The radio decision is upstream of the device decision
Picking a radio technology before you've sized the project's data, range, and power constraints is the source of most failed IoT pilots. Each of the major options below wins in a band and loses outside of it. Here's the practical framing.LoRa / LoRaWAN
Where it wins: kilometre-scale range, battery life measured in years, sub-Hz duty cycle. The right answer for "I have 500 sensors spread across a 5 km industrial site, each sending a few bytes every 10 minutes."Where it hurts: data rate is tiny (typically <1 kbps effective). Latency is high (gateway batching, scheduling). LoRaWAN's regional regulations limit duty cycle aggressively in EU868. Encryption is per-device-key, manageable but operationally heavy.
Use when: sparse data, long range, hard battery budget. Cold-chain monitoring, agricultural sensing, perimeter sensors at large industrial sites.
NB-IoT / LTE-M / 5G RedCap
Where it wins: you don't have to deploy gateways. The cellular operator's network is your infrastructure. Coverage, where it exists, is excellent. Multi-year battery life is achievable on NB-IoT.Where it hurts: cellular subscription costs scale per device. In-building coverage in metal-heavy industrial environments is variable — survey before committing. Operator dependence (and roaming, for international deployments) is a real consideration.
Use when: geographically distributed devices, customer doesn't want to operate gateways, low data rates per device acceptable. Asset tracking, fleet telemetry, smart meters.
Wi-Fi (industrial, not consumer)
Where it wins: high throughput, existing infrastructure (sometimes), familiar protocol. WPA3-Enterprise is a credible auth story for industrial deployments.Where it hurts: power. Always-on WiFi is a battery killer. Roaming between APs is fragile in many implementations. Channel congestion in dense industrial deployments is a real engineering challenge.
Use when: mains-powered devices, high bandwidth (camera, audio, log streams), facility already has industrial-grade WiFi infrastructure.
Bluetooth Low Energy / BLE Mesh
Where it wins: short-range (10-50 m), peripheral interaction, very low power. Pairing with a mobile device for setup or maintenance is solved. BLE Mesh is workable for hundreds of nodes in a building.Where it hurts: not great as a backbone — meshing is complex to debug, throughput drops as the mesh grows. Range is genuinely short, especially in metal environments.
Use when: room-scale sensor networks, beacons, maintenance-engineer-with-a-phone-paired-to-the-device patterns.
802.15.4 / Zigbee / Thread / Matter
Where it wins: low-power mesh, building-scale deployments. Thread / Matter is gaining real traction as the Apple/Google/Amazon-backed standard for building-grade IoT.Where it hurts: industrial deployment is rare; mostly seen in commercial buildings.
Use when: the project is in the smart-buildings space rather than heavy industrial. Crossover is increasing.
The decision tree we use
- Range > 1 km, battery-powered, sparse data → LoRa
- Geographically distributed, no gateway infrastructure → NB-IoT / LTE-M
- Mains-powered, indoor, high data rate → industrial WiFi
- Short-range, battery-powered, occasional comms → BLE / 15.4
- Hard real-time process control → none of the above. Wired EtherCAT / Profinet, full stop.
The thing nobody plans
RF site survey before deployment. Spectrum analyzer, walk the site, log noise floors at the operating frequencies. The "we'll cover it later" approach gets discovered at commissioning, when the WiFi is unusable in half the facility because the nearby variable-frequency drive is radiating broadband.One pattern worth adopting
Hybrid topologies. LoRa for the field, WiFi for the gateway uplink, cellular as a fallback. The cost of redundancy is small compared to the cost of an outage at a remote site.What's your topology? Anyone fully migrated to private 5G in industrial deployments yet?
