In brief
When somebody says “I want an alarm”, they are usually thinking of sensors. And sensors are only the first of three things that have to happen: something has to be detected, somebody has to confirm it, and somebody has to respond. The conversation usually stops at the first. I write about the other two, about the layers of detection, and about a decision that is made before the walls are plastered: whether your system needs the manufacturer’s server in order to be armed. If it does, it works for as long as that server works. And for as long as that company exists. One caveat up front: I do not design alarm systems and I do not pick security grades. What I do see, as the contractor, is which decisions get made before the plastering and at the windows.
Detection, confirmation, response
Sensors are visible. You can count them, mark them on a drawing and price them – which is why the conversation starts there and usually ends there. Except that the signal on its own settles nothing. Somebody has to check it, and then somebody has to turn up.
From my conversations with clients, it is that last part that gets the least thought. The house has sensors, the sensors work, the siren goes off. And nothing follows from it, because nobody agreed who drives out and how they would know it was worth it. I call that arrangement a doorbell. A doorbell is not worthless – noise can cut somebody’s visit short, and the neighbour will sometimes look out of the window. It is only that nobody has promised anyone a response, so there is no point counting on one.
Where the system is actually looking
Detection is built in layers, and that is the whole of it: the boundary of the plot, the shell of the building, the inside. In houses it usually stops at the third layer: motion sensors indoors, sometimes a few opening contacts on the doors. Which means the system speaks up only once somebody is already in the house, a few metres from the bedroom. The detection happens. It just happens at a point where very little follows from it.
Sensors outside make sense for one reason: they give the signal earlier, before anyone reaches the building. If the system knows about someone while they are still at the fence, that person’s whole walk across the plot is left to you – for the notification, for the check, for the decision. On a large plot that walk can be long, and it is the only margin in the chain you do not have to buy from anyone.
Honestly: outside is harder, and there is no point pretending otherwise. Indoors you get false alarms too – a pet, a draught, a fireplace, a robot vacuum – but the environment there is incomparably more predictable. At the boundary you have a deer, next door’s cat, a branch that grows and after two years starts knocking against the railing (nobody thinks about it at handover, because it is not there yet), rain, snow, and the sun rising straight into the sensor. An indoor sensor moved outside unchanged will go off for no reason for as long as it takes you to switch it off. And then the first layer stops existing, even though it is still there in the costings. Outdoor detection is designed differently from indoor detection. It is a separate piece of work and a separate line in the design.
There is one more thing the trade is quiet about: the language used to describe alarms stops at the wall of the building. The standard everyone quotes covers indoor systems and explicitly excludes external installations from its scope. For detection at the fence itself there is no harmonised standard describing it the way the grades describe the system indoors. It is still worth doing. It is only that at the fence line nobody will hand you a number that settles everything, and the requirements, the working conditions, the false alarms and the response procedure all have to be agreed separately.
The glazing decides where the system works
This is where a residence differs most from an ordinary house. There is a lot of glass here: whole walls, sliding panes, glazed corners. Somebody who wants to come in through glass will not force a security door standing in plain view – they will pick a window at the side of the building, somewhere less visible. Locks and doors are usually the strongest part of the shell, and at the same time the part everyone prices most carefully.
On top of that comes a distinction that changes everything in a house with walls of glass: opening a leaf and going through the pane are two different events. They are detected by two different means, and the technical guidance to the standard treats them separately. In practice that means that if the shell rests on opening sensors alone, a house can have them on every leaf and still not detect somebody stepping through a broken pane.
And the consequence people usually learn too late: concealed opening sensors are ordered together with the windows. A factory-fitted sensor, wired later into the alarm installation, is a line in the joinery quote. So the decision about the first layer of the shell is made when you choose your windows, not when the alarm company arrives. You can add them later, surface-mounted, and sometimes that is how it ends. It is just that you can then see somebody was late.
Why one signal is not enough
A system that goes off every week for no reason stops existing. It does not have to break. It is enough that people learn it. The neighbour stops looking out of the window. The security firm treats the next call from that address like the last one. You yourself, after three false alarms in a month, stop arming the house when you go out for an hour. Credibility is the condition of anyone responding at all. Without it you are left with noise.
