DHL port update: Shanghai delays exceed 5 days, Ningbo and Manila 2 to 5

The Analysis

The delta, drawn small, is two ports and one industrial basin. Shanghai's Yangshan and Waigaoqiao are reported above five days of delay in the 18 September port update, Ningbo sits at two to five, Manila at two to five, and Mundra is past five on an equipment shortage. Further west the picture is a pair of labour disputes at Rotterdam and Hamburg that keep working their way out through the inland legs. The rest of Asia-Pacific, the report says, is settling.

Five days is not a big number and it is the wrong number to focus on. What matters is the word the report used, beyond. Beyond five days is not a measurement, it is an admission that the measurement ran out of resolution. A port that says two to five still believes it knows. A port that says beyond five is telling you the queue has stopped being a queue and become a condition.

I want to take the delta apart first, because it is the one node on earth where a five-day delay changes what a whole basin does next quarter, and then look at the two numbers in this same fortnight that contradict each other. That contradiction is the most useful thing in the report and nobody has printed it.

Start with the scale of the thing rather than the delay. Yangshan and Waigaoqiao together are not a port, they are the seaward end of an industrial basin that produces for the whole world, and Ningbo at two to five days is the other end of the same basin. When one end is beyond five and the other is two to five, the basin does not stop. It shifts. Volume that would have sailed from the Yangtze mouth gets dragged down the coast, the trucks and feeders and box trucks reorganise around the new point of exit, and for a few weeks the region's exports keep moving while every individual shipper's plan is wrong.

That shifting is why the headline feels wrong to people on the ground. A European buyer reading sixty-day averages sees a delta that is marginally slower. A shipper with forty boxes and a promise date sees a fortnight of his year disappear. Both are reading the same system. One is reading the flow, the other is reading the variance, and the variance is where the money is.

There is a second reason the basin absorbs so well and it is structural, not managerial. The delta has two deepwater complexes serving one hinterland, which is a redundancy most coastlines in the world do not have. Rotterdam and Antwerp argue about it. Los Angeles and Long Beach share a Breakwater. The delta simply has more quay than any other estuary, so when Yangshan slows, Ningbo takes the load without anybody writing a memo. That redundancy is a genuine asset and it is also why the region's problems tend to arrive disguised as somebody else's problem.

Now the causes, because they are not the same causes and this is where most reading goes wrong. The report does not assign Shanghai's delay to a single event. It reports it as persisting, which is a different kind of statement, and the absence of a named cause after weeks of persistence tells you the constraint is not a storm you can wait out. Manila's two to five is described the same way. Mundra is the only one with a cause printed next to it, and the cause is equipment shortage. Rotterdam and Hamburg are labour disputes. So of five congested nodes in this report, one has a capital explanation, two have labour explanations, and two have no explanation at all, which in a port is usually the signature of volume arriving faster than the yard, the gate and the rail link can be re-paced together.

The wider point the report makes almost in passing is that Asia-Pacific outside these nodes is settling. Read that against Shanghai being beyond five days and a shape appears. A region going quiet while its largest gateway goes past a threshold is not a regional congestion event. It is a local failure at a node that happens to be load-bearing, and load-bearing nodes export their failures through relay. A feeder that misses a Yangshan connection because the mother vessel was held five days does not report the delay at Yangshan. It reports it at the next port, on the next line, in somebody else's lane, as a booking that could not be confirmed.

Here is the contradiction, and it is the one thing in this fortnight I would put in front of a board. One congestion reading dated 16 September put Mundra at 1.38 days of waiting. This report, dated the eighteenth, puts Mundra beyond five days on equipment shortage. Two days apart, four days of disagreement, same port. Both can be true at once, and they are true because they are measuring different ends of the same shipment. A berth wait counts vessels. An equipment shortage counts boxes. When there are no empties and no chassis, the ships berth on time, the quay looks healthy, the vessel-side number prints 1.38, and your cargo sits because there is nothing to put it in. The port is fine. The steel is not.

That distinction is worth more than the buffer because it decides which remedy applies. A berth queue clears by waiting, and waiting is cheap and it ends. An equipment shortage clears when somebody moves boxes from where they are to where they are needed, which is a network decision made by a carrier or a leasing pool, on a vessel that has to have space, at a cost that gets allocated to somebody. It does not clear because time passed. I have watched importers treat an equipment problem as a queue, hold the booking, hold the promise, and lose three weeks to a constraint that was never going to resolve itself on their calendar.

Ask your forwarder one question about Mundra before the next booking and make the answer written: not how long is the wait, but are empties available at the depot for my pickup date. Those are different questions with different answers and only one of them predicts whether your cargo moves. If the answer is a range, treat it as a no.

