EU customs reform names Temu, Shein as importers; 3 euro duty on small parcels

The Analysis

There is one parcel sitting in a picking cage tonight that I would like to talk about, because everything about this news reaches you one box at a time, and the box is always about to miss a truck. Inside it is a nineteen-euro item with a label printed from a marketplace order, a customs form generated by whoever booked the label, and a declared description copied from a listing that somebody wrote eighteen months ago. The whole paper trail for that parcel was created in about ninety seconds, and it is almost impossible to correct once the van doors close.

On 3 September the Council of the EU gave its approval to an overhaul of the customs framework, and the part of it that matters most to anyone shipping direct to European consumers is this: non-EU online platforms, naming Temu, SHEIN, AliExpress and Amazon in the discussion, are to be treated as the importer of record for the parcels they sell. Parcels below the 150 euro line would carry a flat duty of three euro each plus a handling charge, and anyone who keeps getting it wrong can be fined somewhere between one and six per cent of what they brought into the EU over the previous twelve months. That is the factual ground we have. Note what I am not doing: I am not telling you the date this bites, because there is no application date in what has reached us, and I am not going to invent one for you to plan against.

What I can do is the thing I actually do, which is look at a paper trail. Importer of record is not a marketing phrase, it is a question about who receives the assessment and who gets the letter when a number does not reconcile. Whether the new designation moves that burden in practice, and by which route it could flow back through platform terms to a seller in Shenzhen, is something Dr. Ingrid Voss should read off the published text with the exceptions in front of her. My lane starts at the cage: which data is on that parcel, who put it there, who can still change it at eleven at night, and what the change actually costs in hours rather than in opinions.

Start with the most misunderstood part, because almost every client I speak to gets it backwards. The three euro is charged per parcel, not per order, not per buyer, not per week. If one customer order arrives in three boxes because the warehouse picked from two locations, that is three times three euro. Assume a seller doing forty-two thousand parcels a month into the EU with an average declared value of nineteen euro, so nothing in the book touches the 150 euro ceiling today. Forty-two thousand times three euro is one hundred and twenty-six thousand euro a month, about one and a half million euro over a year, before anyone says a word about handling fees or fines.

Now put that against the value of the goods, because that is where the reform stops being a rounding error. Three euro on a nineteen-euro parcel is fifteen point eight per cent of its declared value. Three euro on a sixty-euro parcel is five per cent. The same three euro on a parcel worth one hundred and forty euro is two point one per cent. A flat charge behaves like a steeply regressive tax on the low end of your catalogue, which is exactly where cross-border sellers live. This is the single number worth putting on a whiteboard tonight: take three euro, divide by your own average declared value, and write the percentage on the wall where pricing sits. If that number is above ten, your price architecture needs attention long before any application date arrives.

There is a second component nobody can price yet, and honesty about it matters. The coverage mentions a handling fee on top of the flat duty, without putting a figure to it in anything I have seen. So do not build a business case on one number when the structure has two. Assume conservatively that a handling charge of two euro per parcel applies, purely as an illustration, and repeat the arithmetic: five euro of cost on a nineteen-euro parcel is twenty-six point three per cent of declared value, five euro on a forty-euro parcel is twelve point five per cent. Every downstream decision changes shape depending on whether that second number is one euro, two euro or five euro, which is why your plan should be built as a range with a named owner waiting to swap in the real figure when it appears, not as a single spreadsheet.

Then put it next to what the parcel is actually worth to you. Assume your contribution margin per parcel after goods cost, picking and packing, line-haul, last mile, payment processing and a returns provision is four euro and ten cents, which is typical for a low-ticket category where shipping is cheap and returns are not. Three euro of new duty leaves you one euro and ten cents, meaning roughly seventy-three per cent of contribution gone. Add the assumed two-euro handling charge and the parcel is now contribution-negative by ninety cents before anybody has considered the financing cost of inventory sitting in transit. This is what I mean when I keep saying the flat charge hits the bottom of the price ladder hardest: it is not a percentage problem, it is an absolute-sum problem, and absolute sums do not care how thin your margin already is.

The obvious escape hatch is to bundle, and there is something in it, though less than people hope. Assume twelve per cent of your orders currently ship as two parcels or more, averaging two point four parcels per order for that slice. Moving to consolidated picking could plausibly reduce monthly parcel count by eight to twelve per cent; assume ten, so forty-two thousand becomes thirty-seven thousand eight hundred. That removes four thousand two hundred parcels of duty, roughly twelve thousand six hundred euro a month at three euro each, or about thirty-one thousand five hundred euro a month once you add the assumed two-euro fee. Those are real numbers and worth chasing, and they require nothing more glamorous than a picking rule and a packing bench reorganisation, with a daily reconciliation of parcels shipped against orders fulfilled so nobody quietly drifts back.

