There is one sentence I hear often when I talk about outages: "it's not a big deal, it comes back." It is true for someone checking their messages. It is false for someone whose income runs through the connection.
I belong to the second group. My deliverables go out over the internet. My clients sign off over the internet. Part of what I collect passes through services that need the network to confirm a transaction. The day it stops, I am not inconvenienced: I am at a standstill.
The confusion that costs dearly
We have got into the habit of measuring the internet by the price of its subscription. That is the wrong unit.
The right calculation is the cost of one hour of downtime. Take your monthly income, divide it by the number of hours you actually work in the month. You get what one hour is worth to you. Multiply that by the hours you could not work last month. The figure that comes out usually has nothing to do with the amount on your bill — it is far bigger.
It is a calculation most freelancers and small outfits never make. As long as we don't make it, an outage stays an annoyance. Once we have made it, it becomes a line of loss, and we start treating it as one.
What the context tells us
The country is not disconnected, far from it. Mobile penetration is above 70% of the population, 4G is rolling out in the main cities, and Togo is connected to the ACE submarine cable. Mobile money reaches 57.2% of the population, and more than 80 institutions are now linked to the BCEAO's interoperable instant payment platform.
That is a real base, and it is growing. But it creates a new dependency: the more instant the payment becomes, the more it hangs on the network. A trader who took cash could carry on without a connection. That same trader, once switched to merchant payment, stops when the network stops.
The World Bank, in its 2026 report, ranks connectivity among the structural obstacles to digital development, alongside the cost and stability of electricity. The figure it gives for sub-Saharan Africa says it all: more than two thirds of rural schools have no reliable internet.
In other words: we are building uses that assume a permanent connection, on an infrastructure that is not permanent everywhere yet. That is not a reason to back off. It is a reason to design differently.
Working as if the connection is about to drop
Here is what I have put in place, after paying to learn it.
Two connections, two networks. A main subscription and a second connection with a different operator, on a different technology. Not two plans from the same provider: when its network goes down, both go down together. The second connection does not need to be comfortable, it just has to let me send a file and warn a client.
A way of working that survives offline. I work on local files that sync when the connection comes back, never straight inside a tool that has to be online to save. It is a choice of tools, made once, that changes everything else. If your word processor, your code or your accounting stops without a network, change tools before the next outage, not during it.
Big downloads ahead of time. Updates, source images, libraries: anything heavy gets pulled down while things work, not when it is needed. I have lost hours waiting for a file I could have had the day before.
A written client protocol. When I am cut off, my client hears it from me, not from my silence. A short message, sent over the mobile network: what is happening, when I resume, what it pushes back. Silence turns a technical incident into a trust problem. One sentence keeps it an incident.
Deadlines that build in reality. I add a margin to my commitments. Not a comfort margin — a margin calculated on the downtime hours I actually measured the previous quarter. Promising deadlines that assume a perfect connection is promising what you do not control.
Measure before complaining
The starting point for all of this fits on one line in a notebook: the date, the time of the outage, the time it came back. Nothing else.
After a month, you know what you are putting up with. After a quarter, you can put a figure on it, build it into your prices and show it to a client who finds your deadline long. As long as it stays an impression, it is endured. Once it is a number, it is managed.
Nobody is going to hand me a perfect connection because I deserve one. While I wait for it to improve — and it is improving — I can at least decide that my business does not stop exactly when it stops.
A business that depends on the internet with no plan for its outages is not selling a service: it is selling an expectation of service.