The Shanghai side needs one more piece of mechanics, because it changes the shape of the delay curve. When a gateway holds vessels for more than five days, arrivals bunch. Ships that would have come in on Monday, Wednesday and Friday come in as a flotilla on Friday, and then two things happen in sequence: the yard takes a spike it did not plan for, and the inland legs take a spike a few days later. So a shipper who decides to wait for Shanghai to clear does not return to a normal port. He returns to a port digesting a release, which is a second wave of delay that looks like a new event and is actually the first one arriving late.

Rotterdam and Hamburg are the same lesson with a different mechanism, and they are the reason I keep telling people that a gateway is a port plus its links. The dispute is at the quay. The signature is three hundred kilometres inland, where the box that cleared the terminal sits waiting for a barge slot or a rail path that the labour action has quietly unscheduled. On a port-status dashboard this looks like success. The port did its job, the vessel was worked, the container left the terminal. The gateway failed, and the failure is recorded in a system that nobody in your company looks at.

That is why the report's own remedy for those two ports is about haulage rather than berths. Pre-booking inland capacity is not a way of avoiding the strike. It is a way of not being at the back of the queue that the strike created, because when the corridor reopens the demand for barge and rail slots arrives all at once from everybody who has been waiting.

Ningbo deserves a warning of its own, and it is a warning about the release valve. Every shipper reading this report will do the same arithmetic and move volume to Ningbo, which is exactly the right individual decision and a poor collective one. Valves tighten when everyone uses them. Two to five days is a range, and the honest reading of a range is the top of it, because a port that could say five would say five. Quote Ningbo at five and you will be roughly right whether it prints two or five; quote it at two and you will be wrong most weeks.

Manila at two to five is doing a different job in the network, which is why its number should not be read as a smaller version of Shanghai's. For southeast Asian flows it is often both a destination and a relay, and a relay port's delay multiplies through connections rather than adding to transit. A box that misses a Manila feeder does not arrive two days late. It arrives on the next sailing of that string, whenever that is. The number on the page is a port statistic. The number in your plan is a schedule statistic.

Now scale, because the same report means different things at different sizes. At 200 FEU a month across four lanes you have options that a smaller shipper does not. You can hold contracted space at two delta gateways, you can decide that 60 percent of your volume goes through the port with the better week and 40 percent through the other one, and you can rebalance that split without renegotiating anything. The cost of that structure is a little administrative pain and a small premium on the second gateway. What it buys is immunity to any single node's bad month, which in a year like this one is worth more than the premium.

At 40 FEU a month you have no such structure and it is a waste of a Tuesday to pretend otherwise. Your options are the date you quote, the haulage you pre-book, and the sentence you say to the buyer. The first two cost money and the third costs only nerve, and in my experience the third is the one that determines whether the account survives the quarter. A buyer told at booking that the window is forty days because the delta is past a threshold will build forty days into his plan. A buyer told on day thirty-six that the box is late will remember the day, not the reason.

Category decides which of those three levers matters most. Apparel and anything with a launch date has an immovable calendar, so the buffer has to come out of production or out of the sailing choice, and the cost of a five-day slip is not freight, it is markdown. Consumer electronics have high value density, which means the carrying cost is trivial and the right answer is to pay for certainty and stop optimising, because the margin absorbs the premium and does not absorb a missed launch. Bulky low-value goods are the hardest case in the whole matrix, because the float is too small to justify action and the detention clock is large enough to hurt, so the only lever that works is getting the box out of the terminal before the free time runs.

Then there is the question of which node you actually touch, which people answer with the port they booked from rather than the port their cargo passes through. If you export from the Yangtze basin your exposure is the five-day threshold and the release wave behind it. If your southeast Asian flow touches Manila you are exposed to connections, not transit. If your Indian cargo goes through Mundra you are exposed to steel, not to water. If your European distribution centre sits behind Rotterdam or Hamburg you are exposed to a corridor that a labour dispute has quietly unscheduled, and your port dashboard will never tell you.

The transmission has dates and they are worth writing down. The opening wave is this week: which quotes leave with a buffer, which haulage gets booked, which buyer gets told the truth early. Nothing in the report will change before Monday, and everything you intend to do has to happen before the box is committed, because after that you are managing an explanation rather than a delay.

The second wave is two to six weeks out and it is where the money gets set. A quote built on this week's threshold becomes a rate built on this week's threshold, and by late October nobody remembers why the number is what it is, only that it is. This is also when rotations bend, and a rotation change moves your transit further than the port delay ever did, without anybody sending you a note about it. The third wave is a season, and it is behavioural: if every shipper in the delta moves 25 percent of volume to Ningbo for eight weeks, Ningbo's two to five becomes four to seven and the report prints it as a new problem in November. Networks learn. That is the part of congestion that is caused by the response to congestion.

