EU Publishes 2027 Combined Nomenclature, Applying From 1 January

The Analysis

Every autumn the same document lands and every autumn half the industry misreads what it is. It is not a tariff announcement. It is a renumbering exercise, and renumbering is where declarations go wrong.

On 30 October the European Commission published the 2027 Combined Nomenclature in the L series of the Official Journal, as Commission Implementing Regulation (EU) 2026/2163. The CN is the working tool that fixes the customs duty rate and the statistical treatment of every declared goods line, on import, on export and for intra-EU trade statistics, and Brussels rewrites it every year. The 2027 edition applies from 1 January 2027.

That leaves importers and brokers two months. Two months on paper. In practice the working window is closer to seven weeks, because half of Europe stops taking new classification instructions in the third week of December. I have spent enough Januarys cleaning up after this to know which step people skip.

Start with the three hard items and with what each one forces. The regulation number is (EU) 2026/2163. The publication date is 30 October 2026, in the L series of the Official Journal. The application date is 1 January 2027. Those three are the only facts in this story that nobody argues about, and everything below is built on them.

What the CN actually does is narrower than people assume and wider than they remember. Narrower, because it does not set policy; it is the classification table. Wider, because it is the same table for three separate obligations: the import declaration that sets your duty bill, the export declaration, and the intra-EU statistical return. One table, three filing families, three places where a stale code turns into a wrong filing. Most importers remap the first one and forget the other two, then spend March answering questions about statistical returns nobody looked at in December.

Why the end of October. The CN has to be in force before the year it labels begins, and the practical publishing slot has settled in the last quarter. Upstream there are two engines pushing it. One is the harmonised system maintained by the World Customs Organization, revised roughly every five years, which the EU transposes into its own eight-digit structure. The other is Brussels itself: new product categories need lines so they can be seen in statistics, policy measures need lines so they can be applied, and trade agreement schedules are drafted against CN codes, so a change in a schedule eventually shows up in the table. Neither engine is dramatic. Together they guarantee that some of your codes will move every single year, whether or not anybody calls it a big revision.

Do the calendar arithmetic before anyone asks you for a plan. From 30 October to 1 January is 62 days. Take out the holiday shutdown, say 20 December to 4 January, and 47 days remain. Strip weekends and you are at roughly 33 to 35 working days. Brokers in much of Europe stop taking new classification instructions around the second week of December because their own teams rotate and their systems freeze. So the honest number to give your management is 35 working days, and it sits on top of peak season and on top of contract renewals. If you have ever tried to get a supplier in Asia to reissue a technical specification in the last ten days of December, you know how that ends.

Who this touches, and how hard. At the bottom, an importer with thirty stable lines in chapters that rarely move: a two-day check, and the answer is almost always that nothing changed. In the middle, an importer with six hundred to three thousand active lines, which is what a consumer goods catalogue or an automotive parts programme looks like: this is a project with an owner and a deadline. At the top, an e-commerce importer with tens of thousands of listings and attributes generated partly by suppliers: this is not a project, it is a data problem, and the code is the least of it because the goods description has to move with the code, or you will file a new number under an old sentence. Then there are two groups that always get forgotten. Operators holding binding tariff information, who bought certainty against a code that may no longer exist. And anyone filing export declarations or intra-EU statistical returns, who will swear they are unaffected because they do not import.

The transmission date is the part that catches people, and it has nothing to do with shipping. Which edition applies is decided by the date the customs declaration is accepted. Not the invoice date, not the contract date, not the bill of lading, not the day the vessel tied up. A box discharged on 29 December and declared on 2 January is declared under the 2027 edition. Goods that arrived in November and sat in a customs warehouse until they were released for free circulation in the second week of January are declared under the 2027 edition. If you run a bonded operation, your January release schedule is your real cutover date, and it is the one date nobody puts on the project plan.

Now the piece the headlines leave out, and the reason I am writing this at all. The CN is eight digits and you do not declare eight digits. What goes on the declaration is the ten-digit code: the CN line plus the European measure layer sitting on top of it, which is where duty suspensions live, along with anti-dumping and countervailing duties, safeguards, surveillance, tariff quotas and prohibitions flowing from sanctions. The eight-digit table tells you the skeleton. The ten-digit code tells you what is actually assessed and which certificate is actually demanded at the counter.

When the annual revision splits, merges or retires an eight-digit line, every measure attached to that line has to be re-anchored onto the successor. That re-anchoring is administrative work, it is done by different hands than the ones who wrote the table, and it does not always finish on 1 January. So the exposure is not the duty rate you read off the new table. The exposure is the gap: a remedy duty that has not yet been hooked onto the new line, or a measure that has been hooked onto a sibling line and now catches your product by accident.

If you import steel fasteners, aluminium extrusions, ceramic tiles, plywood, electric bicycles or anything else that carries a trade remedy, that gap is where your money is. Where the legal line sits when a measure has not been carried across is Dr. Ingrid Voss's call and not mine; what I can tell you is that the declaration will still be accepted, the duty will still be assessed, and it will be assessed against whatever the ten-digit code says on the day of acceptance, with your name on the box.

Binding tariff information deserves its own paragraph, because it is the one asset people assume is safe. A BTI is bought certainty: you asked the customs authority to confirm a classification and you can rely on the answer, normally for around three years depending on what your own decision says. That certainty is anchored to a code. If the 2027 table retires or renumbers the line your decision cites, the certainty you paid for does not travel across automatically. Authorities review holdings after a nomenclature change and they can amend or revoke. Pull the list of every decision you hold, put the cited code in one column and the 2027 code in the next, and put a name against each row. Replacement is slow as well: a fresh application takes weeks to months by convention, which is longer than the window you have. Anything you discover in the last week of December cannot be fixed by 1 January.