Raising average order value works on the same arithmetic from another direction. Assume you do nothing but introduce a two-unit bundle on your best-selling low-value item, lifting declared value from nineteen euro to thirty-four euro while the parcel stays one parcel. The duty stays three euro and the assumed fee stays two euro, so the total per parcel is unchanged at five euro, but it now has two units to carry it: two euro fifty per unit instead of five. That is halving the per-unit burden without touching the carrier or the platform. Assume thirty-five per cent of eligible orders accept the bundle, an assumption you should test rather than trust, and the blended per-unit cost across those lines drops substantially. The catch is that bundling changes your marketing economics and your returns profile at the same time, and returns into Europe carry their own paperwork questions, so the number that matters is contribution after returns, not contribution before.

Pushing declared value above the 150 euro line is the move people suggest most confidently and least carefully. Once a parcel is above the ceiling it stops getting the simplified treatment, so it is no longer three euro plus a fee; it goes onto whatever the ordinary duty rate for its classification is, plus the ordinary clearance process and its own handling economics. Assume a heading carrying a four per cent duty rate for illustration, and assume a parcel declared at one hundred and fifty-nine euro: the duty there is about six euro thirty-six, before any clearance or handling cost, against five euro under the flat structure on the assumptions above. So crossing the line is not automatically expensive, and it is not automatically cheap either; it depends entirely on the classification and on what the clearance chain charges at your volume. Anybody telling you there is a universal right answer to this without opening your tariff heading is guessing, and guessing with your money.

That last point is where my own fingerprints are all over the problem. Everything above assumes the classification on each line is right, and for most small-parcel operations it is not. Marketplace listings get written by people optimising for search, they carry descriptions like alloy decorative pendant for home, and the classification that gets generated from that text depends on whatever logic sits behind the label generator. Some of those descriptions are fine, some are the kind of thing that draws questions, and almost none of them were reviewed by somebody who has read a tariff schedule.

Assume you have one thousand eight hundred active EU-facing SKUs and a realistic review capacity of one hundred lines a day with a merchandiser who can answer questions about materials and construction. That is eighteen working days of solid work, call it just under four weeks, and it produces something you will use for years: a single table with classification, English product description, material composition, country of origin and net weight, owned by one person, versioned, and feeding every label.

Four weeks is also the honest answer to a question I get asked constantly, which is what can be done tonight versus what cannot. Tonight you can change what happens to parcels not yet manifested. You can hold a SKU out of EU listings, you can block a declared value below a floor, you can stop printing labels for items with no classification on file, you can reprint a day of labels if you catch an error while the vans are still loading. You cannot put classification data into a system you do not have, you cannot reconcile a year of import value in an evening, and you cannot renegotiate platform terms before tomorrow's truck. Sorting those two categories explicitly, in writing, with times against them, is most of the difference between a seller who absorbs this calmly and one who discovers in week six that half their catalogue has no data behind it.

Now the part that should worry grown-ups more than the three euro, which is the fine. The penalty range of one to six per cent is not assessed per parcel; it is measured against the value of EU imports over the previous twelve months, which is a turnover-scale number, not a box-scale number. Take my illustrative seller: forty-two thousand parcels a month at nineteen euro average is nine million five hundred and seventy-six thousand euro of declared value a year, and if every parcel counts as an import then that figure and your EU revenue are close to the same thing. One per cent of it is about ninety-six thousand euro. Six per cent is about five hundred and seventy-five thousand euro. In other words, the top of that range is roughly one twentieth of whatever you sell into Europe in a year, and it lands because somebody repeated a mistake, not because the maths changed.

Sit with the implications of that structure for a minute. The exposure sits at entity level, so it does not scale with a single bad decision, it scales with everything you sold while the decision was wrong. And the number it is measured against is one you have to be able to reconstruct from your own records if a question ever arrives. Try it this week: ask three people in your company for your EU import value over the last twelve months and see whether you get three answers within five per cent of each other. My experience is that you get something from the marketplace settlement reports, something different from what logistics invoices imply, and a third figure from whatever was declared at the border, and nobody has ever reconciled them. That reconciliation is unglamorous work, one owner, monthly cadence, tolerance written down, and it is precisely the kind of record that turns an unpleasant conversation into a short one.

It also changes how you read your commercial terms. If the platform is the importer of record for these parcels, the platform is the party that carries the weight of getting the data right at the border, and platforms respond to that by pushing requirements down the chain to whoever supplies the product, usually at speed and usually through updated seller terms. I am not going to tell you what your agreement says or what it permits, because reading a contract and reaching a conclusion is somebody else's job and I have no interest in pretending otherwise. What I will tell you is operational: find the clauses in each platform agreement that deal with product data, customs information, indemnity and audit rights, screenshot them into a shared folder with the date you captured them, and put one named person in charge of watching for changes in those sections. Terms get updated quietly, and the notification usually arrives in a dashboard nobody opens on a Friday.

Data flow is where parcels actually go wrong, so let me lay out the trail for one box. A listing holds a description and sometimes a classification. An order carries a price and a destination. A label generator turns the order into a customs form, translating whatever it has into whatever fields the form needs. A carrier consolidates labels into a manifest and lodges it. A customs representative or the platform's own agent deals with what comes back. Every one of those hops can lose fidelity: a discount applied at checkout not reflected in declared value, a gift wrap line item lost, a material field blank, a quantity expressed in sets rather than pieces. The person who can fix each hop is different, which is why fixing parcel data always takes longer than the eighteen minutes anybody expects, and why I always ask clients to draw that map with real names and real cut-offs on it before they start sending emails.

