries a noticeably lower duty rate, it's a genuinely human instinct to lean toward it. But there's always a correct answer, arrived at through a defined legal sequence, not a choice between two acceptable options. HMRC doesn't see a one-off judgement call when a lower-duty code gets used consistently across a product line; it sees a pattern, and a pattern is what moves a penalty from careless toward deliberate. Working through that same sequence properly, rather than stopping at the first plausible-looking answer, is the difference.

4. Settling for "Other"

Every heading has a residual bucket for goods that don't fit anywhere more specific, and it usually carries the highest duty rate in that heading. Landing there isn't wrong in itself; plenty of products genuinely belong in it. The mistake is accepting it without checking whether a more specific, better-fitting option actually exists. That check happens automatically here before an "Other" result is ever treated as final.

5. Trusting an ungrounded answer

Asking a general-purpose AI chatbot for a commodity code will get you an answer. It won't tell you whether that code actually exists in HMRC's current data, or whether it's still current at all, because nothing's checking. That's the one thing this entire approach is built around: a candidate code is proposed, then checked live against HMRC's own data, and if the two don't agree, you're told that plainly rather than shown a confident-sounding guess.

What the £2.99 report adds

The conversation itself is free, and ends with a suggested code you can copy and use right away. The report shows how the code was arrived at and why. It's important for HMRC that you have the rationale not just the answer.

It's the same conversation, turned into a document: every candidate code that was considered, including the ones ruled out and why, each with the specific GRI rule that decided it and the HMRC check that confirmed it. Alongside that: duty, import VAT, and any preferential rate that applies under a trade agreement, controls explained rather than just flagged, and a direct link to the code's own page on the official Trade Tariff so you or anyone checking your work can verify it independently.

It's built to be handed to a customs broker, kept on file against a future HMRC query, or just read once so you understand why the code is what it is, not just what it is.

What this doesn't replace

If it genuinely can't confirm a match, you don't just get told no. You get an honest breakdown of what is known up to that point and the specific, distinct leads worth following up. From there, you can start a new conversation providing more detail, or, if there's genuine nuance or ambiguity, take what's been found to a specialist to speed up their own review.

A live, HMRC-checked conversation is a genuinely different starting point from guessing or copying a code from somewhere else, but it isn't a legally binding ruling. For anything genuinely finely balanced, or where the value at stake justifies certainty, an HMRC Advance Tariff Ruling is still the only answer that protects you if it's ever questioned later. That protection comes with a real time cost, though: HMRC's own guidance puts the typical response at 30 to 120 days. Part of what a proper check should do is tell you plainly when you've reached that point, rather than presenting a best guess with the same confidence as a confirmed one.

Check or find your code