Put numbers on the effort, with the assumptions stated. Assume a mid-size importer: nine hundred active lines into the European Union, twelve forty-foot containers a month, average declared value of 120,000 euro per container. Assume, on past revision patterns, that 7 per cent of the line list falls inside a range the revision touches, which is about sixty-three codes. Assume twenty minutes per code for a competent person to reconfirm the classification, check the ten-digit measure layer and update the master record. Sixty-three codes at twenty minutes is twenty-one hours, which is three working days for one person or a day and a half for two.

At an internal cost of 400 euro a day, the exercise costs about 1,200 euro. Now price the failure. Miss three of those sixty-three codes. Assume each carries 80,000 euro of import value a year and that the slip moves the rate by three percentage points. That is 2,400 euro a year per code. Customs can look back three years. Three codes over three years is 21,600 euro, before interest and before any penalty. That is eighteen times the cost of doing the work, and the work is a December afternoon.

Scale the other direction, on one container. At 120,000 euro declared value, a move of 2.5 percentage points in the applicable rate is 3,000 euro on that box. At twelve boxes a month that is 36,000 euro a month and 432,000 euro over a year. I am not predicting a rate move. I am putting the sensitivity on the table so you can judge how much checking your margin can carry. Note also what a wrong code costs beyond the duty itself: a wrong declaration is a liability that outlives the financial year, it contaminates the statistical return, and if you claim preference under a trade agreement it puts the origin evidence for that line in question too, because the agreement schedules are written against the same codes you are about to change.

The commercial side of the same arithmetic gets less attention than it should. If you sell delivered duty paid into the European Union, a rate move on a remapped line lands on you and not on your customer, because your price was quoted against last year's table. Go through your 2027 price lists and find every line where the duty was baked into a fixed customer price, then decide whether the contract carries a tax-change clause or whether you simply absorb it. A clause that lets you pass through a change in the applicable duty rate is worth more than any amount of careful classification, and it is much easier to insert in November than in February.

The parcel and consolidation side has its own version of this. Low-value consignments and consolidated groupage are declared at a coarser granularity, with several hundred catalogue items sitting under one code. That looks like less work and it is not, because a revision that splits the parent line moves the whole basket at once and you will not see which items went the wrong way until somebody queries a month's worth of entries in a single batch. If you run consolidated e-commerce flows, ask your broker for the mapping at item level rather than at code level, and do it in November, because December is when their consolidation team is at its thinnest.

There is a record-keeping angle that only shows up when something goes wrong later. If customs queries a line in 2028, the question will not be whether you were right, it will be whether you can show how you decided. For every line you remap, keep a one-line reason: the old code, the new code, the source you relied on, and the date. Four fields, and they turn an argument into a document. Operators with authorised economic operator status already have this discipline written into their own procedures, so for them it is a matter of using the template they already have. For everyone else it is the difference between a twenty-minute answer and a three-week one, and it costs nothing in December because you are already sitting in the table.

And ask your suppliers to do the same thing you are doing, inside the same window. A declaration that disagrees with the commercial documents behind it is a problem whichever edition applies, and the classification printed on an invoice or a packing list issued by a factory in Asia will not update itself on 1 January. If your suppliers print codes on their paperwork, tell them which edition to print from, and tell them in November. Otherwise you spend the first quarter of 2027 reconciling documents that were wrong before the regulation existed.

One practical thing to do this week, and it is the email I would send my broker. Ask four questions in one thread: when their tariff database switches to the 2027 edition, whether they are re-running the ten-digit measure layer for every line they hold for you or only the ones you flag, whether any of your lines sit under a binding tariff information decision, and what their instruction cut-off is in December. Four questions, one thread, and the answers tell you whether this is being handled or assumed. If the answer to the second one is that they only re-run what you flag, you have just learned that you are the one doing the diff, and you should start on Monday rather than in the last week of November.

What would make this a non-event. If your entire line list sits in chapters the revision leaves alone, this is a two-day confirmation and then you go back to work. Two conditions turn it back into a project: you hold binding tariff information, or you sit in a category carrying a trade remedy duty or a tariff quota. One condition makes it urgent for everybody: goods arriving in the last ten days of December that will not be declared until January, or stock sitting in a customs warehouse scheduled for release in January. And the honest counterfactual: the revision is uneven from year to year, it may be a light year for your range, and you may spend 1,200 euro to confirm that nothing moved. That is still a good trade. Unused insurance is cheaper than a three-year look-back.

The sequence I would run, in this order. Get Annex I of the regulation, where the actual table lives, and not the press note. Diff it against your own goods master, not against your broker's summary of your goods master. For every line that moved, open the ten-digit code and read the measure layer out loud, because that is where the money is. Update the master record, the goods description that travels with it and the broker's tariff table in the same pass, or you will have three versions of the truth by February. Test-file five to ten lines in December against the 2027 codes to see what the system now asks for that it did not ask for before. Then freeze changes in the last week of December, because a half-finished master data change is worse than a late one.

The traps, in the order people fall into them. Waiting for a letter from the customs authority; there is no letter, the table is the notice. Remapping only the first six digits and calling it done, when the European structure is eight plus two. Remapping the import side and leaving export declarations and statistical returns on last year's codes. Letting the systems team push a bulk update on 31 December with nobody available to read the errors. Assuming the broker's tariff database flips on 1 January; ask the vendor for the date and get it in writing, because some of them lag by days and those days are your declarations. Accepting a supplier's code as your defence; the declarant carries the classification, not the factory. And the quiet one: changing the code without changing the goods description, which is the fastest way to invite a request for information on a line that used to sail through without a question.

↑ Back to the key points

— By Derek Xu