Supply Chain Action Points
Read this first — the conclusion, and the moves to make:
- By 10 August 2026, pull twelve weeks of arrival data on your two main lanes and compute your own on-time percentage plus the 90th and 95th percentile miss in days, instead of quoting the global 62.6 percent.
- By 14 August 2026, reset every customer-facing ETA to your own 90th percentile and put the percentile in the contract rather than a fixed number of days.
- By 21 August 2026, for south China origin, pre-book trucking and berth windows ten days out and hold 15 percent of volume switchable to an alternate load port, activatable within 24 hours of a typhoon signal.
- By 31 August 2026, split North Europe discharge across two gateways with at least 20 percent on the secondary one, and pre-clear the inland leg by truck or rail for the top 20 percent of SKUs by margin.
- By 30 September 2026, re-measure the lane; if reliability is still below 70 percent, move one weekly sailing to another service or origin port for the next contract year and budget the extra five days of pipeline stock permanently.
Rotterdam, Antwerp-Bruges and Hamburg are running elevated yard utilisation with inland bottlenecks and mounting vessel delays, prompting carriers to omit calls and shift to alternative gateways. Typhoons and heavy rain in southern China are separately delaying berthing and trucking connections. Sea-Intelligence put global schedule reliability at 62.6% in June, squeezing Asia-Europe transit buffers.
The Analysis
Spread the chart from the South China Sea to the North Sea and you see the same weather in two places at once. At the eastern end, typhoons and heavy rain are breaking up berthing windows along the south China coast and cutting the truck legs that feed them. At the western end, Rotterdam, Antwerp-Bruges and Hamburg are running yards at high utilisation, with inland connections jammed and vessels waiting. Carriers are doing the only thing left, which is to omit calls and point cargo at alternative gateways. Sea-Intelligence put global schedule reliability at 62.6 percent for June.
Sixty-two point six is not a port number. It is a network number, and it is being read as if it were a service level.
Start with what the figure measures, because most of the argument about it uses the wrong definition. Schedule reliability counts the share of port calls arriving inside the proforma window. It is a frequency, not a magnitude. A call six hours late and a call nine days late are the same event inside that percentage, so 62.6 tells you how often the schedule failed and nothing at all about how badly. For inventory, the badly is the entire question.
This is the sentence the coverage skips. Plan to the proforma date and you are wrong 37.4 percent of the time, with no idea by how much. What a planner needs is the shape of the tail, the ninety-fifth percentile of arrival, because that is the number that decides safety stock, promise dates and penalty exposure. A carrier's on-time percentage is a marketing statistic. The tail is an engineering one, and only engineering can be planned against.
At the European end the failure is not at the berth. Yards high, inland connections jammed, vessels waiting: read that list backwards and you have the causal chain. The yard fills because boxes cannot leave, not because they arrived too fast. Inland capacity is the binding constraint, and inland capacity is the least elastic part of the system, since a barge slot, a rail path and a truck appointment are all permissions granted by somebody else.
Omitting a call is not a remedy, it is a transfer. A skipped call does not delete the cargo; it stacks it on the next call at the same port or pushes it onto a feeder or another gateway, and the gateway receiving it was chosen because it had spare berth capacity, not because it had spare inland connections. That is how congestion migrates. It moves to the node with the most headroom and consumes the headroom first, which is why the secondary port always looks fine for a fortnight and then does not.
The eastern end fails differently. Typhoons and heavy rain compress berthing windows and break the truck legs, and both effects look temporary enough to dismiss and cost enough to remember. A berth window lost to weather is not recovered by working harder the next day; it is recovered by rolling cargo, and rolled cargo is a promise broken at the other end of the line. What makes this season unusual is only the coincidence: the weather at the load end and the yard at the discharge end are both seasonal, they overlap, and a network keeps one buffer for both.
Who absorbs it depends on how many gateways they can use without renegotiating anything. A large importer with three ports and an inland network treats an omitted call as a routing decision taken on a Tuesday. A mid-sized importer with one gateway and one warehouse has no such move and pays in demurrage and in broken promises. An LCL consolidator suffers most, because a rolled box inside a groupage container is not one shipper's delay, it is a container's worth of delays on one bill. Exporters at the south China end face the mirror problem: the truck cannot reach the gate, the gate cannot reach the berth, and the vessel is the only one this week.
The transmission runs fast at the front and slow at the back. Berthing delays show in days. Truck and gate problems in a week. Yard utilisation in two to three weeks. Inland dwell and empty repositioning in about a month. Schedule reliability, the number everyone quotes, is published monthly and describes a month that has already ended, so it is a lagging indicator being used as though it were a warning.
Do the arithmetic, assumptions mine. Take a shipper moving forty FEU a month. At 62.6 percent reliability, roughly fifteen of those boxes sit on a call that misses its window, assuming the global figure applies to the lane, which it may not and which probably flatters a good lane and hides a bad one. Assume the average miss is four days and the ninety-fifth percentile is seven, both mine and not the article's. Fifteen boxes times four days is sixty box-days of schedule nobody planned for. Hold a promise at the ninety-fifth percentile and you carry seven days of demand as pipeline stock, which at forty a month is about nine FEU sitting in the water permanently, doing nothing except protecting a date.
Without the omissions, the queue at Rotterdam would lengthen until vessels waited at anchor and the delay compounded around the rotation. Omission is the valve and it is the right valve. The cost is simply forwarded to the shipper whose box was on the skipped call, and nobody sends that shipper an invoice with the line item on it.
The trap is reading a seasonal coincidence as a structural decline and signing three-year capacity at the wrong node. The other trap is the mirror image, assuming the schedule will recover and writing customer dates at proforma while the tail is still seven days long.
Chokepoints on a chart are rarely closed. They are avoided. Avoiding one for a season is contingency; avoiding one for two years is a new map, drawn by people who never meant to draw one. The deep structure here has not moved: the same water, the same canals, the same three river mouths. What moved is the tolerance for a promise. Watch the tail, not the percentage.
↑ Back to the key points