Durban Backlog: 30,352 Moves, 11.3 Days of Quayside Work, 251-Hour Anchor Waits

The Analysis

Spread the map out and run a finger down the south-east coast of Africa, to where the warm current turns and a city built as a coaling station still decides the fate of a sub-continent's boxes. Ten years ago this was the door: every container for Johannesburg, every spare part for the mines, every crate of citrus going the other way crossed these quays because there was nowhere else to cross. The map has changed since. Walvis Bay has woken up on the far side of the continent, Nacala and Beira pull the northern inland, Maputo has opened a north-eastern alternative, and Durban is no longer the only door. It is the biggest door in a wall that now has four.

The Cargo Movement Update with data to 15 October tells you what that door looks like this week. Eleven vessels are still at anchorage outside Durban Gateway Terminal. The average wait at anchor runs about 251 hours, and once a ship gets alongside it spends another 127 hours on the berth. The backlog stands at roughly 30,352 container moves, which the port itself calls about 11.3 days of quayside work. Waterside throughput at the terminal rose 6% on the week to about 4,734 TEU a day, and across the country the terminals handled 88,926 TEU in the week of 5 to 11 October, also up 6%. Transnet moved 460,166 TEU of containers in September, 20% above August and still 5% below a year earlier.

Add the two waiting numbers and you get 378 hours, which is 15.75 days. That is the figure to hold in your head, not 251 and not 127. It is how much of a vessel's life evaporates at this single node, and on a service whose round trip runs somewhere near eight weeks it is close to a quarter of the rotation burned in one port. Headlines will call it a backlog. It is a capacity event wearing a backlog's clothes, and for anyone with boxes on the water the difference between those two readings is money.

Start with the arithmetic, because the report hands you enough of it to check itself. Transnet's 460,166 TEU across September works out at about 15,339 TEU a day spread over thirty days. The week of 5 to 11 October, at 88,926 TEU, is 12,704 a day. Put them side by side and the current week is running roughly 17% below September's daily rate, on the assumption that both series cover the same set of terminals. That assumption is worth confirming with your agent, because if it holds, the celebrated 6% weekly gain is a rebound off a weak base rather than a recovery. Durban Gateway's own 4,734 TEU a day is about 37% of the 12,704 the country moved that week, which is the other number to carry: more than a third of South Africa's container flow passes through one gate.

Now the figure everyone quoted, and the one nobody checked. Thirty thousand three hundred and fifty-two moves is not 30,352 TEU. A move is a lift, and in a trade dominated by forty-foot boxes heading inland each lift averages somewhere near 1.7 to 1.8 TEU. Take 1.75 and the backlog is roughly 53,000 TEU. Divide that by the 4,734 TEU a day the terminal is currently running and you land on 11.2 days, which is the 11.3 days the port published. The two agree almost exactly, and that agreement is the story: the official clearing estimate assumes the terminal gives every unit of its present daily output to the backlog and does nothing else for eleven days.

No terminal does that. Eleven ships are at anchor right now and more are behind them, and every one of those ships wants a share of the same 4,734 moves a day. The honest reading of 11.3 days is that it is a floor, the time clearing would take in a world that stopped arriving. If half of daily capacity goes to vessels turning up now, the backlog needs more than twenty days, and it grows on any day the cranes underperform. That is the line the headlines skipped, and it is the one to plan against: not eleven days, but a range whose bottom edge is eleven.

Why now? Because September was a catch-up month and October inherited the bill. Transnet's 460,166 TEU was 20% above August, which reads like a recovery until you notice it was still 5% below the same month a year earlier. A month that beats a bad month and still loses to last year is not growth, it is a rebound, and a system that rebounds this hard inside thirty days spends capability rather than adding any. The 6% weekly improvement is the same phenomenon at terminal scale: better than last week, not yet better than the year.

Look at the year-on-year line instead of the month-on-month one and a different picture appears, and it is the one that explains why this backlog is deep rather than shallow. A terminal network that has been handling less cargo than the year before, quarter after quarter, does not keep its surge capability intact while it shrinks. Gangs get smaller, maintenance gets deferred, the crane fleet gets older, and the slack that used to absorb a bad fortnight quietly disappears. Then a peak arrives and a system that once flexed cannot. That is what a minus 5% year and a plus 20% month look like when they turn up in the same table, and it is why I would not plan on a fast normalisation. Ports do not lose the ability to flex in a month, and they do not get it back in a month either.