Put numbers on the small case, with the assumptions on the table. Assume 40 FEU a month, average value 60,000 dollars a box, carrying cost 6 percent a year for money and shelf space only. That is 3,600 a year per box, 9.86 a day, call it 10. Five days of Shanghai threshold is 50 dollars a box, 2,000 dollars a month across forty boxes. A shipper looking at that will conclude the delay is affordable, and on this arithmetic he is right. That is exactly why the arithmetic is dangerous.

Assume, purely for illustration, that recovering a late box costs ten times the daily float, through a rescue shipment, a discount, an expedited inland move. Recovery is then 100 dollars per box-day, and five days across forty boxes is 20,000 dollars rather than 2,000. Whether ten is your multiple I cannot know. What I can tell you is that recovery cost is not linear in the delay, it is a cliff, and it sits where a promise date is missed rather than where a box is late.

Scale the same case up. Assume 200 FEU a month, 60 of them on the delta lane, same 60,000 dollar value, same 6 percent. Five days on sixty boxes is 300 box-days a month and about 3,000 dollars of float, which is still small against the freight spend. But run the calendar instead of the money. If those sixty boxes serve a ninety-day selling season and each loses five days, you have removed 5.5 percent of the season's selling days from 60 boxes at once, and you have done it in the fortnight when the season's sell-through is decided. Nothing in the carrying-cost line will ever show that. The margin line will.

The failure mode, if none of this is done, is worth stating plainly because it is the default. The quotes go out on the range's midpoint, which is what everybody quotes and what nobody believes. The boxes arrive late, four or five of them at once because the release wave is real. The inland bookings were made after discharge, so they are at the back of the corridor queue. The buyer is told on the day of arrival that the date slipped. The quarter ends with a discount, an airfreight rescue on two SKUs, and a buyer who will now ask for a longer window on every future order, which is a permanent cost disguised as a lesson learned.

The condition that would make me wrong is worth naming, because a forecast without a falsifier is a mood. If the next monthly update puts Shanghai back inside two to five and Mundra's equipment position is described as restored, then this is a nodal pulse, the delta did what redundant basins do, and the correct response was a five-day buffer and nothing more. If instead the delta is still beyond five in the November update, then this is not weather and not a bad month. It is a capacity and corridor problem that will not be solved by quoting differently, and the response changes from buffering to re-contracting for the first quarter while the market still thinks it is temporary.

One more thing, on the ten-year scale, because it explains why the remedy feels wrong. Ports solve their constraints in a fixed order, and the order takes decades. Water first: you deepen the channel, you build on islands, you run a bridge out to them because the city has taken the shoreline. Then land: you fill, you automate the yard, you stack higher. Then the corridor: the rail link, the barge terminal, the road that has to cross a city to get to a gate. By the time a port is world class at the first two, its binding constraint has moved to the third, and the third is the one that cannot be solved with capital alone because it runs through somebody's labour agreement and somebody's municipal politics. That is precisely where the delta and the German gateways are now. The cranes are not the problem. The path away from the cranes is.

So, concretely. Before Monday 28 September, rebuild the promise date on every delta lane, using a 35-day base transit as the stated assumption: forty days where Yangshan or Waigaoqiao is involved, thirty-five where Ningbo is, and never the bottom of a range. For Rotterdam and Hamburg, book the barge or rail leg ten to fourteen days before the vessel is due, not after discharge, and keep the booking reference in the shipment file so somebody can prove when it was made. For Mundra, ask for empty availability in writing against your pickup date and route anything with a fixed shelf date elsewhere until you have the answer. For your delta split, no single gateway above 60 percent of monthly volume for the next eight weeks, with a live Ningbo option at 25 to 30 percent.

Then make the reading a habit with an owner and a time, because the report is monthly and the threshold moves between editions. One person, the day the update drops, fifteen minutes, a table of the six gateways you actually touch with three columns: the number, the cause if the report gives one, and whether the number is vessel-side or cargo-side. That third column is the one that would have caught Mundra. It takes a minute to fill and it is the difference between reading a delay and reading a port.

Ports are slow objects and their failures are slow too, which is why a five-day number can be a season's problem. The map has not changed shape this month. What changed is that one node on it stopped telling us how long, and started telling us how much. When a gateway crosses from a duration into a condition, the answer stops being a buffer and becomes a decision about where your cargo lives. Draw the basin, mark the two exits, and know which one you would use on a Monday when the other one is beyond five.

↑ Back to the key points

— By Tomás Ferreira