UPU: US Parcels Still Down 36.8% in July as EU Volumes Drop 31% Under the EUR3 Duty

The Analysis

Yesterday a client rang me at 07:40, before I had sat down. Four hundred and twelve parcels into Belgium, handed over on a Friday, and by Tuesday morning 214 of them had not moved. He wanted to know whether the Belgians had stopped buying phone accessories. The Belgians had not stopped buying anything. Two hundred and fourteen parcels were sitting in a shed because nobody had declared three euros on each of them, and his marketplace clock had been running the whole time.

That call is the whole story behind this month's UPU numbers, and almost nobody is reading the numbers that way. The 2026 State of the Postal Sector report puts US-bound small parcels down 36.8% year on year in July, after the fall had narrowed to 33.8% in June, and it puts EU-bound small parcels down 31% in July, the month Brussels ended the EUR150 exemption and put a flat EUR3 on every item instead. Every headline I saw wrote that up as a collapse in demand. Demand had nothing to do with it.

The line in the same report that nobody quoted is the only line that tells you what to do on Monday morning. Postal operators sending to the United States went from 176 to 133 and were back only to 159 by July 2026. That is not a demand curve. That is a route map with seventeen lights still off.

Three numbers decide how you read this month, and only one of them made the headlines. US-bound small parcels fell 70.3% year on year in September 2025, the month the de minimis rule ended. That fall narrowed to 33.8% by June 2026 and then widened again to 36.8% in July. EU-bound small parcels fell 31% in July. Dutch customs show 46% and Belgian customs 53% for the same month. Run those in that order and you do not have a demand story. You have a channel that lost its exemption and then lost its shape.

A de minimis threshold was never really a tax break. It was a queue removal. Under it, a parcel did not get declared, did not get assessed, did not wait for anybody to pay, and did not need an importer-of-record story. It went from the export scan to the last mile without stopping. That is how a postal channel moved a EUR14 phone case out of a warehouse and onto a doorstep in six days for less than the price of a coffee. What the exemption bought was not a discount. It was the removal of three stops in the middle of the timeline, and the middle of the timeline is where parcels die.

So let me run the clock the way I run it with a client, because the 36.8% lives in the clock and not in the market. Hour zero is the order, and in most marketplaces the promise starts there and is counted in days, which is already a mistake, because everything that goes wrong later happens in hours. Hours zero to six are pick, pack and label. Hours six to twenty-four are handover and export scan. Then the linehaul, which on a postal small parcel is anywhere from thirty-six hours to five days depending on whether it rides a scheduled consolidation or whatever belly space was spare. Then the arrival scan. Then the declaration. Then the assessment. Then the payment. Then release. Then last mile, which is usually the only leg anybody has ever actually measured.

Under the old rule the declaration, the assessment and the payment took zero hours. Zero. Now they take real hours, and nobody put those hours into the promise. So when a seller reads 31% and concludes that Europeans stopped buying, what actually happened is that a queue appeared in the middle of the timeline and their parcels are standing in it while the promise clock runs.

The hour that blows first is the hour after the arrival scan, and it does not blow the same way for everybody. If the duty was settled at checkout and the data travelled with the manifest, the parcel walks through in a few hours and you will never recognise this story at all. If nothing was settled and nothing was filed, the parcel waits for somebody to notice it, and then it waits for somebody to pay, and those are two different waits. One is a data wait. The other is a money wait. Postal operators were built to move envelopes, not to chase importers for three euros, and a fair number of them simply stopped chasing. I would bet that is a good part of the 43 operators that dropped out between 176 and 133.

Now put it against a promise. Say you sell on five days, 120 hours, and you hand over on a Friday. Linehaul takes 60 to 72. That leaves 48 to 60 hours for arrival, declaration, assessment, release and last mile, and last mile eats 12 to 24 of those. The middle therefore has somewhere between 24 and 36 hours of slack. A declaration queue running at 72 hours eats the promise three times over. Your cancellation rate is not a demand signal, it is a queue signal, and it lands in your defect metric four to six days after handover, which is long enough that you will spend a fortnight blaming the wrong thing, exactly the way my Belgian client did on that Tuesday morning.

