East Asian gateways log long vessel waits as Typhoon Saudel hits

The Analysis

Nine days at Shanghai. Four at Ningbo. Three at Busan. Two at Singapore. Read that column the way most people read it, as a ranking, and you will conclude that the answer is to use the bottom of the list until the top recovers. The column is not a ranking. It is a photograph of water finding its level.

The figures come from a port update covering 26 August to 1 September, and they follow Typhoon Saudel. The same update notes that pre-peak cargo was already pressing on landside truck capacity and on berth buffers, and that the risk for Asia exports now shows up as rolled cargo and slipping schedules.

What interests me is not the nine. It is the spacing between the four. A ladder that decays that smoothly is telling you something about how the nodes are wired together, and the message is less reassuring than it looks.

Start with what a waiting time measures. It measures the gap between when a ship wanted a berth and when it got one. That gap is not a property of the port alone; it is a property of the port plus everything upstream of it. When Shanghai closed, the ships outside did not vanish, they held. When it reopened, they queued. Every one of those ships had a rotation to keep and a cargo plan to honour, and a carrier with a rotation to keep will do arithmetic you would rather it did not: which call to shorten, which port to skip, which boxes to leave on the quay. The skipped call gets chosen by volume, so the small port pays. That is how a queue at one node becomes a roll at another, and the roll never appears in a waiting-time table, because it happens in a booking system, not a harbour.

The half of this that gets under-reported is the landside half, which the update names directly. A quay works around the clock. A gate does not. Cranes have no appointments; trucks do. So when a terminal reopens, the water side of the queue clears at a rate set by crane productivity, and the land side clears at a rate set by how many trucks can be processed in a working day, which is a far smaller number. The visible metric, vessel waiting, recovers first and recovers fastest. The invisible one, truck turn time and empty return, does not. If you are reading the update to decide whether to load, you are reading the number that will look better before the situation is better.

The four rungs are also a size ladder. Nine days at the biggest gateway, two at the one that is mostly a transshipment hub. Descend and the vessels get smaller, the direct calls get fewer, and the number of times your box is handled goes up. Moving a container from Shanghai to Ningbo is not a lateral step to a port with a shorter queue. It is a step onto a different string with a different rotation, and it pays for the shorter wait in touches: one more lift, one more gate transaction, one more set of papers, one more chance for something to be misread. The wait you save is counted in days. The touches you add are counted in risk. Those are not the same currency and they do not convert at a fixed rate.

Pain follows the cargo. Exporters of heavy, slow-moving, low-value goods hurt first, because their boxes fill before their weight allowance does and they cannot fly. Time-definite retail and promotional cargo hurts second and louder, because a roll is a missed date rather than a bigger bill. Nearly untouched are high-value goods that can go air, intra-Asia traffic that never touches these gateways, and anyone holding a premium allocation on a direct service. LCL shippers are the most exposed of all, for a structural reason: a consolidator protects its base and rolls its tail, and the tail is where small shippers live. Size is not the variable. Position on somebody else's priority list is the variable.

Two clocks run at once. The vessel clock is the one in the update: a ship that waited nine days is nine days late, and everything downstream of it is nine days late with it. The booking clock runs ahead of that, and it rings the moment a carrier decides to cut a call, which is a decision taken days before the ship arrives and taken at the expense of whoever holds the lowest priority on that service. Behind both of them sits a third, slower clock: equipment. Boxes that should have been repositioned are on the wrong side of the ocean, and you learn about it three to five weeks later, at an inland depot, when a booking cannot be covered. The update is a snapshot of the first clock. Your problem will arrive on the third.

Here is what the numbers are not saying, and it is the part I would pin to the wall. A genuinely diversified network should produce a ragged profile: one node congested, the next nearly empty, because the congestion has nowhere to go. What we have is a smooth decay, nine, four, three, two, which is the signature of a connected system rather than a redundant one. Each rung took on what the rung above could not hold, and that is why the profile is smooth. It is good news for this week and poor news for the next one, because the buffer that produced the smoothness has now been spent. There is a second layer under that: all four numbers come from one typhoon. Your alternates are correlated by weather before they are correlated by commerce. A plan that says use Ningbo if Shanghai is bad is not a hedge; it is a bet that the same storm will be gentler two hundred kilometres down the coast.

Put numbers on it, all of them mine and not the source's. Assume thirty FEU a week out of Shanghai and a seventy-thirty split between Shanghai and Ningbo. Nine days against four means the Ningbo third runs five days ahead of the Shanghai two-thirds. Nine FEU times five days is forty-five FEU-days of inventory you are not paying to carry. Assume fifty thousand dollars a box and a cost of capital of eight per cent, and forty-five FEU-days comes to roughly five hundred dollars. That is not a reason to rebuild a supply chain; it is a rounding error next to a missed retail window. Which is the point. The split is not worth doing for the carrying cost. It is worth doing because five days is the width of the gap between making a date and missing one, and because the split is the only lever you can pull without asking anybody's permission.

Suppose Saudel had arrived two weeks later, when the pre-peak build was already loaded rather than still being planned. The same nine days would have landed on committed cargo with customers already holding dates, and the rolls would have been expensive instead of inconvenient. Suppose instead the ladder had been steep, Ningbo at eight and Busan at seven: the overflow would have stopped at the second rung and Shanghai's damage would have been contained, but the second rung would have failed in the same week. Neither version is comfortable. What we got is the one where the network absorbed the shock and sent the bill to October.

What to do is unglamorous. Separate the two numbers in your head, waiting at the water side and turn time at the land side, and stop using the first as a proxy for the second. Then decide in advance, in writing, how much volume can move to a second rung without adding a truck leg you do not have, because the truck leg is where substitution quietly becomes more expensive than the wait. Anything past that is a bet on weather.

The trap is reading a smooth ladder as a healthy network. Smoothness is what a system looks like while it is sharing load, and sharing load is what it does right before it has none left to share. The other trap is booking the alternate in your head. An alternate that exists as an idea costs nothing and saves nothing; it becomes capacity only when there is a named booking on a named rotation with a container allotted to it. Until then it is a rumour you repeat to a customer.

The map does not show this. Four ports, four numbers, one storm crossing a coast. What the map shows, if you read the spacing rather than the values, is a chain of nodes leaning on each other in the same direction, which is fine until the wind does the same thing. Watch the gap between the rungs, not the height of the top one.

↑ Back to the key points

— By Tomás Ferreira