Layer the impact and it does not fall evenly. The 15.75 days bites anyone with a box on the water, but it bites differently depending on what is inside and where the box goes afterwards. A Johannesburg importer of vehicle components running to a line schedule eats it twice, once in the extra stock needed to keep the line turning and once in the premium paid to fly the single part that cannot wait. A citrus or table-grape exporter eats it worse, because that clock is not a production clock but a shelf-life one, and a week at anchor in October is a week out of a European market window that shuts on somebody else's calendar. Mining inputs sit in between: heavy, low value per tonne, indifferent to delay right up to the hour a plant runs dry.

Then there is the inland map, which is where this stops being a South African story. Durban is the door for Botswana, Zimbabwe, Lesotho and Eswatini, and for a good share of Zambia and the southern DRC. Those corridors have no substitute that works at short notice. Walvis Bay is two thousand-odd kilometres of road from Johannesburg, Maputo is a real alternative for the north-east but its corridor has its own ceiling, and Dar es Salaam and Nacala drain a different catchment altogether. When this door narrows, the landlocked interior pays twice, once in days and once in a trucking rate that climbs when trucks have to queue.

On timing, count backwards from the berth rather than from the anchorage. The 251 hours at anchor is mostly the carrier's exposure in the narrow contractual sense, because demurrage free time on most import bills starts at discharge rather than at arrival. The 127 hours alongside is where your own clock starts, and the yard behind it is where it compounds: a terminal running flat out to clear a backlog has a full yard, rationed truck appointments, and a gate that is the slowest link in the chain. In practice that means adding ten to fifteen days to any Durban transit time you are currently planning, assuming the variance is uglier than the average, and treating the day the box is discharged as day one rather than the day the ship turned up.

Here is the step that turns a berth delay into a network event, and it is the one most importers miss because it happens on somebody else's balance sheet. A weekly service on a 56-day round trip needs eight ships. Every extra day burned at one port has to be absorbed by stretching the rotation, by adding a vessel, or by dropping a call. Fifteen and three-quarter days at Durban is the equivalent of about 2.25 vessels on a string that size. Nobody holds 2.25 spare ships, so what actually happens is that the rotation stretches, the buffer call disappears, and the service runs late into every subsequent port for the following two or three months. Which means the damage is not confined to South African cargo: a box discharged three calls after Durban is riding the same broken rotation. If any of your volume sits on a string that touches this port, this is your problem even if your container never goes ashore here.

One exposure deserves its own line because it never appears in a container count. A reefer sitting at anchor for 251 hours runs on ship's power and is monitored by the carrier's crew, and most of the time that is fine, until it is not, and then the failure mode is a claim rather than a delay. Ten and a half days is long enough for a small temperature excursion to turn into a quality dispute, long enough for a gap in the monitoring record to become a documentation problem, and long enough that cargo insurance rather than the carrier ends up being the instrument you have to rely on. If any part of your book moves under temperature control through this port, ask for the reefer monitoring record and the plug-in log at discharge rather than at claim time, and check that the policy treats a long wait at anchor as an insured event rather than an excluded delay. Perishables do not wait for a backlog to clear.

What would turn this is not dramatic, and that is exactly why it deserves watching. The terminal sits at 4,734 TEU a day and is improving 6% a week. Compound that for three more weeks and you are near 5,600, at which point 53,000 TEU of backlog really is eleven or twelve days of work and the queue begins receding by mid-November. Lose a week instead, to a crane failure, to wind, to a stoppage, and the gain vanishes, because a queue this deep has no slack to absorb a bad week. So the trigger is not the weekly percentage. It is whether the daily rate holds above 5,000 TEU for two consecutive weeks. One good week is weather. Two is a trend.

One more read of the weekly pair, because it is easy to get wrong in both directions. National throughput rose 6% and Durban Gateway rose 6%, the same step in the same week, and that co-movement supports two very different stories. Either the country's terminals are recovering together because the constraint was temporary and is passing, or Durban is being partly relieved by spillover and the network is shuffling the same volume between gates. The monthly number settles it. Transnet moved 5% less than a year earlier in September, so the system is not handling more cargo than last year, it is handling less, which means the second story is the likelier one: spillover, not recovery. For an importer that changes the question from when will this clear to which gate is taking the overflow and what the inland leg from there costs me.

