Wi-Fi vs Cellular vs LoRa vs BLE: Choosing IoT Connectivity

Picking the radio is one of the few hardware decisions you cannot quietly fix in a firmware update. It drives your antenna design, your power budget, your bill of materials, your certification bill, and whether your product has a recurring carrier cost for its entire life. Get it wrong and you either ship a device with two-day battery life or a SIM bill that eats your margin.

Here is how I reason through Wi-Fi, cellular, LoRa, and BLE before committing.

The five factors that actually decide it

• Range. BLE: ~10-30 m indoors. Wi-Fi: ~50 m. LoRa: 2-15 km line-of-sight. Cellular: wherever there's a tower.
• Power. BLE and LoRa can run years on a coin cell or a couple of AAs. Wi-Fi and cellular pull hundreds of mA in bursts and want a real battery or wall power.
• Data rate. Wi-Fi: tens of Mbps. Cellular (LTE Cat-1): a few Mbps. BLE: ~1 Mbps practical. LoRa: a few hundred bytes per message, and you're rate-limited by duty cycle. LoRa cannot stream anything.
• Recurring cost. This is the one engineers forget. Cellular means a SIM and a data plan per device, forever. Budget $1-5/device/month even on IoT plans. Across 5,000 units that's $60,000-300,000/year. Wi-Fi, BLE, and LoRa (on your own gateway) have zero airtime cost.
• Certification burden. Every radio you put in the box needs intentional-radiator approval: FCC Part 15 in the US, RED (EN 300 328 etc.) in the EU. Cellular adds the worst of it — carrier certification (PTCRB / GCF) on top of regulatory, which can run thousands of dollars and weeks of lab time. Two radios = roughly double the test matrix.

How the choice rewrites your sourcing

The decision doesn't stop at the schematic — it changes which factory you go to and how hard certification is.

Use a pre-certified module and you inherit its modular grant. An ESP32 module ships with FCC/CE IDs already; you only test the host as an unintentional radiator. Roll your own RF on a bare chip and you own full intentional-radiator testing yourself. For a first product run, the pre-certified module is almost always the right call.

For LoRa, sourcing is also where you can claw back margin. On a Japan LoRa gateway project, China Sourcing Agents found the SX1302-based module and gateway directly at the factory in Shenzhen instead of through a distributor, cutting unit cost about 22% on the run. The module you pick is the long pole; the supplier you pick decides what it costs.

A decision shortlist

Run these in order and stop at the first "yes":

1. Does it stream audio/video or pull MB of data? → Wi-Fi (mains/large battery) or cellular (mobile).
2. Does it need to work anywhere, with no local gateway? → Cellular. Accept the per-device monthly cost and PTCRB testing.
3. Battery-powered, sends small readings, fixed-site, long range (farms, meters, asset tracking)? → LoRa, paired with a gateway you control.
4. Battery-powered, talks to a nearby phone or hub, short range? → BLE.
5. Mains-powered, in a home/office with Wi-Fi already there? → Wi-Fi.

Two practical notes: if you're tempted to add a second radio "for flexibility," cost out the extra certification first — it's rarely worth it on a first run. And always pick a module with a published reference design and a part already used in shipping products; a slightly more expensive module with a proven layout beats a cheap chip that needs three RF respins.

Choose the radio for the job in front of you, source the module that already carries its grants, and treat recurring airtime as a line item from day one. That's the difference between a product that ships and one that stalls in the test lab.