Low declared values deserve their own paragraph, because this is the trap with the longest tail. Under-declaring was always a bad idea and structurally it is getting worse, not better: a three-euro flat charge does not care what you declared, so cutting declared value buys you nothing on the duty side while increasing the size of the contradiction between your settlement records and your customs records. Assume a seller has been declaring at sixty per cent of transaction value on some lines to stay under psychological thresholds. Those twelve months of suppressed figures are also the base any penalty would be measured against, so the same habit that saved nothing now looks like a documented practice. If you have any of that in your history, deal with it deliberately rather than hoping nobody looks, and take advice from someone qualified before you make representations that contradict what you filed earlier.

Incoterms and the buyer-facing contract have a role here too, smaller than sellers expect. For direct-to-consumer parcels the platform route generally decides who deals with customs, so the main lever left is your price and your assortment rather than your terms. If you also ship bulk into the EU for warehouse replenishment, that is a different file entirely, with its own importer, its own declarations and its own values, and it should not be reconciled using the same spreadsheet as parcels, because mixing the two is how people end up quoting the wrong twelve-month number. Keep the two books separate, label them clearly, reconcile both monthly, and when somebody asks for the figure that a penalty would be calculated on, be very sure which of the two books they are talking about.

Storage and dwell cost are the other quiet drain. A parcel held because its data does not satisfy anybody is a parcel accruing handling charges and failing a delivery promise, and for a nineteen-euro item the arithmetic turns ugly fast. Assume a held parcel costs four euro in additional handling and re-delivery attempts, and assume one point five per cent of parcels get held once the new structure is in place, which is a generous assumption given how much classification data is missing in this trade. That is six hundred and thirty parcels a month on our forty-two thousand, roughly two thousand five hundred euro in extra handling, plus whatever refunds and replacements the resulting negative reviews cost you. Two thousand five hundred euro will never make a headline, but it is a predictable recurring amount that stops entirely once the master data is correct, which makes it one of the highest-return projects in the whole programme.

On timing, let me be concrete about the week ahead rather than vaguely urgent. Assume today is the fourth of September, the day this passed. Within one working day, freeze any new EU listing that lacks classification, origin and English description, and stop propagating bad records into your catalogue at the source. Within three working days, produce the first version of the EU twelve-month import ledger and reconcile it against settlements and logistics invoices, tolerance five per cent, gaps flagged rather than silently smoothed. Within one week, pull the current platform terms for every EU-facing channel into a dated folder and assign one owner per channel. Within four weeks, finish the one-thousand-eight-hundred-line classification and origin table. Two of those four items cost nothing but attention, and they are the two most likely to be skipped.

About the alternatives people will propose, I have opinions. A European fulfilment warehouse where you import in bulk and ship domestically changes the shape of the problem completely: you move from forty-two thousand individual import events a month to perhaps ten or fifteen bulk entries, each with proper documentation, and importantly you stop paying a per-parcel charge on every order. Whether that pencils out depends on the duty rate on your headings, the VAT treatment of those domestic sales under local rules, storage rents and pick-pack fees, and it also brings a whole new record set to maintain, so it is not a shortcut and I would not sell it to anybody as one.

Dropping the lowest-value tail of your catalogue is faster and cheaper than any structural change: assume your bottom fifteen per cent of SKUs by declared value account for twenty-two per cent of parcel volume and eleven per cent of contribution. Cutting them removes parcels, improves the average and costs you a little less than you fear. Changing the transport mode does not help at all here, because this is a customs structure, not a network problem; nobody has ever solved a per-parcel charge by finding a faster plane.

Before the weekend I would want three things in writing, and each of them is small enough to finish in an afternoon. Put a declared-value floor in the operating manual, with nothing going to the EU below it, and choose the number by looking at your own contribution per parcel rather than at somebody else's blog. Start a daily reconciliation between parcels shipped and orders fulfilled, hold one person to it and write down the tolerance you accept, because split shipments show up completely in that comparison and nowhere else. Name one owner for each EU channel's data relationship with the platform, including whoever is responsible for noticing changes in the terms, because otherwise every update lands in the inbox of somebody who left the company. None of these require the application date, none of them cost money, and each of them reduces exposure whether the final text is softened, delayed or implemented exactly as described.

The last thing I would say is about where this leaves the paperwork habits you already have. A three-euro flat charge and a turnover-based penalty reward exactly one behaviour, which is having one clean record of what you sold, what it was, where it was made and what you told customs at the time. That record cannot be built during a crisis, only maintained, and maintaining it is dull monthly work with a tolerance number and a person's name against it. Cross-border sellers are very good at everything except that kind of dullness, because dullness does not show up in conversion dashboards. This particular reform is the kind of change that quietly puts a price on it. So yes, tonight's parcel can still be fixed if you catch it before the manifest goes out, but the reason tomorrow will be calmer is not tonight's heroics. It is whether somebody started building that table last month.

↑ Back to the key points

— By Derek Xu