Evri Buys Cross Border Connect, Puts a 30% Cost Cut on the UK-US E-Commerce Lane

The Analysis

Common enough scene: twenty to eleven on a Tuesday night, somebody on your customer team has three tracking pages open at once and all three say the same thing. Arrived at gateway. Fourteen hours unchanged. Nobody has slept properly since Monday because an American customer was promised five days, and the paragraph underneath the status line explains nothing at all.

This week Evri Group said it will acquire Cross Border Connect, a Boca Raton platform built for getting British e-commerce into American hands and getting it back again. Six US ports of entry. End-to-end clearance run on their own broker network. Once a parcel clears, direct injection at the nearest carrier hub rather than a truck ride across town to be sorted by somebody else. Evri's commercial chief put the achievable cross-border cost cut at up to 30 per cent, stacked on top of Evri's own air rates. Evri handles more than 1 billion parcels a year and absorbed DHL eCommerce UK in 2025, so this is not a minnow nibbling at a lane; it is the biggest parcel operator in Britain buying the front end of its own American pipeline.

I am not going to spend your time on trade policy or freight indices. Neither of those will rescue a parcel sitting in a queue tonight. What matters on this lane is the clock, because the clock is the promise, and 30 per cent off an invoice nobody can read is worth less than 18 hours back on your worst day of the quarter. So let's cut the thing open hour by hour and find the joint that actually moves.

Start with what was actually announced, because this deal has already been reported in a way that makes it tidier and bigger than it is. Evri is buying Cross Border Connect. CBC runs what its own side calls an asset-light network: 6 US ports of entry, its own clearance capability end to end, and injection into whichever carrier hub sits closest to the consignee once the parcel is released. Asset-light is the phrase doing the heavy lifting and I will come back to it. Evri merged with DHL eCommerce UK in 2025 and moves more than 1 billion parcels a year, most of them domestic. The number everyone has put in a headline is 30. The number almost nobody has interrogated is 6, because 6 is where the hours are.

Here is the version of this lane most retailers live with today, written in hours because days are how you lie to yourself. An order lands at ten in the morning and is picked and scanned by one in the afternoon, call it T+3h. First-mile pickup scan at four, T+6h. The van reaches the UK gateway around midnight and spends the small hours being screened, consolidated and built onto a pallet, T+14h. Wheels up around eight the next morning, T+22h. Seven to eight hours of flying plus an hour of taxiing and the pallet is on the ground at the one port everybody uses, T+30h.

Then comes the interesting stretch. Eighteen hours of average wait before release, with a bad tail that runs past 60. Up to 24 hours trucked to a national or regional facility that is not part of the delivering carrier's network. Then injection. Then 24 to 72 hours of sorting, line-haul and a final attempt at somebody's front door. Add it up and the median is around 120 hours, with the ninetieth percentile somewhere near 168. If you are promising five days, you are promising your median, and you will break that promise about half the time.

So which hour goes first? Not the flight. Every operator on this lane flies roughly the same metal at roughly the same speed, and when somebody tells you their aircraft is faster than the competition, ask them what time it lands and what time it is released. The hours are lost on the ground, in exactly two places. The stretch between wheels-down and release, where deconsolidation meets a queue and a building full of people waiting for another building to say yes. And the intermediate sort sitting between release and injection, a facility your parcel did not need to visit, run by people whose throughput targets have nothing to do with your customer. Those two together are where the median goes to die and where the ninetieth percentile gets manufactured. Everything else is rounding error.

6 ports of entry is about the first problem plus a routing decision that most retailers have never been allowed to make. Landing port is the reset button on the whole clock, because the port decides which direction the truck points next. Clear everything on one coast and every parcel bound for a cluster two thousand miles away buys itself 48 to 72 hours of cross-country line-haul before it even starts being delivered. Clear it somewhere the delivering carrier already runs a hub near the destination and that line-haul shrinks into a local run or vanishes under a zone skip.

6 turns where do we land from a fixed constraint into a choice, and a choice is the only variable anybody in your operations room can actually pull. Common situation this time of year: the retail calendar says the goods have to be delivered by a certain date, the whole shipment went through one gateway, and a third of those boxes have to cross three time zones of road before they reach anybody. That is not weather. That is a landing-port decision nobody got to make, and this acquisition is a bet that deciding it is worth real money.

