Who this is for
Engineers and operations managers planning an IoT rollout across South African sites — factories, farms, water networks, mines, fleets — who need to choose a connectivity strategy before committing to hardware, and don't want to discover the coverage gap after installation.
Why this is harder in South Africa
Two local realities dominate the decision. First, coverage is uneven: cell networks are strong in metros and along corridors, then thin out fast on a farm or at a remote pump station. Second, load shedding means any link that depends on grid-powered infrastructure — including the cell tower itself — can go dark, so resilience and local buffering matter as much as raw throughput.
The four contenders, compared
| Network | Best for | Range / coverage | Power | Data per device | Watch-out |
|---|---|---|---|---|---|
| 4G / LTE | Few devices, anywhere with cell signal; gateways; video/rich data | National carrier coverage | Higher — mains or solar | High | SIM/data cost per device; tower depends on grid |
| LoRaWAN | Many low-data sensors over a site you control | Several km from your own gateway | Very low — years on a battery | Low | You run the gateway; not for rich data |
| NB-IoT | Low-data sensors where carrier has deployed it | Carrier-dependent, patchy | Low | Low–medium | Coverage varies sharply by area |
| Sigfox | Very simple, very low-data, low-cost devices | Operator network footprint | Very low | Very low (tiny payloads) | Single-operator lock-in; minimal data |
The technical parameters, side by side
The summary table above decides the shortlist; this one settles the arguments a shortlist produces:
| 4G / LTE | LoRaWAN | NB-IoT | Sigfox | |
|---|---|---|---|---|
| Frequency in SA | Licensed carrier bands | 868 MHz ISM (licence-free) | Licensed — Vodacom Band 8 (900 MHz) | 868 MHz ISM (operator network) |
| Typical range | Per tower footprint | 2–5 km urban, 10–15 km rural line-of-sight | Per tower, with deep indoor/underground margin | Tens of km to operator base stations |
| Battery life (sensor duty) | Days–weeks; mains/solar in practice | 5–10 years typical | Multi-year (PSM/eDRX dependent) | Multi-year |
| Payload per message | Effectively unlimited | ~51–222 bytes (data-rate dependent) | ~1,000+ bytes practical | 12 bytes up, 8 bytes down, ≤140 msgs/day |
| Latency / downlink | Real-time, full duplex | Seconds; downlink limited by class | Seconds; supports firmware-over-air | Very limited downlink |
| Ongoing cost shape | SIM/data per device | No per-device fee; you power the gateway | Low per-device SIM | Low per-device subscription |
| Who owns coverage | Carrier | You (or a public network) | Carrier | Single national operator |
| SA status (Sep 2026) | Universal; 5G growing | Public + private networks mature | Vodacom commercial only; MTN testing | Single-operator network (Sigfox South Africa, since 2022) |
What changed on the South African networks (2026 status)
- Sigfox still exists, under a different operator. SqwidNet was wound down in 2022 and the network passed to Sigfox South Africa, which continues to run it. The technology constraints are unchanged — 12-byte payloads, minimal downlink, one operator's footprint — and they are why, for the fixed-sensor problems Sigfox targets, we deploy LoRaWAN (coverage you own) with 4G backhaul instead.
- NB-IoT remains a one-carrier story. Vodacom is the only commercial NB-IoT network (Band 8, 900 MHz), with most of its LTE sites targeted for enablement; MTN is still in testing. LTE-M has no committed SA deployment — specify it only for hardware that will roam elsewhere in the world.
- The 2G/3G sunset is now a design constraint, not a rumour. Government policy stopped type-approval of 2G/3G-only devices in 2024 and barred new activations from the end of 2024; the shutdown itself, originally gazetted for end-2027, is now left to each operator's timing. The practical rule for anyone buying trackers or telemetry hardware today: nothing new goes on 2G or 3G — the cheap GSM module that still works this year is a stranded asset on its way to happening.
