Nobody sells a single point of failure. Every proposal that crosses your desk will describe a resilient, diverse, redundant network, and the diagram will have two lines on it. The question that matters is where those two lines stop being two lines — because somewhere they usually do, and the place they converge is the thing you are actually buying.

This is not a criticism of providers. In a market where most capacity is bought and resold several times, a supplier often does not know where the convergence is either. That is precisely why the questions below are worth asking in writing.

1. Name the two systems

Ask which submarine cable system carries the primary service and which carries the protection, by name. Not “our international capacity” — the system name.

If both answers are the same system, you have capacity redundancy on one cable, which protects you against an equipment failure and not against the event everyone actually worries about. If the answers are two different systems, the next four questions still apply.

2. Where do they come ashore?

Two cables that land at the same station share that station’s power, its terminal equipment, its building and its staff. That is a meaningful reduction in risk compared with a single cable — a cable break on one is genuinely independent — but it is not the full separation a diagram implies.

In Somalia this is checkable. The National Communications Authority publishes a register of cable systems, landing cities and landing station operators. Cross-check what you are told against it.

3. Where does the backhaul diverge?

This is the question that catches the most people. Two cables, two landing stations, and then both services ride the same terrestrial route to your building — because there is only one duct into that part of the city, or because both providers lease from the same third party.

The honest answer to this is sometimes “it does not diverge, here is where it converges, and here is what that means.” That answer is far more useful than a diagram that implies otherwise, and it lets you decide whether to accept the risk, pay for a second route, or put a satellite path in as the genuinely independent leg.

4. Direct or through a party?

Capacity can be held directly on a system, or reached through a shareholding or a wholesale relationship with someone who holds it. Both deliver real capacity on real cables. They are not the same instrument in a crisis, because only one of them puts your supplier in the room where restoration is prioritised.

We publish our own split for this reason: of the four systems we reach, EASSy and 2Africa land at our own station in Mogadishu, and WACS and EIG are reached through our shareholding in WIOCC. A customer designing against a specific failure is entitled to know which of those is carrying them.

5. What actually happens in the first hour?

Not the SLA — the sequence. Who detects it, what reroutes automatically, what requires a human, which humans, and where they are. An SLA is a commercial instrument that describes what you get paid if things go badly. It does not describe what happens, and the credit rarely covers the cost of the outage it compensates.

A provider who can walk you through the first hour concretely has thought about it. One who moves the conversation to the service credit has answered a different question.

What good looks like

A design that survives these questions usually has three properties. The two paths separate at the cable, at the landing station and on the terrestrial leg — or, where they cannot, the convergence is documented and agreed rather than discovered. At least one path is genuinely independent in kind, which in practice often means satellite, because a second fibre in the same trench is a second fibre in the same trench. And the party you pay controls enough of the path to act rather than escalate.

None of that is exotic. It is just uncomfortable to specify, which is why it usually is not.