Home security that is not theatre: what the evidence says deters burglars

A great deal of home security hardware is sold on fear and bought on impulse, and the relationship between what people spend and how safe their home actually is turns out to be weak. The good news is that the research on residential burglary is reasonably consistent, and it points at a small number of measures that work and several popular ones that mostly do not.

Here is what the evidence supports, and why.

How burglary actually happens

The picture that emerges from police statistics and offender interviews across many countries is remarkably consistent, and it is not the picture in most people's heads.

Most residential burglary is opportunistic rather than planned. The typical offender is looking for a property that appears empty, offers easy access, and can be entered and left quickly without being seen. Selection happens in seconds from the street.

Most entry is through a door rather than a window, and a substantial share involves no forced entry at all — an unlocked door, an open window, a key in an obvious place. Forced entry, when it happens, is usually crude: a kicked door, a snapped cylinder, a levered frame.

And most burglaries are short. Offenders consistently report wanting to be inside for only a few minutes, and anything that extends the time required or the chance of being seen is a strong deterrent.

Three implications follow, and they shape everything below. Looking occupied matters. Physical delay at doors matters. And being seen matters.

What works, in rough order of value per pound spent

1. Locking what you have

Unglamorous and free. A meaningful proportion of burglaries involve no force whatsoever. A door sensor that tells you the back door has been left unlocked, or a reminder when everyone leaves, addresses a real and common failure.

2. Good physical door security

The cylinder is the weak point on most doors. Anti-snap cylinders resist the technique of snapping the lock barrel, which is fast, quiet and requires no skill. Look for a cylinder tested to a recognised anti-snap standard, add a reinforced strike plate and long screws into the frame, and you have removed the commonest forced entry method for a modest sum.

This is the least exciting recommendation on this page and probably the best.

3. Looking occupied

Since selection is based on apparent emptiness, credible occupancy is a direct deterrent — and this is where automation is genuinely good, as opposed to merely convenient. It gets its own section below.

4. Visible cameras and doorbells at the approach

Offender interviews suggest cameras do deter, largely because they suggest a resident who takes security seriously and because they raise the risk of identification. The deterrent value comes from being seen, which means placement at the approach matters more than image quality.

Cameras also help after the fact, though rather less than people expect: footage of a hooded figure at night is often not identifying. Value them for deterrence and for knowing what happened, not for guaranteed detection.

5. Lighting that responds

Lighting that comes on when someone approaches is more deterrent than lighting that is on all night, because the change is what signals attention. Permanently lit gardens simply help people see what they are doing.

6. Layered detection inside

Door and window contacts plus a couple of interior motion sensors give you fast, reliable detection with very few false alarms — because a door opening followed by interior movement is a much stronger signal than either alone.

What matters less than the marketing implies

Very high camera counts. Beyond covering the approaches and entry points, additional cameras add footage to review rather than security.

Extremely high resolution. Useful, but low light performance and sensible placement matter far more for identification than pixel count. A well-lit 2-megapixel view beats a badly placed 4K one.

Loud sirens alone. Alarms that nobody responds to are background noise in most neighbourhoods. A siren is worth having as part of a system that also notifies you and records; it is weak on its own.

Fake cameras and alarm box dummies. Cheap, and offenders are often better at spotting them than homeowners expect.

Hidden cameras indoors. Rarely useful for burglary, frequently a legal and ethical problem, and a genuinely bad idea in any space where guests, cleaners or lodgers have a reasonable expectation of privacy.

Making an empty house look occupied, credibly

This is where a smart home does something a traditional alarm cannot, and doing it badly is worse than not doing it.

The classic failure is a lamp on a timer that comes on at 7pm every single day, including in June when it is broad daylight. That signals absence rather than presence.

Credible occupancy has four properties:

It follows the sun, not the clock. Lights that come on relative to sunset track the seasons the way a real household does.

It varies. Different rooms, slightly different times, some randomisation. Real households are inconsistent.

It moves through the house. Kitchen in the early evening, living room later, a bedroom and a landing last, all off by a plausible hour. A single room lit all evening reads as a timer.

It includes more than lighting. Curtains or blinds that open and close, and radio or television sound in the evening, are strong signals. An upstairs window with the curtains still closed at 4pm is a strong signal in the other direction.

Two things not to automate: do not post about being away, and do not let deliveries pile up visibly. Both undo everything above.

The false alarm problem, which is really the trust problem

Every alarm system faces the same failure: if it cries wolf, people stop responding to it, and a system nobody responds to protects nothing.

The commonest causes of nuisance triggers are worth designing out from the start. Pets set off passive infrared motion sensors constantly; either use pet-immune sensors, mount them high and angled downward away from floor level, or rely on door contacts in rooms the animals use. Sunlight and heat sources confuse infrared sensors, so avoid pointing them at windows, radiators or anything that warms rapidly. Draughts move curtains and blinds, which reads as movement.

The most effective structural fix is confirmation: requiring two independent signals before treating something as an intrusion. A door contact opening on its own might be wind or a badly fitted frame. A door contact followed within thirty seconds by interior motion, while nobody is home, is a genuine event. That single rule removes most false alarms without weakening real detection, and it is the reason layered sensors beat a single clever one.

Finally, be honest about what you will actually do when notified. An alert that arrives while you are in a meeting and gets dismissed unread is not protection. Decide in advance which events warrant an interruption, and let the rest simply be logged.

Where to start

1. Walk your own property as an opportunist would. Which door is least overlooked? What is visible through the front window? Where would you go?
2. Fix the physical basics — anti-snap cylinders, reinforced frames, window locks, and anything left unlocked out of habit.
3. Add contacts to the doors and windows you identified, plus one or two interior motion sensors.
4. Put a visible camera or doorbell at the main approach.
5. Set up credible occupancy lighting that follows sunset and varies.
6. Test the notification path by triggering it deliberately. A system you have never seen alert you is a system you do not know works.

Steps one and two cost the least and prevent the most. Everything after them is refinement.