The shape of the fall tells you who is actually bleeding. It is not spread evenly. The EU average is 31% and the two big gateway customs are at 46% and 53%, which means the damage sits where parcels enter the continent and not where they are bought. Demand does not fall hardest at the shed door; queues do. Inside your own book the split is just as uneven. A seller of EUR12 baskets has lost two thirds of gross margin to a fee that is 25% of the goods. A seller of EUR120 baskets has lost 2.5% and barely noticed the fee, and is only fighting the dwell. Sample and spare-part shippers sending two or three parcels a week are almost untouched, because their problem was never the exemption. The postal operators themselves took the worst of it, with 43 of them leaving the US lane inside a year and only 26 of those coming back by July. Consolidators and brokers picked up whatever the operators dropped.

And note the direction of travel, because it kills the comforting idea that this heals by itself. The US fall was 70.3% in the first month, then narrowed all the way to 33.8% by June 2026, and then widened again to 36.8% in July. Twelve months in, the channel did not recover, it re-worsened. The operator count went 176 to 133 and crawled back only to 159. If you have been telling yourself this is a transition that will settle, the July number is your answer: it is not settling.

Here is what I have not seen in one single write-up of this report, and it is the line that will actually cost you money. The EUR3 is per item, not per parcel. A parcel with one phone case pays three euros. A parcel with four phone cases pays twelve. Read a flat fee against a basket and it is savage in one direction only: three euros is 25% of a EUR12 item, 15% of a EUR20 item, 5% of a EUR60 item and 2.5% of a EUR120 item. One rule, four completely different businesses. It is nothing at all to whoever sells mid-priced goods and it is an execution to whoever sells cheap ones, and no coverage I read wrote that sentence, because a single flat number reads as small.

The second thing nobody said out loud is where the 36.8% is measured. It is a handover-side number. It counts what leaves postal operators. It says nothing about what reached a door. A seller whose parcels moved from post to express looks like a 36.8% collapse in the UPU table and looks like flat business in their own book. A seller whose parcels are stuck in a shed looks identical in the UPU table and is having a disaster in their own book. The report cannot tell those two apart. You can, in about an hour, by pulling dwell time by lane out of your own system and sorting it, and that hour is the highest-return hour available to you this week.

Then there is the triage question, which is the one I get asked most and the one the report cannot answer: which parcel do you save first, and which customer do you steady first. I do it on three keys and I do it in the first ninety minutes of the working day. Parcels past 48 hours of dwell go first, because past 48 they stop being delays and start being cancellations. Then parcels with the highest declared value inside that group, because those are the ones carrying the biggest marketplace penalty and the angriest buyer. Then customers with a repeat rate worth keeping, because a first-time buyer who waits six days writes a review and a repeat buyer who waits six days stops buying. Everything still under 24 hours of dwell can wait until tomorrow; it is not in danger yet and your attention is not free.

The clock the customer is watching is shorter than the one you are watching, and it decides whether a delay turns into a refund or just into grumbling. Assume a buyer who ordered on Monday evening expects to see movement by Thursday. If tracking has not moved 36 hours after handover, they write a message. If it has not moved 72 hours, they open a case, and a case costs you more than the parcel does on most platforms. So the move is to send the explanation at hour 30 rather than at hour 80, and to send it with a new date you can actually hold. A note at hour 30 with a real date costs you a discount code. The same note at hour 80 costs you the order and a defect point.

There is also a clock inside your own four walls, and it is the cheapest one to fix. A parcel handed to the operator at 18:00 on Friday and declared at 09:00 on Monday has burned 63 hours of a 120-hour promise before anything whatsoever happened to it. That is more than half your promise, spent on a weekend, and the shed does not care. Move your internal cut-off on the gateway lanes to Thursday and you buy the whole weekend back, for the price of a scheduling argument with your own warehouse.

There are two ways this reading could be wrong and I want both on the table. It might be a demand story after all, in which case the right move is to cut spend and re-plumb nothing. I do not think the numbers say that, because the fall is concentrated in the two gateway customs at 46% and 53% rather than spread evenly across the EU's 31%, and demand does not fall hardest exactly where parcels enter a continent. It might also be undone by Brussels, if the flat fee becomes ad valorem or a threshold comes back, and then the basket sort flips and cheap goods return first. What decides which world you are living in is your own average declared value, and you can read it this afternoon out of your own order table.

