A fixed-wireless network is hundreds of radios on roofs and towers, all talking to each other through the air, in the sun, wind and rain. In Gqeberha and Durban we run about 100 sector radios on our towers, plus a radio on every wireless customer's roof, and we manage about 300 customer Wi-Fi routers remotely. Any one of them can have a bad day.
For years the main question our monitoring asked was simple: is it on or off? That catches the big failures. But it misses the slow ones: the link that's still up but getting worse, the radio that's working on half its strength, the customer whose connection is technically fine but feels terrible every evening.
So during October and November 2026 we rebuilt how we watch the network, around one idea: find the problem before the customer has to tell us. This is how it works today.
Layer 1: the sectors, every hour
A sector radio on a tower serves a slice of the area around it. The first check runs every hour and looks at our sector radios as a whole.
The most important thing it catches is a sector that goes quiet. If a sector that had three or more customers connected suddenly has none, that's not normal, even if the radio itself still looks fine. Since 18 Oct 2026 that raises an alert straight away, and a second message when the customers are back.
The check also knows what a power cut looks like. When three or more sectors go off in the same hour, the alert says to check the municipality's power notices first. Several sectors dropping together is usually power in an area, not several separate faults, and treating it that way sends the right people to the right problem.
Layer 2: every link, every hour
On 6 Nov 2026 we added a second hourly check that looks much deeper: at every radio link, every chain inside each radio, every network cable's speed and the power supplies of our network routers.
It looks for faults that don't switch anything off, for example:
- A radio working on one of its two antenna chains. It still connects, but with less margin, and it can fail completely later.
- A signal that has dropped well below its own normal. We compare each link with its own history, not with a fixed number, because every roof is different. A drop of 6 dB or more from its usual level gets flagged.
- Links that hear one way better than the other, a sign of interference or a misaligned radio.
- Lots of resent data (retries), which makes a link feel slow even when its signal looks fine.
- A network port running at 100 Mbps instead of gigabit, usually a damaged cable or connector, quietly capping everything behind it.
- Unexpected restarts, power supply warnings, high temperatures and ports that keep going up and down.
Every new fault, every fault that lasts three days and seven days, and every fault that clears goes into our findings log, so nothing sits forgotten. A fault that needs someone on site becomes a job with an engineer's name and a date on it.
There's one more rule we hold ourselves to: when someone finds a new kind of fault by hand, it becomes a new automatic check, with a test. The same problem should never have to be found by hand twice.
Layer 3: every customer, every five minutes
Radios and links are only half the story. What matters is how the internet feels in your house. Since 8 Nov 2026 we work out an experience score every five minutes for every connected customer on our Gqeberha network.
Each score starts at 100 and loses points for the things that make the internet feel bad:
| What we measure | What it tells us |
|---|---|
| Delay (latency) to your router | Slow-feeling browsing, laggy calls and games |
| Lost packets | Video calls that freeze, downloads that stall |
| Jitter (how much the delay jumps around) | Choppy voice and video |
| Your radio's link speed against what it should be | A weak or poorly aligned link |
| Signal strength | How much margin your link has |
| How busy your sector's airtime is | Evening congestion on the tower |
| Data dropped in your speed queue | Your plan's limit being hit hard |
| Your connection dropping and reconnecting | Power, cabling or router trouble |
| Delay inside your home | A Wi-Fi or router problem rather than the line |
We measure the delay to both your roof radio and your router. That one comparison tells us a lot: if the radio answers quickly and the router doesn't, the trouble is inside the house, not on the air.
Two rules keep the score fair. First, anything we can't measure is shown as "not measured", never counted against the link. Some routers simply don't answer pings, and that isn't an outage. Second, one bad reading is noise. A customer is flagged as "likely having a bad experience" only when their score is under 60 in three of the last four rounds. Scores of 80 and up count as good.
The result is a short, ranked list for our team: the customers most likely to be unhappy right now, with the reason next to each name. Often the fix is on our side (a sector that needs tuning, a radio that should move to a quieter tower) and we can start before anyone picks up the phone.
Layer 4: knowing what normal looks like for you
Some people are online until midnight; some switch everything off at 21:00. An offline alarm that ignores that is just noise. So our operations screen learns each customer's own pattern from four weeks of history, and only flags someone as offline when it's unusual for them. It also rolls customers on the same sector together: if three or more drop within a few minutes, that's one line on the screen, one problem to solve, not a wall of separate alarms.
What this means for you
- We often know before you do. Many faults are spotted and assigned before the first customer calls.
- Slow problems get caught. A link that's fading gets attention while it still works.
- Your report still matters. Monitoring sees the network; you see your home. When you tell us what's happening, with the lights on your equipment and when it started, we can match it with what we see. Our guide Is it a power cut or a fault? shows what to check.
These checks are plain rules, written down and tested, and they keep growing every time we learn something new. The network is monitored around the clock; read more about it on our network page.