There is a triage question hiding inside all of this that most retailers answer by accident. When a hundred parcels slip, which twenty do you chase? The instinct is to chase whoever shouted loudest, and whoever shouted loudest is rarely the most expensive failure. Sort the slipping ones by how far past their own point of no return they already are against the cut-off printed on the storefront, then by whether the consignee sits inside one of the clusters that a different landing port would have served better. On the distribution I just described, the parcels worth intervening on are the ones somewhere between hour 36 and hour 72, because before that they recover on their own and after that nothing short of re-shipping changes the outcome.

Throw every hour of discretionary capacity at those and spend not one minute on anything already past hour 120, other than writing to the customer before they write to you. Twenty minutes of sorting at 8 in the morning beats 4 hours of firefighting at 4 in the afternoon, and almost every operation I have stood in does it the other way round.

Direct injection at the nearest hub is the second piece, and it is the one that shows on a tracking page as a straight line instead of a zigzag. Every extra facility a parcel touches costs roughly 12 to 24 hours and adds one more building where somebody has to find it. Take out one intermediate sort and one intermediate line-haul, which is what nearest hub means in practice, and 18 hours come off the median on every parcel routed that way, with considerably more off the tail, because tails are made of missed connections and backlogs in buildings that were never designed for your volume. It reads like plumbing when you see it written down and it feels like oxygen when you have ever had to explain to a customer why their parcel went to a city they do not live near.

Here is why I keep talking in hours rather than percentages, and it is the reason I distrust every write-up of every parcel deal. Hours at the front compound into days at the back. Push your warehouse cut-off 2 hours later and you might catch one more evening departure; catching that departure is not 2 hours, it is 24, because the alternative leaves tomorrow. Miss an injection window by 40 minutes and you have bought a whole extra day. Nobody loses 20 minutes on this lane. You lose 20 minutes now and 24 hours later, and that window closes and you are done. Every single time. The reason it happens is that the window belongs to whoever owns the building, and until this deal you did not have much choice about which building that was.

Right, the 30 per cent, because somebody in finance has already put it in a model. Let me write my assumptions down so you can argue with them instead of with my conclusion. Assume a 1.2 kilogram apparel parcel, 40 pounds of retail, product cost 11 pounds, platform commission 15 per cent at 6 pounds, payment processing 3 per cent at 1.20, leaving a contribution pool of 21.80 before logistics. Assume today's cross-border cost per parcel is 12.34, built this way: air line-haul at 3.20 a kilo is 3.84; US port and clearance handling 1.60; one unnecessary intermediate sort 0.70; last-mile injection and delivery 4.90; returns amortised at a 12 per cent return rate against 7.50 of reverse cost is 0.90; service and exceptions 0.40. Contribution per order 9.46, which is 23.65 pence of every retail pound. Take 30 off the whole 12.34 and you save 3.70, contribution goes to 13.16 and your margin moves from 23.65 to 32.9. 9.25 points. If that arrived I would be writing a much calmer column.

It will not, and here is the honest arithmetic. The pieces a new owner can genuinely change are the line-haul arrangement, the port and clearance handling, and that unnecessary sort: 3.84 plus 1.60 plus 0.70 is 6.14, and 30 per cent of that is 1.84. The last mile at 4.90 is not theirs to cut. It is set by whoever delivers inside the United States, and no acquisition in Florida changes what an American carrier charges to put a box on a porch. So the realistic first cut is 1.84 off 12.34, which is 14.9 per cent of your bill, taking margin from 23.65 to 28.25. 4.5 points. Worth having. Roughly half of what was promised.

Getting closer takes two things that are not automatic. A zone-skip out of hub injection: enter the network at the right facility instead of being line-hauled across a zone boundary and assume 0.55 off that 4.90, which only exists for parcels whose destination cluster matches the port they landed at. With that and nothing else you are at 2.39 saved, 9.95 of cost, 19.4 off the original bill and a margin of 29.6. That, not 30, is the number I would put in a plan.