Which leaves the question I keep returning to on this coast. Is this a pulse or a turn? The honest answer is that it is a pulse in the data and a turn in the map. Backlogs clear; that is what backlogs do. What does not clear is the memory of them. Durban is no longer the only door, and every fortnight of 251-hour waits is an argument shippers will remember at the next corridor review. Ten years ago this stretch of coast was a story about how many berths there were and how deep they had been dredged. Today it is a story about which door the cargo trusts, and trust, once it moves, comes back slowly if it comes back at all.

Put a number on it, assumptions first. Say you import forty forty-foot containers a month through Durban, average goods value USD 38,000 a box, carrying cost 8% a year, and you currently hold safety stock against a transit variability of plus or minus three days. If Durban pushes realistic variability out to plus or minus fourteen, the extra cover you need is eleven days of consumption. Forty boxes a month is 1.32 boxes a day, so eleven days is roughly 14.5 extra boxes parked in the pipeline permanently, which is about USD 550,000 of working capital you did not previously need, or around USD 44,000 a year of carrying cost before a single box is late.

Add the detention exposure. Assume five days of the delay lands inside your free time rather than the carrier's, at USD 120 a day per forty-foot box: forty boxes, five days, USD 120, is USD 24,000 a month, and it arrives as an invoice three weeks later with no decision attached to it. Set that against the obvious alternative. Rerouting through Maputo for a Johannesburg delivery adds trucking cost and takes days off the clock; if the premium is USD 1,200 a box and it buys eight days, forty boxes costs USD 48,000 a month in order to avoid USD 24,000 of detention and a large slice of the inventory hit. That trade is close enough that it has to be run per commodity rather than per company: it wins on high-value time-sensitive lines and loses on heavy low-value cargo.

Air freight, for completeness. A forty-foot container carries something over 20 tonnes of realistic payload, and at USD 4.5 a kilo that is upwards of USD 100,000 against a sea leg costing a few thousand. A thirty-to-one premium is not a routing decision at all; it is a decision about the two or three stock-keeping units whose absence stops a line or loses a customer. Make that list now, while nobody is shouting.

And then the layer that never makes it into a budget line: the commercial fallout of a delay you did not cause. Most sales contracts fix a delivery date, and most force majeure clauses do not cover port congestion in terms broad enough to actually protect you. If your box lands fifteen days late and the contract treats late delivery as a breach, the argument about whose fault it was happens after the margin is gone. Two checks are worth running before the next shipment and neither costs anything. Look at whether your incoterm places risk at a point you can influence at all, and look at whether the congestion carve-out in your customer contract is written narrowly enough to be usable or vaguely enough to be refused. This is not legal advice, only an observation about arithmetic: when transit time swings by fifteen days, a delivery clause stops being boilerplate and becomes a live term.

The traps are predictable, and every one of them is a version of mistaking a floor for a forecast. The one that costs customer trust is telling a consignee their box will move in eleven days; it will not, and 11.3 was never a promise. The one that costs money is booking around Durban to a terminal you have never used without checking the inland leg, because a port change that takes days off the ocean and adds them back on the road is not a saving. The one that wastes a quarter is treating this as a Durban problem and shifting to Ngqura or Cape Town inside the same symptom: national weekly throughput rose 6% alongside Durban Gateway's 6%, which tells you the constraint travels with the network rather than sitting at one gate. And the quiet one, the one that hurts longest, is doing nothing for a month because the weekly number improved. Six percent a week is the sound of a system running flat out, not the sound of slack coming back.

Durban has been written off before and it will be written off again, usually by people reading a spreadsheet from another hemisphere. Ports are stubborn things. They have cranes, a city behind them, and a deepwater habit a century in the making. Networks are patient in a different way, and they remember. Every fortnight of 251-hour waits is a small instruction to the people who choose routes, and by the time it surfaces in an annual volume table it will look like a decision that was made all at once. It was not. It was made on weeks like this one.

↑ Back to the key points

— By Tomás Ferreira