Ningbo-Zhoushan Clears Its Typhoon Backlog With Berth Waiting Back to About One Day

The Analysis

Spread the map out on the table and put a finger on the southern edge of Hangzhou Bay. Ten years ago this stretch of coast was a story about berths — how many, how deep, how many cranes per metre of quay. The question that decides whether this node holds is no longer that one. It is how fast it comes back after the sky falls in.

Ningbo-Zhoushan has just answered that question in public. In September it worked 1,926 container vessel calls and moved 4.27 million TEU, up 11.5% on a year earlier and 15% on August, a monthly record. Average daily throughput went from 130,000 TEU to 142,000, and across the seven days of Golden Week it ran at 145,000 a day. January to September reached 34.67 million TEU, up 6.8%. By 9 October berth waiting was back to roughly one day, which is normal, and in-berth productivity was running 5% above the January-to-August mean.

Read those lines as a recovery story and you have read them the way every headline did. Read them as a bill and you get somewhere more useful, because none of it came free. Clearing that typhoon backlog took 74 vessel call changes, 56 berth switches and 18 double calls. The record is the headline. The price is the story.

Start with the arithmetic the port has handed us, because it checks out and that matters. September's 4.27 million TEU over thirty days is 142,333 TEU a day, which is the same number the port gives as its daily rate of 142,000 — so the record is not a rounding exercise, it is a real step up from the 130,000 baseline. Spread the month across its 1,926 calls and each vessel moved an average of 2,217 TEU. Now set September against the year: 34.67 million TEU over nine months is about 127,000 a day, so the month that just closed ran roughly 12% above its own annual average. A port does not do that by accident. It does that because two things arrived at once and both had to be finished inside thirty days.

The two things were the typhoon and Golden Week. The backlog from the storm was still sitting on the yard when the pre-holiday export push began, and the port had to work off the debt and run the peak in the same month. That is the honest explanation for a month that grew 15% on August and 11.5% on a year earlier while the nine-month number sits at 6.8% — September ran well ahead of the year's own trend, and the gap between 11.5% and 6.8% is where the catch-up lives. I will come back to that gap, because it is the hinge on which the whole thing turns.

What deserves attention is not the height of the number but the speed of the recovery, and specifically the fact that the recovery happened under load. Berth waiting back to about a day almost always means one of two things: the port caught up, or the ships stopped coming. Ningbo-Zhoushan's own figures rule out the second. Vessel calls at 1,926 and throughput at a monthly record are not what a quiet month looks like. This port absorbed a typhoon, paid the backlog down, and held a queue of roughly one day while running at 142,000 TEU a day. Call that what you like; I call it elasticity, and elasticity is the scarcest thing in this network.

How was it paid for? The numbers are sitting right there in the same release and almost nobody quoted them. Fifty-six berth switches. Eighteen double calls. Seventy-four vessel call changes. Think about what each of those actually is. A berth switch is the port treating its own berth calendar as a re-arrangeable resource rather than a fixed appointment book — the vessel does not lose its slot, it loses its address, and the yard plan, the crane plan and the truck gate all move with it. A double call is the same logic pushed one step further: one hull worked through two windows in the same port complex instead of one. Neither of these shows up as capacity anywhere. Both of them are borrowed capacity, and you borrow it from the schedule.

That is where the 5% comes from. In-berth productivity running 5% above the January-to-August mean did not arrive with a new crane. It arrived because for a few weeks in September the port ran itself the way a chess player runs a bad position, moving pieces to make the board fit the position rather than waiting for a better board.

One small thing worth flagging, because it tells you how these numbers are made. The Chinese version of this story reports 74 vessel call changes and three container services adjusted. The English version reports the same 74 call changes but breaks the rest into 56 berth switches and 18 double calls. Neither is wrong. One counts the actions, the other counts the outcome in liner services — and a port can describe the same fortnight either way depending on whether it wants to be read as an operator or as a network. Hold on to that, because the rest of this piece is about exactly that gap: the throughput number is an outcome ledger, and the price is an action ledger, and they are never printed on the same sheet.