Hence the rule everything else stands on: two independent events mean more than one. If something was noticed in two places, or by two different means, certainty goes up. And only then does it make sense to put people in a car. Monitoring stations tell a confirmed alarm from an unconfirmed one for exactly that reason: somebody has to decide whether to send a patrol. Outside, there is no conversation to be had without that rule. With a single sensor on the fence, the wind and the deer will get into the statistics faster than a person.
How it is put together in a particular house, I will not write. Specific detection layouts are not knowledge to publish on the internet.
Whose system is it, really
An alarm has one feature that sets it apart from the rest of the house: it has to work when other things fail. Which is why the question “what does this system depend on” matters more than the question “what can this system do”.
In some of today’s solutions – the good-looking ones with an app especially – the panel in your plant room talks to the manufacturer’s server, and you talk to the same server through your phone. Convenient. And that is where the line runs, the one you have to ask about directly, because the brochure will not show it: in some systems the app is only an addition and you arm the house on the spot, whatever the network is doing. In others the server sits in the control path. Then arming your own house is a service provided by a company that sits somewhere else, answers to a different jurisdiction, and owes you nothing as to how long it will keep that service running. That server runs for as long as it pays someone to keep it running. When it stops, a perfectly good device on your wall loses whatever passed through it: sometimes only the app and the notifications, sometimes everything. That is not a fault – it is the end of support, and nobody has to ask your opinion.
Companies fail, get bought, change their model. The house stays where it is (equipment bought for a residence is meant to work longer than many a start-up).
So I put the requirement simply: the system has to work without the internet and without the manufacturer’s server. Arming, disarming, detection and signalling have to happen on the spot, in the house, without asking anyone’s permission. Notification is the one element that by its nature has to go outside – which is exactly why it needs its own independent path, and I come to that in a moment. Remote access is fine and I have nothing against it, as long as it is an addition rather than a condition of the system working, and as long as the channel is yours.
And honestly: a local system does not make you safe. It removes one route – the remote one. The physical route remains, and the rest of this text is about that.
Blinds, lights and the alarm
Home automation can run everything these days: lights, blinds, the gate, heating, ventilation, and monitoring and the alarm along the way. Precisely because it can do everything, this one thing is better not handed to it.
Blinds are supposed to be convenient, an alarm is supposed to be predictable. Those are two different aims, and you see it most clearly in the people. An automation system lives: the installer comes back two or three months after you move in, asks how you are getting on, and adjusts the settings. Then he comes back many times more, because the children, the furniture and the habits all change. That is automation’s strength – it is meant to be tuned. Just think for a moment what it means if the alarm sits on the same bus: the man adjusting the dimmer in your living room has access to the layer your alarm panel works on.
The automation installer has no bad intentions in this. The point is a rule applied everywhere that anything matters: everyone gets access to what they need for their work, and to nothing else. The lighting engineer does not need the alarm to do his job and should not get it thrown in.
Then there is something clients are rarely told: automation devices are certified first and foremost on whether they talk to each other properly. They may well have protections of their own, but nobody assesses them the way alarm components are assessed: for sabotage, line supervision and alarm procedures. A device working well within its own system does not mean it meets what is asked of an alarm system. That does not make it worse. It just does a different job.
So the two systems are run separately and joined at a single, defined point of contact: automation sees only what someone deliberately showed it, and an automation failure has no bearing on the alarm. That is a decision from the installation design stage, because two systems mean two cable routes.
That separation has a price and I will say it plainly: sensors get duplicated, there is more cabling, and sometimes two scenarios have to be reconciled so the lights do not come on while the alarm is armed. It looks like excess, and I understand the objection. I think it is worth it, but I want it to be your decision, made knowingly.
Power and the line
In any serious scenario you have to assume the power and the line may be the first things to go. There is no secret in that and nobody in the trade pretends otherwise – which is why battery backup and a second, independent notification channel are not top-shelf extras. They are the condition of the system working at the one moment it is needed.
There is a mechanism here I like more than anything else in this field, because it is the opposite of what you would expect. If the transmission path is supervised, the line speaks for itself: the panel and the monitoring station keep exchanging a sign of life, so cutting the cable or jamming the network does not silence the system, it wakes it. The absence of a signal is the signal. The condition is a hard one, though: the path has to be supervised and the station has to have a procedure for loss of contact. You have to ask about that directly, because from the outside such a set-up looks identical to one that will simply go quiet.