There is also a weight at which even that evaporates, and that is where up to earns its keep. Run the same model on a 0.4 kilogram accessory retailing at 18 pounds. Line-haul collapses to 1.28 because it scales with kilos. Nothing else does: clearance and port handling are per parcel, the intermediate sort is per parcel, last mile stays flat around 4.90. Total 9.31, addressable pieces 3.58, 30 per cent of which is 1.07, or 11.5 per cent of the bill. Run 3.5 kilogram boots at 80 pounds retail. Line-haul 11.20, bulky last mile 6.80, returns heavier at 1.35, total 22.05, addressable pieces 13.50, saving 4.05, which is 18.4 per cent.

Nothing in that spread reaches thirty except by rounding. My honest range for a 1.2 kilogram 40-pound garment is 15 to 22 per cent, and the gap between that and 30 is the gap between somebody's press line and your own profit and loss. Notice who said 30: the commercial chief of the company selling you the service. Nobody is lying to you. Thirty is a ceiling quoted by a person whose job is the ceiling.

Now the part nobody has written, and it is the one thing I would carry into my next negotiation. Every headline has counted the outbound side of this deal. Not one of them has counted returns, even though returns decide whether a British apparel retailer makes money in America at all. Try it with the same assumptions. 12 per cent come back, at 7.50 each. That 0.90 per outbound parcel sitting in my cost stack is the reason a generous returns policy eats the entire margin on a 40-pound sale. If CBC's network moves the reverse leg through the same six ports, the same clearance arrangement and the same injection logic, and there is nothing in the structure of this deal suggesting it does not, then the addressable cost is not only going out, it is coming back.

Assume reverse cost drops from 7.50 to 5.20 because the parcel re-enters through a port and a hub that were already being paid for: 2.30 back on every return, 0.28 on every outbound parcel at 12 per cent, which takes total saving from 2.39 to 2.67, cost to 9.67, the discount to 21.6 and margin to 30.3. Push the return rate to 25 per cent, which is what fashion with a loose size policy actually does. Reverse per outbound parcel today is 1.875, after the change 1.30, saving 0.575, total saving 2.965, cost 9.375, 24 per cent off, margin 31.1. The higher your return rate, the more of the promised 30 lives on the reverse leg rather than the outbound one. Nobody put that in their headline because returns are nobody's favourite slide.

The counterfactual first, because it changes how much of this you should believe. Had Evri not bought CBC, what would this lane look like twelve months from now? Almost exactly like today. One nominated gateway per retailer, one broker relationship, one injection point, and every November the whole British e-commerce sector queuing behind itself in the same building waiting for a pallet to be released. The change is real even where the 30 per cent is not. Then the argument I have with the deal's own logic: asset-light is a polite way of saying we rent.

In an ordinary week renting is cheaper, easier to reconfigure and better than owning assets you only use eleven weeks a year. In the third week of November all six boarding houses fill up at once and the operators who own their own buildings get served first. Optionality that depends on somebody else's spare capacity tends to evaporate exactly when you need it, which is why I want every promise here written down as a number with a date on it rather than a ceiling.

And yes, there is a question people are politely stepping around: Evri absorbed DHL eCommerce UK in 2025 and now owns the American entry points as well, which means the number of people you can phone about this lane went down again. Consolidation always shows up wearing the costume of savings.

Timing, in hours and in dates, because knowing when this reaches you is the difference between planning and being surprised. For the next two weeks nothing moves; your parcels travel on existing contracts through existing gateways and anybody telling you otherwise is selling something. Between week two and week twelve the routing rules change or they do not, which means the port-choice option either appears inside your shipping software or stays invisible, and if it stays invisible none of this is available to you no matter what got announced. This peak season is the first honest test, because single-port tail risk is what gets ugly under volume: the 18-hour average wait before release becomes 60 in the third week of November, and anyone who has lived through that week knows the average is the least interesting number in the distribution.

If you want any of this before then, the decision closes inside two weeks. Past that, out to roughly 180 days, the interesting change is reverse routing: a returns network takes longer to stand up than an outbound one because labels, consolidation points and the customer's own returns portal all have to agree with each other, and none of that is a switch anybody flips. Anything you hear about returns capability before January should be read as intent rather than capability.

Which leaves the only bit that matters. Somebody will put 30 per cent in a deck. Somebody else will tell you direct injection is a technology story. It is not technology. It is geography and floor space and whose building your parcel sleeps in overnight, and it shows up on a tracking page as 12 to 24 hours at a time. Get the parcel back first and argue about the ceiling afterwards.

↑ Back to the key points

— By Nadia Pryce