Supply Chain Action Points
Read this first — the conclusion, and the moves to make:
- By 15 September 2026, qualify one alternative corridor end to end, covering rate, transit, border and haulier, and move at least 15% of your South Africa volume onto it before the berth queue clears.
- For every open booking through Durban, collect the carrier's arrival-to-alongside estimate in writing each week, and write your own delivery promise against the twenty-day figure rather than the berth figure.
- Set a formal trigger: if the published berth wait is still above five days at the eight-week mark, move a second 15% tranche, because the evacuation tail runs weeks behind the berth.
- Book inland truck and rail slots for the discharge week plus three rather than the discharge week, and confirm them 14 days out; the gate is where the second queue forms.
Vessel waiting times at Durban's Gateway Terminal have stretched to 8-12 days at berth following its NAVIS N4 terminal-operating-system migration, with the South African Association of Freight Forwarders escalating the 'operational crisis' to the presidency. Productivity has fallen to under one-third of normal, compounded by Red Sea diversions, while some carriers warn of waits up to 20 days.
The Analysis
Durban's Gateway Terminal has been running waits of eight to twelve days at berth since it switched to NAVIS N4, a new terminal operating system. Productivity is below one-third of normal. The South African Association of Freight Forwarders has escalated what it calls an operational crisis to the presidency, and some carriers are telling customers to plan for waits of up to twenty days, compounded by Red Sea diversions.
Two things in that paragraph deserve a long look, and neither of them is the eight to twelve days. One is that a software migration turned into a capacity outage, which means the port had no parallel path to run on while it changed its own plumbing. The other is that the trade is quoting two numbers eight days apart and treating them as if they described the same thing.
Durban is not an unusual port. It is an unusually honest one, because the failure is large enough to be measured and loud enough to reach a presidency.
A gateway port is a funnel. Everything bound for a hinterland goes through one neck, and a funnel is efficient right up until the neck closes. On a map that looks like a virtue. In a network it is a single point of failure with a national economy standing behind it, and it survives only because the alternatives cost more than the queue does. The alternatives here are real: Cape Town on the same coast, Coega to the east, Maputo and Walvis Bay across borders. Each of them charges a price that has nothing to do with berth productivity. A border crossing, another terminal's own backlog, and a set of hauliers who have never carried your cargo and have no particular reason to prioritise it next week.
Here is the queue arithmetic, with my assumption marked. A terminal running at about 80% normal utilisation has a spare service rate of about 20%. Knock capacity down to a third and the backlog builds at the gap between arrivals and that third, which at 80% utilisation is close to half of normal capacity per unit of time. Restore capacity and the backlog drains at the 20% spare rate, not at the rate it built. Do the division and clearing takes roughly 2.3 times as long as the build-up lasted. If the disruption window is three weeks, the berth tail runs past seven. The 80% is my assumption and it is not in the report; the ratio is the point, and it does not need my number to be exact in order to be right in direction. Queues are never symmetric. They fill faster than they empty, always, because filling happens at capacity minus a third and emptying happens at capacity minus demand.
Eight to twelve and twenty are not a contradiction and they are not a range. They are two rulers. Eight to twelve is a berth metric, measured from arrival to alongside. Twenty is a booking metric, which is what a carrier puts in front of a customer who has to promise a delivery date, and it contains the anchorage time before a berth window is even allocated plus the buffer the carrier adds because it would rather under-promise. You will only ever experience the second number directly. Your customer will only ever experience the third, which is whatever you chose to tell them. When two numbers sit eight days apart, do not average them. Ask which one your contract is written against, because a service commitment written against a berth metric is not measuring the thing that will make you late.
The Red Sea deserves a fair hearing before it gets dismissed as an excuse, and then it needs to be put back in its box. A diversion changes which ships arrive, in what order, and how full they are. It cannot triple the time it takes to move a box across a quay. Productivity under one-third is a fact measured inside the gate, at the crane and in the yard, and it is an organisational fact before it is anything else. The two multiply, and more boxes arriving into a slower terminal is far worse than either on its own. But they do not explain each other, and the distinction decides what you do next. A software and process recovery can complete in weeks. A routing change around the Cape will not revert because a rollback finished.
Who feels this first is not who complains loudest. The exporter with a fixed vessel cut-off and a penalty clause in a supply contract feels it first and hardest, because a missed cut-off is not a delay, it is a breach. The importer whose cargo is already on the water feels it at discharge, where the options are narrower than they look from an office. The party with the most freedom is the one who has not booked yet, and that party almost always waits, because waiting is free right up until it is not. Scale splits the impact in the usual way. A buyer with a monthly programme and a standing allocation gets re-accommodated. A buyer with thirty containers and a shipment date gets a phone call.
What the reporting has not said is that the constraint does not stay at the berth. While the berth queue drains, boxes are being discharged faster than the hinterland can take them away. The yard fills, the storage capacity that was designed for normal throughput is consumed by abnormal dwell, and two to four weeks after the berth wait starts falling the binding constraint moves to the gate: truck slots, rail slots, and the storage clock that starts when a box hits the ground whether or not anyone can collect it that day. A shipper who plans to re-enter when the queue clears arrives exactly at the moment the queue has moved rather than gone. The planning horizon is the berth wait plus the evacuation tail, not the berth wait. On my arithmetic above that is a berth tail near seven weeks followed by several weeks of gate congestion, and the second number is the one nobody publishes.
Step back to the decade and a more interesting risk appears. Corridors are habits, and habits are harder to move than boxes. A shipper who spends two quarters routing through Maputo or Walvis Bay builds a customs rhythm, a haulier relationship, a different set of people who answer the phone at seven in the morning. When Durban recovers, that habit does not automatically revert, because re-qualifying a corridor costs money and nobody inside a large company gets rewarded for switching back. That is how a gateway loses share permanently to an event that lasted a quarter. Ports that have lived through it know the pattern, which is one reason this escalation went to a presidency rather than to a customer service desk. The queue is the visible problem. The corridor is the asset.
The counterfactual is uncomfortable because it is so cheap in hindsight. Had the cutover been staged, with parallel running and a controlled volume ramp, the productivity line would have been a curve rather than a cliff, and the queue would have stayed inside the berth window instead of outside it. The same logic applies on the shipper's side. If 20% of volume had been dual-routed before the switch, you would have a warm alternative corridor today instead of building one under pressure, in the worst week, alongside everyone else doing the same thing to the same neighbouring terminals. Capacity discovered during a crisis is not capacity. It is a queue in a different postcode.
A small account with assumptions marked. Take an importer moving 60 FEU a month into a hinterland market through this gateway. Assume, and this is my assumption rather than a figure from the report, that the berth wait adds ten days to vessel arrival and that half the book misses its inland slot by one week. That is 30 FEU pushed a week, or 210 FEU-days a month of extra float, before anyone counts the storage clock on boxes sitting in a yard that cannot evacuate them. Multiply by your own carrying cost per FEU-day. I will not put a number there, because the number that matters is in your books and mine would only be a guess dressed as arithmetic. The shape is what matters. The delay is a stock, not a rate. It shows up as inventory and as expiring commitments, which is why finance feels it a month after operations does.
Maps do not change because a port gets faster. They change when a shipper decides a corridor is no longer worth the risk, and then they stay changed for a decade. Durban's terminal will be fixed, because terminals almost always are and software eventually does what software does. The question worth asking this month is not when. It is whether the cargo that leaves during the wait ever comes back, and what you will have standing when it does.
↑ Back to the key points