Which radio for which job

The four questions that decide every case#

Before the table, the logic. Every device class answer below falls out of these four:

1. Is it battery powered? If yes, Wi-Fi is disqualified. A Wi-Fi radio has to keep an association alive, renew DHCP leases and answer the AP; a sleepy 802.15.4 or Z-Wave end device turns its receiver off between reports and parks incoming messages at a parent. That is the difference between changing a cell every few months and changing it every few years.
2. Is it mains powered and in a permanent location? If yes, it should be a mesh router, not just a mesh client. Every mains-powered Zigbee bulb, Thread-capable plug and Z-Wave switch you install makes the mesh stronger for the battery devices around it.
3. Does it need to work when the internet is down? If yes, avoid cloud-only Wi-Fi devices and keep the automation local. See local vs cloud control for where the command is actually processed.
4. Is there a wall, a driveway or a slab between it and the hub? 908.42 MHz penetrates construction materials better than 2.4 GHz, and Z-Wave Long Range trades the mesh for raw link budget. That is the whole reason Z-Wave still exists.

The decision table#

Class by class, with the reasoning#

Ceiling lights: switch first, bulb second#

The switch-versus-bulb decision matters more than the radio. A smart bulb behind a dumb switch fails the moment somebody flips the switch, which is why smart lighting that stays compatible starts there and switches, dimmers and the neutral wire covers the US wiring reality.

Once you have decided, the radio follows. In-wall switches and dimmers are the best mesh infrastructure you can buy: mains powered, permanently located, one per room. Z-Wave switches keep that traffic off 2.4 GHz entirely, which is worth real money in an apartment building. Zigbee switches are cheaper and more widely available. Either is right; Wi-Fi dimmers are the one option that gives you nothing back, because they add load to your access point and repeat nothing.

If you are buying bulbs, Zigbee is still the volume answer, and every mains-powered bulb acts as a router for the sensors near it. Hue runs Zigbee under its own bridge, which is a slightly different trade covered in proprietary radios.

Battery sensors: never Wi-Fi#

This is the least ambiguous line in the table. Motion, contact, leak and temperature sensors report a few bytes, rarely, and then sleep. Both 802.15.4 (Thread and Zigbee) and Z-Wave were designed for exactly that pattern: the device keeps its receiver off, wakes on its own schedule or on an event, transmits, and goes back to sleep. Its parent holds any message addressed to it in the meantime.

A Wi-Fi sensor has no such role. It either stays associated, which costs power, or it re-associates on every event, which costs latency and power. Pick Thread if your hub is a Thread border router and you want Matter native devices; pick Zigbee if you want the widest and cheapest catalog; pick Z-Wave if the sensor is far away or behind masonry. The battery math and reporting intervals are worked through in sensors: motion, contact, temperature, leak.

Locks: Z-Wave first, Thread second#

Locks are the case where 908.42 MHz earns its keep. The radio is in a metal-heavy assembly, usually in an exterior door, often at the far end of the house from the hub, and it is battery powered. Z-Wave gets through the door and the wall better than 2.4 GHz does, and Z-Wave S2 pairing uses a device-specific PIN or QR code (the DSK) so the join cannot be spoofed by proximity alone.

Thread is a legitimate second choice and is where Matter locks are heading. What both share is a hard rule: do not buy a lock whose only radio is Wi-Fi or Bluetooth. A Wi-Fi lock stops working when the router reboots. A Bluetooth-only lock cannot be automated from anywhere but your phone. Smart locks and the Aliro standard covers credentials and offline behavior in detail.

Outdoor gate, well, mailbox: Z-Wave Long Range#

Once a device is outside the building envelope, mesh stops being an advantage and starts being a problem, because there is nothing out there to put a repeater on. Z-Wave Long Range answers that directly: it drops the mesh and uses a star topology, every device talking straight back to the hub, with no hops. It allows up to 30 dBm transmit power, up to 4,000 nodes, and the Z-Wave Alliance cites 1.5 miles line of sight and up to 10 years on a coin cell with dynamic power control.

The catch is that your hub and controller firmware must support LR inclusion. An 800-series stick on its own is not enough. Z-Wave and Z-Wave Long Range covers what to check before you buy.

Wi-Fi HaLow is the technology that should own this niche and does not yet. It is certified, sub-1 GHz, and ranges over a kilometer, but as of September 2026 there are no mainstream smart home products and no ecosystem support. Treat it as a thing to watch, not a thing to buy. The reasoning is in Wi-Fi smart devices: the hidden costs.

Cameras and doorbells: bandwidth wins#

There is no debate here, only arithmetic. IEEE 802.15.4 carries 250 kbit/s shared across the whole channel. A 1080p stream needs one to two orders of magnitude more than that. So cameras are Wi-Fi or Ethernet, full stop, and the interesting decisions are about recording and subscriptions rather than radio. Run them on 5 GHz if the signal supports it, because that keeps them out of the 2.4 GHz band your sensors live in. Wired PoE is better again. See cameras and doorbells and Matter cameras for what the standard actually exposes.

Blinds, shades and garage doors#

Battery blinds have an unusual power profile: near zero for weeks, then a hard motor draw for a few seconds. That favors any radio with a cheap idle state, which means the mesh radios and not Wi-Fi. Wired shades should be treated like in-wall switches: permanent, mains powered, therefore repeaters.

Garage controllers are usually mains powered inside a metal box in a detached or semi-detached structure. That is a bad place for 2.4 GHz and a decent place for Z-Wave. If you do choose a Wi-Fi controller, insist on documented local control, because a garage door that depends on a vendor cloud is a garage door that stops opening when the vendor has an outage. Blinds, shades and garage doors covers the position reporting and safety side.

What a coherent stack looks like#

Most well-behaved US homes end up with the same shape:

• Wi-Fi for cameras, doorbells, TVs, speakers and anything with a screen, ideally on a dedicated 2.4 GHz IoT SSID as described in the network setup a smart home wants.
• One 802.15.4 mesh (Thread or Zigbee, not both if you can avoid it) for lights, plugs and indoor sensors.
• Z-Wave for locks, and for anything outside or behind masonry.

Running Zigbee and Thread together is fine, but remember they share the same PHY and the same 16 channels. Put them on different channels from each other and from your Wi-Fi, or they will collide. That is what 2.4 GHz interference and channel planning and the channel planner are for.

When the right radio is not an option#

Sometimes the correct answer is unavailable: your hub does not speak Z-Wave, or the device you want only exists in Wi-Fi. Two honest fallbacks:

• Add a second controller rather than compromise the radio. A Z-Wave stick on a local controller alongside your main platform is cheaper than replacing a lock that never responds. Running two platforms on purpose covers the pattern.
• Accept Wi-Fi, but insist on local control. A Wi-Fi device with a documented local API and no cloud dependency is a reasonable citizen. A cloud-only one is a subscription waiting to happen, and when a smart home company shuts down is the history of what that costs.