The communication choice is made on a drawing and fails at the meter position. Pick it after you know what is reachable where the meter will actually sit.
NB-IoT
Runs on licensed cellular spectrum with a SIM in the unit, so no gateway of your own is needed. Strong penetration into basements and meter pits, which is where most other options fail. It needs cellular coverage at the meter and a data plan per unit, and the operator has to support NB-IoT in your country.
LoRaWAN
Your own network on unlicensed spectrum — you control the gateway, the coverage and the running cost, with no per-unit subscription. Best where you already run a LoRaWAN network or are building one across a district. It needs a gateway in range of every meter, which is worth testing on the worst position in the pilot rather than the best.
4G
The widest coverage and the highest bandwidth of the three, which suits meters sending frequent or larger payloads — dual-channel ultrasonic on a DMA boundary, for example. It uses the most power, so battery life is the constraint to plan around.
How to decide
- No network of your own, meters scattered across a city → NB-IoT or 4G.
- You already run LoRaWAN, or one site with many meters → LoRaWAN.
- Basements, pits and chambers → test penetration first; NB-IoT usually wins.
- Frequent upload intervals or larger payloads → 4G, and size the battery accordingly.
Mixing is normal
A single project can use two: LoRaWAN across an estate where the gateway reaches, NB-IoT on the blocks it does not. Decide per position rather than per project, and test at the meter during survey rather than at street level.