Now follow the bill outwards, because a port does not pay its own bills. The carrier pays first and it pays in rotation. Seventy-four call changes inside one port in one month means the order in which vessels visit that port over the following fortnight or so is not the order that was published when the bookings were taken. Some of those changes are permanent for the rest of the season: once a service has re-sequenced its calls to recover a week, it does not always go back. The shipowner books the cost as schedule integrity. The shipper experiences it as a vessel schedule that no longer matches the booking confirmation.

The forwarder pays next, in feeder and second-leg connections. A mainline vessel that arrives a day and a half off its published window does not just deliver late; it misses the feeder it was supposed to hand to, and transhipment boxes sit in Ningbo for another three to five days waiting for the next one. Inland, the trucking side pays in a quieter way: a berth switch can move the box from one terminal to another inside a port complex that is spread along a long stretch of coast, and the short-haul run, the gate slot and the driver's hours all change with it.

And there is the neighbouring set of ports, which is where the network view starts to matter. Shanghai, Taicang, Jiaxing and the other Yangtze Delta gateways take the diversion when Ningbo-Zhoushan is full, and they are the reason a backlog here clears in weeks rather than months. Ten years ago we would have described that as competition. Spread the map out and it looks more like redundancy, and redundancy is what you build when the thing you are protecting is not market share but continuity.

Timing is where this becomes a task rather than a story. Call changes land inside hours of the decision. Rotation changes propagate over two to three weeks — that is roughly the window in which the affected services re-sequence and settle into a new pattern. A shipper's ETA, which is derived from the old rotation, is stale for fourteen to twenty-one days after the change and often nobody sends a notification. Add seven to ten days of ocean leg for a Europe service and the arrival shift lands at the consignee a month later. By the time it shows up in a European or North American replenishment cycle we are thirty to forty-five days out, and by then the original typhoon is a footnote in a report nobody reads.

Here is the point I have not seen in any of the coverage, and I will put it plainly so there is no mistaking where it sits. The cost of clearing that backlog was not absorbed by the port. It was transferred onto the vessel schedule, and it does not appear in any of the numbers the port reports. Throughput, berth waiting and in-berth productivity are the three figures every headline carried, and none of them records rotation changes, because rotation changes are not a port metric. They are a carrier metric, and eventually they are a shipper's inventory problem. So the same fortnight that produced a monthly record for the port produced, for the people whose goods move through it, a schedule that quietly stopped being true. "The port has recovered" and "the schedule no longer matches" are not two stories. They are one story told from the two ends of the same container.

Seventy-four is the number that carries that cost. Spread across a dozen trunk services it is only about six call changes on the lane you actually use, which sounds small until you remember that a re-sequenced call does not shift by an hour. It shifts by a sailing.

Two judgements in this piece could be wrong, and both are testable inside six weeks. The first is whether September's 11.5% was demand or catch-up. If a meaningful part of it was backlog being released, then late October throughput should fall back below 140,000 a day and possibly toward the 130,000 baseline, and the record becomes a pulse rather than a turning point. If it holds, the 6.8% annual trend is the number to plan against and the record was an acceleration. Watch the daily rate, not the monthly total.

The second is where the 5% productivity gain came from. If it came from scheduling — the 56 berth switches and 18 double calls — then it is a one-off, drawn from a reserve the port cannot spend every month, and the January-to-August mean remains the honest planning number. If part of it came from real works, yard or equipment, then it persists, and it persists in exactly the conditions that matter. A port that can run 5% harder when it is recovering from a storm is a different asset from a port that can only run 5% harder because it borrowed three weeks from its own schedule.

And the counterfactual that should worry anyone shipping out of this coast: the timing was lucky. The typhoon arrived early enough that the backlog could be cleared before the holiday peak rather than into it. Had it landed two weeks later, a one-day berth queue would not have come back before Golden Week, the diversion to the neighbouring ports would have been real rather than theoretical, and the rotation damage would have run into November. Nothing about that was designed. It was weather.

Which brings me to the decade view, which is the only view that makes this more than a month's news. Put the map down and look at what a hub is actually being asked to be. Berth count used to be the answer, and on that measure this coast won years ago. What is being asked now is different: the ships are larger, so a single berth lost costs more than it used to; the weather is less predictable, so the interruption comes more often; and the tolerance for a late arrival at the other end has gone down, because inventory models were built on schedules that held.

