Nationwide COSATU Protest Disrupts South African Port Operations

The Analysis

Spread the map out and run a finger down the south-east coast of Africa. Durban sits on that coast and Port Elizabeth sits further along it, and on 19 June both of them reported the same three things: access constraints, landside delays, terminal congestion. Transnet Port Terminals had put the warning out that a nationwide COSATU protest action would disrupt transport and logistics movements to and from port facilities. That is the factual base of this piece, and there is not much more to it than that, which is exactly why it is worth dwelling on. The interesting half of this story is the half that did not happen.

The half that did not happen is on the water. Waterside operations largely continued. Ships kept being worked, berths kept turning, and if you had been watching nothing but the vessel schedule you would have concluded that South Africa was having an entirely ordinary week. What broke was everything on the land side of the gate: the trucks that were not available, the roads that were disrupted, and the containers that as a result sat where they were for longer than usual. I have stood on enough terminals to know which kind of trouble a nationwide protest is, and it is never the kind you can see from the water.

That distinction is the whole reason I am writing this rather than filing it as one more congestion note. Most importers track the ocean leg well and the land leg badly, and almost all of them measure the wrong end of the port. If your visibility stops at discharge, a disruption like this one stays invisible to you until a truck does not arrive, and by then the queue in front of your box has already been built.

Two gates on one coastline reporting the same symptoms on the same day is not bad luck, and it is worth saying why. A country with a small number of container gates has concentrated its risk whether or not anyone ever decided to. Those gates share a road network, a truck market, a labour market, and, for a large part of the inland volume, the same corridors going up-country. A nationwide action does not pick a port. It lands on all of them at once, and the operational report from each one ends up reading like a copy of the others. When you are choosing how to enter a market shaped like this, the question is never which gate is better on an ordinary Tuesday. The question is whether the two gates you are counting on fail at the same time for the same reason.

Ten years ago this was a different place, and I do not mean that nostalgically. Ask me then what constrained Durban and I would have talked about berth capacity, crane productivity, the depth of the entrance channel, and the terminal operating system that was supposed to tie all of it together. Those were the investment stories of that decade and a great deal of concrete and steel went into them. The binding constraint has moved inland. It is the gate, the road behind it, the truck fleet available on a given morning, and the corridor that carries the box up-country. That is not a failure of the terminal. It is a sign that the terminal got better than the country around it, which happens more often than port people like to admit.

What has really changed is the network behind the port, and dwell time is where you can see it. Dwell is not a fee line to be argued about afterwards. It is an inventory line and a reliability line. Every extra day a box sits behind a gate is a day of goods you own and cannot sell, and, more expensively, a day of variability that somebody in your planning team has to cover with stock somewhere else. A port can improve its berth productivity every year for a decade and still hand you a worse door-to-door service, if the land side behind it is where the variability now lives.

The signature is worth memorising, because it tells you whether what you are looking at heals by itself. Tell me which side of the gate failed and I will tell you how long the tail is. A berth problem heals when the ship sails; the queue is on the water and it leaves with the vessels. A gate problem heals only when the road is open, the trucks exist and the shift roster is whole again, and even then the terminal has to work off everything that piled up while those things were broken. Container dwell extending while vessel operations largely continue is the cleanest signal you will ever be handed that the failure is on land. If you take one operational idea out of this piece, take that one, because it changes which number you watch.

Now the part that touches your door-to-door options. If your inland leg is built around one gate, that gate's dwell is your lead time, and the number you should be managing is not the ocean transit your carrier sells you. It is the ocean transit plus whatever happens between discharge and the moment your box is actually moving on a road you control. Two gates on one coastline give you redundancy only if the corridors behind them differ. If both gates feed the same road into the same inland market, you have changed the port and kept the risk, and the second gate is decoration. Ask your inland carrier to draw you the two routes before you call either of them a backup.