Let me price a business that looks like the ones I sit with, and I will show you the assumptions because they are mine and not the report's. Assume 20,000 small parcels a month into the EU. Assume 40% of them, 8,000, enter through the Dutch or Belgian gateways, because that is where the linehaul lands. Assume an average declared value of EUR18 and 1.4 items per parcel. Assume a fixed declaration and handling cost of EUR1.80 per parcel on top of the duty. Assume a five-day, 120-hour marketplace promise.

The duty line: EUR3 times 1.4 is EUR4.20 a parcel, and across 20,000 parcels that is EUR84,000 a month. Against an EUR18 basket at a 35% gross margin of EUR6.30, the duty takes two thirds of gross margin before a single cent of freight is paid. That is not a tax problem. It is a pricing problem in a tax costume, and it cannot be fixed by finding a cheaper operator.

The dwell line bites harder, because nobody budgets it at all. Assume the parcels you did not pre-declare sit 72 extra hours. Assume each extra 24 hours in transit cancels 1.5% of what is in that window. Three days is 4.5%, and 4.5% of 8,000 gateway parcels is 360 parcels a month. Cost a failure at EUR14: EUR9 of freight you will never recover, EUR2 to restock, EUR3 of service time at six minutes a parcel. That is EUR5,040 a month, and 360 parcels at six minutes each is 36 hours of your service team's month spent typing where is my order. Thirty-six hours is a person for a week, every month, absorbing a queue you never priced.

The fix side is quick. Assume collecting the duty at checkout and filing it with the manifest takes 60 of those 72 hours out. Cancellation exposure drops from 4.5% to roughly 0.75%, so about 300 parcels a month survive: EUR4,200 back and 30 hours of service time back. Against that, assume your team needs 20 hours to rebuild the checkout and the manifest template. You are whole inside a month, on one lane, and every lane after that is margin.

What I told the Belgian client that morning, in this order, and I would tell you the same. Get the manifest to the broker that day with a declared value on every line, because a broker holding data can clear in hours and a broker waiting for yours will wait exactly as long as you make them. Hold every Belgium-bound parcel that is not declared until it is, because six hours sitting at your dock costs you nothing and seventy-two hours in a shed costs you the order. Then stop selling a five-day promise into a lane that is running a three-day queue; lengthening the advertised promise is cheaper than the cancellations, and narrowing the market you sell into is cheaper than both. Then open the book and sort it by declared value, lane by lane, and ask whether three euros is 2.5% of what you sell or 25% of it.

Four ways out, and each one has a bill. Moving from post to express does not touch the EUR3 at all, it only buys a shorter queue and a broker who chases, so you are paying freight to convert a data wait into speed, and above EUR60 baskets that trade is usually the right one. Consolidating into bulk and declaring once saves the fixed declaration cost and not the duty, because the duty counts items, and it adds twenty-four to forty-eight hours at the consolidation point, so it only pays where the promise has slack. Holding stock inside the EU turns thousands of duty events into one and makes your last mile a domestic one, which is the real answer for anything you sell over and over, and the bill is inventory risk plus a building you now have to run. Raising the basket is the cheapest of the four and the hardest to get signed off internally.

The pits are all one pit wearing different hats. Reading 36.8% and cutting ad spend is the biggest, because you would be cutting into demand that never left you. Waiting for the postal operator to tell you is another, since the operator counts parcels and you need hours, and parcels are not hours. Declaring at arrival instead of at checkout is simply choosing to pay the queue. Assuming those seventeen dark routes came back quietly on their own will cost you a quarter.

Save the parcel first, argue about the rule later. The UPU's own estimate of what slow decisions cost this sector is US$70.8 billion a year, and I do not need to defend that figure to make my point. My client did not lose 214 sales because Belgians stopped wanting phone cases. He lost them in the ninety hours between an arrival scan and a declaration, and that window closed on him while he was still reading the headline about demand.

↑ Back to the key points

— By Nadia Pryce