Your signal dies on a Tuesday night. There’s no message from your provider, no cause, no estimate of when it will come back. You restart the router, check the app, nothing. In the morning you call the contact centre, wait, and come away with a reference number and a callback that never arrives. Two days later the service returns as quietly as it left.
Here is what the industry has not absorbed: that week, not the outage inside it, decided whether you stay. Every network fails; that is the physics of running infrastructure across a big, unevenly powered country, and no operator escapes it. So, failure cannot separate a good provider from a bad one. The silence around it can, and silence is cheap to end. So, why does almost no operator end it?
The answer is not that executives haven’t noticed; they have read the same decks as everyone else. It is structural. Operators go silent during faults not because someone chooses silence but because they are built in a way that makes speaking hard. Three features do most of the work, and none is fixed by good intentions.
The first is the contact centre. The industry decided long ago that customer contact was a cost to minimise, not a capability to own, and outsourced it. Outsourced centres run to the contract’s metrics, and those reward handling time, cost-per-contact and calls closed, not problems solved. So, the agent logs the fault, issues a reference number, closes the ticket, takes the next call. The customer feels processed, not helped. The agent is not failing but is hitting the wrong target, and you cannot train your way out of a measurement problem.
The second is the org chart. The people who know a tower is down and the people who talk to customers sit in different parts of the company, under different bosses, on different systems. Network operations is measured on uptime; customer communications, under marketing or care, on campaigns and sentiment. Faults fall into the seam between them. Nobody’s job is to tell customers what broke and when it returns. Split an outcome across two silos and it belongs to neither. Silence is the default.
The plumbing
The third is the plumbing. Even an operator that wants to send proactive updates often can’t: the systems that detect faults were never wired to the systems that reach customers. Fault-management platforms alert technicians and track repairs; they were not built to message every affected customer. Wiring the two competes for budget against projects with clearer revenue. So, the ability to communicate at scale during an outage does not exist in most operators, and nobody notices until the outage hits.
Put those together and the silence stops looking like negligence and starts looking like an emergent property of the architecture: a contact centre built for cost, no owner for the customer’s outage and no path from “a fault occurred” to “the customer was told”. That machine produces silence reliably, however sincerely its leaders say they care.
Read: The MVNO trap deepens as the battle moves to data
There is a way to test this. If complaints tracked network quality, the operator with the most negative network conversation would have the worst measured network. It has the best. In Opensignal’s 2025 South African report, built on billions of handset measurements, MTN won 11 of 15 awards, including every overall experience award; Rain managed only a shared voice category. Ookla puts MTN’s median mobile download at 74.76Mbit/s, the fastest in the country, and Icasa records national 4G population coverage of 99.5%. This is not a country where the network is the bottleneck, and MTN’s lead holds across benchmarks and years.
Now set that against how customers feel. DataEQ’s operational customer experience score runs almost exactly counter to the rankings. Rain, dead last on the network, records the highest customer sentiment in the industry at +16%; MTN, first on the network, sits at -17%. The best-measured network trails the worst on how its customers feel. Cell C is positive, too, at +11%, and there the comparison is cleanest: Cell C owns no network, it roams on MTN’s and Vodacom’s, so its customers sit largely on MTN’s own towers. Same infrastructure, opposite sign. When the same towers flip the score depending on the name on the bill, the score is not measuring the towers.
Read: The fragile joint in the Capitec machine
Read the gap carefully, because the two datasets do not measure the same thing. The benchmarks gauge speed and coverage from handsets. DataEQ measures what people post, which folds in billing, pricing, fixed-line service and the ordeal of getting a fault fixed. So, sentiment is not measuring the network; it is measuring everything around it. And the report says which parts do the damage: network coverage sits at -27%, bad but survivable, while customer service is -85%, account admin -80% and the call centre, the worst channel in the industry, -84%, with only 39% of urgent messages getting any reply. The network is not what sinks satisfaction. The service wrapped around it is.
One caveat: Cell C’s score appears to have been lifted by a coverage campaign – narrative control rather than service. The cleaner signal is Rain, whose score rested on agents resolving problems the same day. What customers reward is the experience around the network: being answered, being told and having their problem resolved.
The fix is cheaper than it sounds. If the machine produces silence, you change the machine, not the messaging: one owner for fault communication with authority across operations and care; a contact function measured on resolution, not throughput; the fault systems wired to the customer-facing ones so a proactive update is automatic, not heroic. None of that needs new spectrum or a rebuilt network. It needs treating communication during a failure as part of the product. The operators still treating it as an afterthought are not short of goodwill. They are short of an owner, an incentive and a wire. – © 2026 NewsCentral Media