This is also why a change in a hub's berth count or its throughput so rarely shows up in your door-to-door options in the way people expect. A port that adds a berth and cannot add a road has given you a better port and the same supply chain. Berths are the part that gets built, photographed and announced; the road behind the gate is the part that takes fifteen years, several budgets and a set of decisions nobody in shipping controls. So when you read that a gateway has expanded, the question to ask is not how many boxes it can now handle at the quay. It is how many boxes a day can leave it, on how many corridors, when a quarter of the trucks do not turn up.

One more thing about the decade, because it is the part that quietly decides all of this. A berth is roughly a three-year object: design, tender, build, commission, and at the end of it there is something you can stand on and photograph. A road behind a gate is a fifteen-year object, and it is not one decision but a sequence of them, spread over several budgets, a set of land acquisitions and a political cycle that has no reason at all to finish on your schedule. A rail path is not an object in the first place; it is a negotiation between a terminal, an operator and a timetable that belongs to somebody else.

That is why a port can announce an expansion, complete it on time and still hand you the same door-to-door service three years later, and why the question to ask when you read the announcement is not what is being built at the quay. It is what is being built, or quietly not being built, on the land side of the same project, and whether the two sit in the same budget.

Long-run port selection in a market like this should be judged on things that are not in the port's own brochure. How many gates there are and whether they share a hinterland corridor, which you test by asking for the actual routing rather than by looking at a map. Whether your box has any way of leaving the gate that is not the road, and what your own contract gives you on that in writing, under stress, rather than what your agent hopes for. The distribution of dwell rather than the average, because an average hides precisely the tail that ruins your month; ask for the nine-out-of-ten figure and keep it yourself if nobody will hand it over.

How deep the truck market behind the gate is: how many operators, how much of the capacity is one vehicle and one driver, and what happens to your bookings when a quarter of it disappears for three days. And the labour calendar, meaning when the next wage round falls and what the last one cost the country in operating days.

On when to actually redraw the network, because waiting for a catastrophe is the most expensive version of this decision. By the time something has genuinely broken, every alternative is expensive and none of them has been tested. The rule I use is duller. Take the gate you depend on most and write down three answers, so that somebody else can argue with them. How long the constraint has been named in operational reports: one week is weather, one quarter is structure. How much of your volume sits behind a single inland corridor: past roughly half, the corridor is the decision and the port is only where the corridor starts.

And what the second gate costs you in transit days on an ordinary week, which is the price of the option before you need it. If the first answer is a quarter, the second is above half and the third is under two days, the second gate stops being a contingency and becomes part of the plan, and it should be carrying boxes before anything goes wrong.

There is a second reason to look at dwell rather than at anything on the water, and it is the one that changes how you plan a year. A national protest is not a weather event. Weather is random and it does not read the calendar; a wage round, a national action and the dates around them are announced, negotiated and known weeks in advance. That means a meaningful part of the variability you are absorbing at this gate is not random at all. It is seasonal in the way a public holiday is seasonal, and it should be in the same place in your planning calendar. Importers who treat South African dwell as noise end up buying stock against it forever, at 8% a year, to protect themselves against something they could have scheduled around. Importers who treat it as a calendar move a dozen shipments a year and hold the same service level with less stock.

One more thing about measuring it. Do not let anybody sell you a single dwell number. A terminal that reports an average is telling you about the box that behaved; the box that costs you money is the tenth one, and the only way to see it is a distribution. Keep the series yourself, because the series is yours in a way no report will ever be: your boxes, your gate-out timestamps, your own tail. It takes one column in a spreadsheet and about ten minutes a week, and it is the difference between arguing about whether a gateway has deteriorated and knowing that it has.

Signals, because a judgement you cannot trigger on is a mood. Dwell running above your own recent median for a run of consecutive working days is the base signal, and you should be keeping that series yourself from gate-out timestamps rather than trusting anybody else's. A disruption that is still touching your boxes two days after the event that caused it has ended is the second: it means the failure was on land and the queue is real. A third is the shape of the failure itself. If vessel operations are described as continuing, then everything wrong with your shipment is on land, and every lever you have is on land too, which rules out half the phone calls people normally make. A fourth is inside your own inventory: if one event eats more than a third of the stock cover you hold against transit variability, the cover is wrong, not the event.