Under those three conditions the things that decide whether a hub keeps its traffic are the ones you cannot photograph — how many separate port areas can stand in for each other, whether the approach channel takes the largest ships at all states of tide, how much anchorage sits behind the fairway when a queue does form, and how fast the whole complex can be re-planned.

Ningbo-Zhoushan is spread across a string of port areas along one deep-water coast — Beilun, Chuanshan, Meishan, Daxie, Jintang and the rest — and the structure matters more than the sum of its parts, because a typhoon does not close all of them equally. That is the bone this network has grown in the last ten years: not more quay, but more ways to keep working when part of the quay is out. Ten years ago this was a different question entirely. The hubs that understood that early are the ones whose throughput dips and recovers. The ones that did not are the ones whose traffic leaves and does not come back.

Let me put a number on what this costs a shipper, with the assumptions standing where you can argue with them. Assume you are an exporter moving 40 containers a month to Europe out of Ningbo. Assume the 74 call changes spread across a dozen trunk services, so about six fall on or near the service you use. Assume each re-sequenced call moves the ETA by one to two days, and take 1.5 days. Assume half your monthly volume, 20 boxes, ships inside the affected window, and that for six of them you end up having to gate into Shanghai instead, roughly 200 km by truck at an assumed USD 2.50 per kilometre. Those six reroutes cost USD 500 a box, or USD 3,000.

Assume those 40 boxes carry USD 50,000 of goods each, so USD 2m sitting in transit, and charge 1.5 extra days of transit against a 10% annual cost of capital: USD 2m times 0.10, divided by 365, times 1.5, is about USD 820. Add the reroute money and the shift costs roughly USD 3,820 for the month. Still all assumptions, and still not the end: if your margin is USD 400 a box, 40 boxes is USD 16,000 of gross profit, and USD 3,820 takes about a quarter of it.

The port did not invoice you for any of that. The port's numbers for September are, on their own terms, excellent. The bill arrived on the vessel schedule, and it arrived in a form that no throughput table will ever show you.

So what to do, and by when. Re-verify every Ningbo booking against the carrier's current rotation within fourteen days, not against the rotation you booked on. Anything whose ETA differs from your confirmation by more than a day gets a written correction request, with a target of no unresolved gaps by 20 October. Build a before-and-after rotation table for the three to five services you actually use and send it to your customers and consignees before the end of October; telling a consignee in advance that the schedule moved is a service, and telling them after the vessel arrives is an argument.

Have a tested fallback before mid-November. Ningbo to Shanghai or Taicang sounds like a phone call, and in peak season it is not: trucking capacity is tight on both ends, cut-off times differ, yard availability differs, and dangerous or out-of-gauge boxes cannot simply be walked to another terminal. Write the plan down with names and numbers, and set yourself the standard that the decision can be executed inside twenty-four hours. The pit that everyone falls into is assuming the alternative port has spare capacity because the primary one is full; in a peak week they are usually full together.

Reset in-transit inventory by two days, three on Europe services, in the first week of November, and stop planning on August transit times. Watch the daily throughput figure through late October and treat it as the signal: two consecutive weeks under 135,000 TEU and September was backlog release, so do not build next year's plan on the record; sustained above 140,000 and you plan to the 6.8% trend and book capacity accordingly.

And before the annual contract round, put storm resilience into the questionnaire. Ask the carrier which alternative port areas it can berth at in this complex, under what conditions a call change or a port change is free, and how many hours of notice you get on a rotation change — forty-eight hours is a reasonable floor to ask for and worth more than a small rate concession, because a rate concession you can predict and a schedule you cannot are not the same kind of cost.

Spread the map out one more time. This network has grown quickly enough that we got in the habit of measuring it in records, and September gave us another one. Records are kept by ports. Accounts are kept by the people whose goods are on the water. The storms will come again, and next time this coast will do the same thing — shift berths, double up calls, re-sequence sailings, and hand the world a one-day queue and a monthly record. Somewhere in that fortnight a rotation will be quietly rewritten, and it will land, weeks later, in somebody's inventory as two days that nobody budgeted for. What has changed here over ten years is not the size of the number. What has changed is the bone: the strength of a hub is no longer how much quay it has, it is how completely it can absorb a shock without handing the bill to somebody else.

↑ Back to the key points

— By Tomás Ferreira