Which network for which site — the scenario map
| Scenario | First choice | Why | Backhaul / fallback |
|---|---|---|---|
| Factory floor (counts, currents, temperatures) | LoRaWAN or wired to edge gateway | Dense sensors, no per-point SIM cost, buffering at the edge | 4G from the gateway |
| Remote borehole / pump station | LoRaWAN to nearest gateway; 4G node if solitary | Battery-for-years where no power; solar + 4G where standalone | Store-and-forward buffering |
| Municipal DMA / bulk water metering | LoRaWAN (own gateways on reservoirs) | Hundreds of points, pulse/level payloads are tiny | 4G backhaul per gateway |
| Farm — soil, tanks, gates, weather | LoRaWAN from the homestead | One gateway covers kilometres; devices live on AA cells | 4G or fixed wireless uplink |
| Mine surface infrastructure | LoRaWAN + 4G hybrid | Wide area you control; rich data (cameras, OEE) rides LTE | Private LTE on larger operations |
| Underground workings | Site network (leaky feeder / private LTE / Wi-Fi mesh) | No public network reaches underground; design is site-specific | Surface backhaul |
| Cold-chain vehicles & trailers | 4G | Mobility rules out your own gateways; loggers buffer between signal | Store-and-forward in the logger |
| Generator / plant-room monitoring in buildings | 4G (NB-IoT is the niche alternative where verified) | Deep-indoor reach decides it; buffering covers the gaps | LoRaWAN if the estate runs one |
| Estate / campus utilities | Private LoRaWAN | One network serves water, power, access and level sensing | 4G backhaul |
| Scattered national retail sites (few sensors each) | 4G per site | No site justifies a gateway; carrier coverage does the work | SIM failover between carriers |
Why 4G is so often the practical answer
For many real projects — a dozen sensors at a plant, a few water points, an edge gateway pushing dashboards — 4G simply removes a problem: you don't build or maintain any network infrastructure, and it works the moment there's signal. The trade-off is per-device data cost and power draw, which is why 4G suits a smaller number of richer devices or a gateway that aggregates many cheap sensors behind it. ("4G" is itself a family — Cat-4 for gateways, Cat-1/Cat-1bis for telemetry nodes — unpacked in our LTE category guide.) In a country where the priority is usually "get it working reliably without a site-wide network project," that pragmatism wins a lot of the time.
When LoRaWAN is clearly better
The moment you have many low-data sensors spread across an area you control — a farm, a mine, a municipal reservoir network — LoRaWAN's economics dominate. One gateway covers kilometres, devices run for years on a battery, and there's no per-device SIM cost. The catch is that you own and power the gateway, and the link isn't for rich data. For soil moisture, tank levels, flow pulses and the like, that's exactly the right shape. For a deeper look at bands, device classes, gateways and security, see our dedicated LoRaWAN in South Africa guide.
The hybrid that usually wins: LoRaWAN (or wired) for the many cheap sensors on-site, aggregated to an edge gateway, with 4G as the backhaul to the cloud — and local buffering so a tower or grid outage doesn't lose your data. You rarely pick one network; you layer them.
The decision, in five questions
- Coverage: Is there reliable cell signal at every device location? If not, you need your own gateway (LoRaWAN) or a different site plan.
- Device count: A handful → cellular per device. Dozens-to-hundreds → LoRaWAN behind a gateway.
- Data per device: Rich (video, high-rate) → 4G. Small periodic readings → LPWAN.
- Power: Mains/solar available → cellular is fine. Battery-for-years required → LoRaWAN/NB-IoT/Sigfox.
- Resilience: Whatever you choose, buffer at the edge so load shedding and tower outages don't lose data.
This is precisely the assessment we run at the start of every addaNet project — connectivity chosen per site, not per brochure — with LoRaWAN and 4G as the stack we deploy. It's especially decisive for remote farms, water networks and mines.