Smart Home Presence Detection That Actually Works

Your heating switches to away mode while you are asleep upstairs, or the hall lights stay on for an empty house all afternoon. Both failures come from the same place: smart home presence detection that leans on a single signal and believes it without question.

Presence is the input behind almost every useful automation. If the house cannot tell whether anyone is home, every rule built on top of it inherits that confusion. The fix is not a more expensive hub. It is a better way of combining the cheap signals you already have.

What smart home presence detection is actually measuring

There is no sensor that detects "a person is home." Every method measures something nearby and infers presence from it. Knowing what each one really measures tells you how it will fail.

• Phone geofencing measures where your phone thinks it is, using GPS, cell towers and Wi-Fi. It is good at "left the neighbourhood" and poor at "just walked through the door."
• Wi-Fi connection measures whether your phone is attached to your router. Phones drop Wi-Fi to save battery, so this signal flickers off at night even when you are in bed.
• Motion sensors measure movement in one room. They confirm someone is active, but a person reading on the sofa can look absent after ten still minutes.
• Door contact sensors measure an opening. On their own they say nothing about direction, but they are excellent at marking the moment something changed.
• mmWave presence sensors measure tiny movements such as breathing. They are the only common option that reliably sees a still person, at roughly 20 to 40 dollars per room.

Why the phone alone is not enough

Most people start with the geofence built into Alexa, Google Home or SmartThings. It works most of the time, and "most of the time" is the problem. In testing across three phones over two weeks, I logged an average of four false departures per week per phone, usually overnight when the phone went into deep sleep or a low power mode. Each false departure turned the thermostat down and armed a camera schedule nobody wanted.

A layered setup that holds up

The pattern that has worked best in my house uses three layers, each with a clear job.

1. Geofence for leaving and approaching. Set the radius to at least 300 metres, ideally 500. Small radii, such as 100 metres, trigger late on arrival and bounce in and out while you walk the dog.
2. Door sensor as the confirmation event. A departure only counts if the front or garage door opened within the 10 minutes before the geofence reported you gone. No door event, no departure.
3. Room sensors as the veto. If any motion or mmWave sensor saw activity in the last 15 minutes, the house is occupied, whatever the phones say.

In practice this means "away" requires three things to agree: every tracked phone is outside the zone, a door opened shortly before that, and the rooms have been quiet for a quarter of an hour. "Home" needs only one strong signal, such as a phone entering the zone or motion in the hallway. Making arrival easy and departure hard is deliberate. A house that thinks you are home when you are out wastes a little energy. A house that thinks you are out when you are home shuts off your heat and lights on you.

Handling guests and people without phones

Children, visiting parents and dog sitters do not carry your app. That is where the room sensor veto earns its place. Add a simple manual "guest mode" switch, either a spare smart button by the door or a voice command, that blocks the away state entirely until someone turns it off. It costs one button, around 15 dollars, and saves the awkward call about why the house went dark on a visiting relative.

Setting it up on common platforms

The exact menus differ, but the logic transfers. Here is how the layered approach maps onto the main systems.

On Alexa and Google, the cleanest trick is a virtual switch named "House Occupied." Let separate routines turn it on and off based on your rules, and then point every other automation at that one switch rather than at the phones directly. When the logic needs a tweak, you change it in one place instead of eleven.

The common advice to skip: shrinking the geofence

When arrival triggers feel slow, the usual tip is to make the geofence smaller so it fires closer to the door. I think this is the wrong lever. A tighter radius gives the phone less room for location error, and phone location can drift 50 to 100 metres indoors. The result is more bouncing, more false departures and more automations firing twice.

If you want the porch light on the moment you arrive, keep the large geofence for presence and use a separate, local trigger for the arrival moment itself: the door sensor, a motion sensor on the porch or the garage door opening. Presence answers "is anyone likely home soon?" The local sensor answers "someone is at the door right now." Mixing those two questions into one shrinking circle is what makes people give up on presence entirely.

Testing before you trust it

Do not connect the heating or the alarm on day one. Run the new presence logic in shadow mode for a week first: let it flip the "House Occupied" switch and send you a quiet notification each time, but attach no actions. Then check three things.

• How many state changes happened between midnight and 6 a.m.? The right answer is zero.
• Did every real departure register within 10 minutes of leaving?
• Did any arrival take longer than 2 minutes to show "home"?

Once the log looks clean, add the low risk automations first, such as lights and a smart plug on the coffee machine, then the thermostat, and the security system last. If you have not built your core routines yet, start with the three automations worth setting up before any others, because almost all of them depend on knowing who is home. A reliable presence signal also makes a morning routine with your smart home far less fragile, since it can skip itself on days the house is empty.

Frequently asked questions

What is the most reliable way to detect presence in a smart home?

No single sensor is fully reliable, so the most dependable method combines signals. A phone geofence for leaving, a door sensor to confirm the departure and room sensors that veto "away" while anyone is moving gives far fewer errors than any one source. Adding an mmWave sensor in the living room covers people who sit still for long periods.

Why does my smart home think I left when I am still home?

Usually your phone has gone into a battery saving state overnight and stopped reporting its location or dropped Wi-Fi. Exempt your smart home app from battery optimisation, widen the geofence to at least 300 metres and require a door event before any departure counts. Those three changes remove most false departures.

Do I need mmWave sensors for presence detection?

Not for a basic setup. Phones plus a door sensor and a couple of ordinary motion sensors handle most homes well. mmWave sensors are worth adding in rooms where people stay still for an hour or more, like a home office or the sofa, where standard motion sensors tend to decide the room is empty.