China-Europe rail from Xi'an ships 34,000 cooling appliances to Europe in 12 days vs 40 by sea

The Analysis

The two places where a China-Europe train stops being a train are the gauge changes, one at the Chinese border and one before Poland, and that is where I would start if you asked me why a box out of Xi'an lands in Mannheim in twelve days on a good week and takes visibly longer on a bad one. At each of those points the box is lifted off one set of wheels and set down on another, and a line of wagons forms behind it. Everything in between is a long, well-run, thoroughly unremarkable run.

The figures from Xi'an are these: more than 34,000 air conditioners and fans to 15 European countries this year, a train that left on 24 August and reached Mannheim in about 12 days against roughly 40 by sea, a departure every 80 minutes on average, and eight consecutive years leading the country on volume and load factor. Read as a system, that is capacity. Read as a shipper, it is a timetable you cannot see from outside.

So let me walk the box instead of the announcement. Where does it stand still, who is standing next to it while it stands still, and which of those waits can be deleted by a decision taken in a different month.

Start at the factory gate and count the links. The units come off a line, get palletised, get loaded into a high cube, get trucked to the Xi'an hub, get declared, get consolidated, wait for a slot, get lifted onto a wagon, cross two gauge changes, arrive at Mannheim, clear, get devanned, get trucked to a distribution centre. The twelve days covers exactly one of those links. Door to shelf is somewhere near thirty days, and that is my planning assumption rather than a figure from the article. It is not a criticism of the railway. It is arithmetic about where the clock starts. The headline clock starts when the train pulls out. Your clock starts when somebody signs a purchase order, and the gap between the two is where seasonal businesses lose a season.

Eighty minutes is a network number and I want to be blunt about what it does not mean. It does not mean a train to your destination leaves every eighty minutes. A hub dispatching at that cadence does it across dozens of destinations, and each one gets its own allocation, which is a fraction of the timetable rather than the timetable itself. In a heatwave every appliance importer discovers this on the same afternoon. The corridor has capacity. Your destination has a slot, and the slot was probably taken by somebody who booked in the spring.

Here is what makes this cargo different from a box of machine parts. A split air conditioner is mostly air. The indoor head is a thin shell, the outdoor unit is a compressor sitting in a box with a great deal of nothing around it, and the cartons add more nothing. Load a high cube and you hit the volume ceiling long before the weight limit. On a train priced by the box that is not a defect, it is the reason the cargo belongs there: you are not paying for weight you are not using. Thirty-four thousand units sounds like a flood. At roughly 120 units to a high cube, which is my assumption and not the article's, it is about 283 boxes for a year across fifteen countries. That is not a market. That is a shelf being kept full.

Who feels it first depends on where the stock sits. An importer with a warehouse inside a day's drive of Mannheim gets the whole twenty-eight days that separates rail from sea, and can place a second order inside the same season. An importer holding stock in northern Italy or southern Spain pays another three to five days of road out of Mannheim, my assumption again, and more importantly loses the second order, because it lands after the weather has moved on. A small retailer with no European warehouse cannot use this product at all, since the minimum sensible unit is a container and a container is a season for them. A large brand with a European third-party warehouse and a six-month plan does not care about twelve against forty. It cares about variance, because the volume is already sold and what it needs is a date it can repeat to a retail buyer.

The heat arrives in days and the supply response arrives in weeks, and no amount of corridor capacity closes that gap. A heat alert buys a retailer a fortnight of extra demand. A rail restock takes a month door to shelf even when the main leg runs in twelve. By the time the forecast is unambiguous the only mode left is air, and putting boxed split units on a freighter only works if the unit is worth more than the freight, which for most of this category it is not. Rail here is not a response tool. It is a pre-positioning tool, and the decision point is the shoulder weeks.

Now the part the coverage leaves out. Twelve days is the train arriving at Mannheim, and the comparison with forty is terminal to terminal. Your promise to a customer is door to shelf, and the last four or five days of that promise cannot be compressed by any railway: the box gets lifted off, cleared, devanned, reloaded onto a road trailer and received at a dock that closes at four in the afternoon. The other omission is that twelve days is one train, and it is reported because it was a good one. Variance on this corridor is manufactured at the two gauge changes, where wagons queue and each box is lifted across. Some weeks that queue moves and some weeks it sits, and nothing on your side of the contract changes which week you get. The number for your planning file is the ninetieth percentile. No press release has ever printed one.

Let me put an account on the table and flag every assumption, because the article gives no freight rate and I will not invent one. Assume a high cube holds 120 split units and each lands at 300 euros. That is 36,000 euros of goods in a box, and twenty boxes is 720,000 euros of inventory moving or waiting. Assume a fifth of it arrives after the selling window and clears at a twenty per cent markdown. That is 144,000 euros gone, and it went because the dates were wrong, not because the freight was dear. Divide by twenty boxes and you get 7,200 euros of markdown exposure per box. That is the only threshold you need from this article. A rail premium under 7,200 euros a box is already paid for by the markdown you did not take. Above that line, the honest comparison is rail against air for the last twenty per cent of the buy, and those are two different products with two different customers.

The temptation this week is to treat twelve days as a service level and promise it downstream. Do not. Promise the ninetieth percentile and keep the difference as your own buffer, because the day you promise a record you have handed a border queue the power to make you a liar in front of a retail customer. The second trap is treating a slot as a booking. A slot is inventory. It has a date, it expires, and in a hot year it is worth more than the rate you negotiated for it.

↑ Back to the key points

— By Ömer Kaya