Here is the arithmetic, with every number exposed so you can replace mine with yours. Suppose you import 30 forty-foot containers a month through Durban to an inland distribution centre. Suppose your normal container dwell, counted from discharge to gate-out, is four days, and suppose this event stretches it to nine. Five days is my assumption; the report says dwell times were extended, it does not say by how much, so put your own observed number in and keep the method. Thirty boxes a month is a flow of one box a day. Five extra days on a one-box-a-day flow means roughly five extra boxes sitting behind the gate at any moment once the queue has built.

Suppose each box carries 40,000 of goods and your cost of capital is 8% a year. Five boxes is 200,000 of goods you own and cannot sell; 8% of that is 16,000 a year if the delay were permanent, and this one is not, so scale it: a ten-day event costs you 16,000 times ten over 365, which is about 440. That is the carrying cost of the delay, and it is small enough that you should stop worrying about it and start worrying about the number underneath it.

The number underneath it is the buffer. Suppose your distribution centre holds ten days of stock against transit variability. A five-day dwell extension eats half of that in one event. Suppose something like this happens twice in a quarter: your cover is gone, and the cost of what follows is priced at your gross margin, not at 8%. So the arithmetic you should actually run is not the carrying cost of five days. It is the cover you have to add to hold the same service level. One extra day of cover on a one-box-a-day flow is one box, 40,000 of working capital, plus roughly 3,200 a year to carry it, forever.

Two days and you have spent 80,000 and 6,400 a year to protect yourself against a protest. That may well be the right answer; I am not going to pretend it never is. But put it beside the alternative before you sign it off. Qualifying a second gate costs a project, an inland tender and a few weeks of somebody's time, and it buys you a corridor that a nationwide action may not close on the same day. One of those two options expires. The other one does not.

Alternatives, and what each of them actually costs. Moving volume to Port Elizabeth is the obvious first move, and it is only real if the inland corridor differs; if the two gates converge onto the same road for most of the distance to your distribution centre, you have bought a different port and the same risk, and the honest thing is to stop the project there. Any way of getting the box off the port that is not the road is worth asking about now rather than during the event, and the question to ask is what your contract gives you in writing, not what your agent hopes for when the gates are quiet.

Pulling shipments forward by a week before a date you already know about costs inventory and buys certainty, and for anything with a production date or a shelf life attached it is usually the cheapest lever on the table. Air freight covers the slice that would otherwise stop a line, and only that slice, which means deciding in advance which boxes those are rather than discovering it in the moment. Changing where you buy from is a decade-scale answer to a one-week problem, and the reason to open that file now is that the file takes years, not that this particular protest justifies it.

There is one alternative I would start this quarter regardless of how the next few weeks go, and it is dull: run a slice of your volume through the second gate on purpose. Ten percent for one quarter is enough. You are not buying a backup, you are buying data, because the only way to know what the second corridor really does to your lead time is to put boxes on it and timestamp them. Most importers never do this, then discover during a disruption that their second gate is a name on a map with an inland leg nobody has ever quoted.

The traps, and they are almost all variations on measuring the wrong thing. Setting your milestone at discharge rather than gate-out means a national protest can unfold entirely underneath your visibility. Working from average dwell rather than the tail means your plan is built on the median box while your pain comes from the tenth one. Treating your contracted collection capacity as a commitment when trucks are short means you will learn what it really is on the worst morning of the quarter. Quoting free time from memory rather than from the contract means the clock you are planning against is not the clock that is running. And planning for the event rather than for the queue behind it means you will reopen the plan two days early, find the terminal still working off what piled up, and conclude that the disruption was worse than it was.

Put the map away. South Africa is a good place to think about all of this because the geography is honest: two gates, one coastline, and a hinterland that has to be reached by land whichever gate you choose. Ten years ago the constraint here sat on the water and you could pour concrete at it. Now it sits on the land, and the things that would fix it are roads, trucks, shifts and settlements, none of which a terminal operator controls and none of which arrive quickly. What has really changed is the network. The port is only the part of it you can see.

↑ Back to the key points

— By Tomás Ferreira