LoRaWAN vs NB-IoT vs Sigfox: Choosing an LPWAN

Choosing a radio is the decision on an industrial IoT project that is most expensive to reverse. Sensors can be swapped, dashboards rebuilt, cloud providers migrated. But the radio determines your gateway estate, your civils, your power budget and your recurring cost — and once a few hundred devices are in the field, changing it means touching every one of them.

This is a practical comparison of the three low-power wide-area networks that come up most often. Not a feature matrix: the question that actually matters is which constraints you can live with for the next five years.

What LPWAN is for

Low-power wide-area networks solve one specific problem: getting small amounts of data from many battery-powered devices over long distances, cheaply. They trade bandwidth for range and power. A LoRaWAN sensor might send 20 bytes every 15 minutes and run for years on a primary cell.

That trade is the whole point, and it is also the most common source of disappointment. If your requirement involves images, audio, firmware updates over the air, or anything approaching real-time control, LPWAN is the wrong family and no amount of tuning will fix it. Use LTE-M, Wi-Fi, or wired.

LoRaWAN

LoRaWAN operates in unlicensed sub-GHz ISM spectrum — 868 MHz in Europe. Its defining characteristic is that you can own the whole stack. You buy gateways, you run a network server, and there is no per-device subscription to anybody.

Where it wins

• Sites you control. A quarry, a plant, a farm, a campus. Put a gateway on the highest structure and you often cover the whole site with one or two units.
• No recurring connectivity cost. This dominates the economics at scale. A thousand sensors on your own network cost nothing per message.
• Private data path. Telemetry never leaves your infrastructure if you do not want it to — which matters more than people expect when the data describes production rates.
• Mature, competitive hardware market. Sensors from many vendors, and it is realistic to build your own.

Where it hurts

You own the network, which means you own its problems. Coverage gaps are yours to fix. Gateway backhaul, power and mounting are real civils work that projects routinely under-budget.

The bigger constraint is regulatory: in the EU, sub-GHz ISM bands carry duty cycle limits — commonly 1% on the bands most devices use. That caps how often a device may transmit, and it caps downlink from the gateway even harder. Any design that assumes frequent acknowledgements or regular downlink commands will hit this wall. Plan uplink-mostly.

NB-IoT

NB-IoT runs in licensed cellular spectrum, operated by mobile networks. You buy a SIM and a subscription, and coverage is somebody else’s problem.

Where it wins

• Geographically scattered assets. Anything spread across a region — utility metering, remote pumping stations, distributed infrastructure — where deploying your own gateways would be absurd.
• No infrastructure to own. No gateways, no mounting, no backhaul, no network server.
• Good building penetration. NB-IoT was designed for deep-indoor and basement scenarios, which is often exactly where meters live.
• Carrier-grade operations. Someone else is on call for the radio network.

Where it hurts

Per-device subscription costs never stop, and at thousands of devices they compound into a serious operating line. Model this over the full asset life, not the pilot.

Coverage is genuinely not universal, particularly rural and underground. Verify with a survey on the actual site before committing — carrier coverage maps are marketing, not measurement. And you are dependent on the operator’s roadmap; network sunsets have caught IoT deployments out before.

Power behaviour also surprises people. NB-IoT transmission draws considerably more current than LoRa. Power Saving Mode and eDRX make multi-year battery life achievable, but only if configured properly — a device that wakes too often will flatten a cell in months.

Sigfox

Sigfox is an ultra-narrowband operator network with deliberately severe constraints: uplink messages are limited to 12 bytes, with a cap around 140 messages per day, and downlink is far more restricted still.

Those limits are a feature for the right use case — a leak detector or a bin-full sensor that sends a handful of bits occasionally is a genuinely good fit, and device cost and power draw are excellent.

The honest caveat

Sigfox’s operating company entered administration in 2022 and the business was acquired by UnaBiz. The technology continues and networks operate, but for a new industrial deployment with a ten-year horizon that history belongs in your risk assessment. We would want a clear answer on national coverage commitments and long-term pricing before designing an estate around it.

How to actually choose

Work through these in order. The first one that gives a decisive answer usually is the answer.

1. Are your assets on sites you control?

If yes, and they are clustered, LoRaWAN is usually the economically correct answer — you amortise a gateway across many devices and pay nothing per message. If your assets are scattered across a region, private infrastructure stops making sense and cellular wins.

2. What is your real payload and cadence?

Write down the actual bytes and the actual interval. Then double the interval, because someone will ask for more frequent readings later. If that number exceeds a few hundred bytes per hour, you are drifting out of LPWAN and towards LTE-M.

3. Do you need downlink?

Reconfiguring devices remotely, acknowledging alarms, actuating anything. Downlink is the weakest capability across all three, and duty cycle limits bite hardest here. If remote control is core rather than occasional, question whether LPWAN fits at all.

4. What does year five cost?

Build both models properly: LoRaWAN’s capital cost in gateways and installation against NB-IoT’s per-device subscription over the asset life. The crossover point moves with device count, and it moves a long way. Small estates often favour cellular; large ones almost always favour private LoRaWAN.

5. Who operates it at 3am?

An honest question about your team. Owning a network means owning its failures. If nobody will have time to diagnose a gateway backhaul fault, that argues for a carrier network regardless of what the spreadsheet says.

The pragmatic answer: often both

Plenty of mature deployments are hybrid. Private LoRaWAN inside the plant where density is high and you control the estate; NB-IoT for the outlying assets where running a gateway is not justified. Both feed the same data pipeline, and the devices do not care.

Designing for that from the start — a transport-agnostic ingestion layer rather than a LoRaWAN-shaped one — costs very little at the outset and saves a rebuild later.

Common questions

Is LoRaWAN better than NB-IoT?

Neither is better in general. LoRaWAN wins where you control clustered sites and want no per-message cost; NB-IoT wins for scattered assets where owning gateways is impractical. The deciding factors are asset geography and total cost over the asset’s life, not radio performance.

How far does LoRaWAN actually reach?

Far enough that the honest answer is “survey it”. Range depends overwhelmingly on antenna height, terrain and obstruction — a gateway on a tall mast in open country behaves nothing like one inside a steel-framed building. Treat any single quoted figure with suspicion and measure on site.

Can I run LoRaWAN without The Things Network or a commercial provider?

Yes. You can run your own network server on your own infrastructure, which is what most industrial deployments with data-sovereignty requirements do. Public community networks are excellent for prototyping but rarely appropriate for production plant telemetry.

Which is best for battery life?

LoRa and Sigfox devices generally draw less energy per transmission than NB-IoT. But configuration dominates: a well-configured NB-IoT device using Power Saving Mode can outlast a badly configured LoRa device that transmits too often. Cadence matters more than the radio.

Where to go next

If you are early enough that the radio decision is still open, that is exactly the right moment for a second opinion — the cost of changing it later is what makes it worth getting right now.

Our IIoT Foundations workshop covers radio selection, MQTT topic design and OT protocol bridging hands-on, with real hardware on the bench. If you would rather talk about a specific site, a free technical audit is a sensible starting point.