Who actually turns up
I ask this question at the start, though in conversations it tends to come back at the very end. Sensor settings can be chosen and tuned – that is still only technology. The measure of a system is who turns up, and how long it takes.
The answer depends on where the house stands. For the residences I build outside town, the real drive time can be nothing like the owner’s picture of it, and it is better to check that at a stage where something can still be done about it than to find out later. The contract with the monitoring station is a separate subject and a separate document; it is worth reading more carefully than usual, because that is where it says what actually happens after the signal.
The people on the list belong to the system too: a neighbour, a brother, someone who holds keys. They are part of the system only if they know what to do. Someone who gets a notification and has no idea whether to drive over, phone, or wait, simply gets anxious. That is all that comes of it.
Detection buys time. That time has to be spent on something. And that is where the conversation about what happens next begins: about barriers, about escape routes, sometimes about a protected room, which I wrote about separately in the piece on panic rooms.
When to decide
An alarm and cameras are extra-low-voltage work, so the cable routes are made together with the electrical installation, before the plastering. After that every change costs: sometimes chasing walls, sometimes a visible trunking, sometimes a compromise. The cameras themselves – where the footage goes and why it should not leave your property – I cover separately in the piece on home CCTV.
So on ordinary houses I tell clients the simplest thing there is: while the electrician is running the installation, ask him to add cables for an alarm and cameras to the place where a panel could sit. It is not a big cost, and if you end up fitting nothing, the cables disappear under the plaster and wait. That advice does not change. On a residence it simply stops being enough, because things come with it that cannot be run “just in case”: sensors in the joinery, ordered with the windows; power and routes at the boundary of the plot; a second route for a system meant to be independent of the automation. All of those are design decisions.
Good wireless systems can save a finished house from being chased open, and on an ordinary house they genuinely save the day – you can decide after the fact. It is worth checking the range, the batteries and the supervision of the link, because the differences between systems are large. On a residence that door narrows: the more of the first layer you want and the more metres of it outdoors, the more of it needs a cable and power.
When this is not worth it – and when I advise against it
When a simpler set-up gives the same result. I have said it for years and I do not take it back for bigger houses: even the simplest alarm puts people off, and on a longer trip away it lets you sleep. If the house stands empty three weeks a year and that is the whole problem you have to solve, you do not need three layers of detection. The question only changes once the house is not empty – because then deterrence settles the smaller part of it.
When the alarm answers the wrong problem. Sometimes the real problem lies somewhere else entirely: the plot can be walked into like a park, or the family keeps a routine you can read from the street. No system fixes that, and an elaborate alarm is very good at talking over the conversation. So I start by asking what exactly you want to defend against, and it ends often enough with me advising against half the scope the client came for.
When the alarm is meant to replace a barrier. It will not. Detection says something is happening; it does not slow it down by a second. The fence, the joinery and the structure are there to slow it down. The alarm is there so that somebody finds out.
And the last one: an alarm will not undo decisions that were not taken before the plastering. In a finished house a great deal can be done, and it is done regularly. At a cost that did not have to be paid.
What I do not promise: that the system will catch every scenario; that a local installation is resistant to everything, because it is not, it is only independent; and that I will tell anyone, publicly or privately, what solutions of this kind look like at particular clients.
What I ask before I say yes
- What exactly is this system meant to defend against, and who assessed that besides the company that wants to sell it?
- Can the house be armed and disarmed when the internet is down and the manufacturer’s server is switched off for good?
- Which layers are we doing – boundary, shell, inside – and what happens when the first one speaks up?
- Does the alarm share a bus with the home automation, and if not, at which point do the two meet?
- Who has service access to the alarm and who to the automation, and are those the same people?
- What does notification look like when the power and the network are down, and is the line supervised?
- Who turns up, how do they know it is worth it, and how long does it really take them to reach the gate?
- Which of the household and of the contact list knows what to do when the phone rings at three in the morning?
Frequently asked questions
Can an alarm system without the cloud still be operated from a phone?
Yes. Local means the panel does not need someone else’s server in order to work. It does not have to be cut off from the world. Remote access can be built on a channel that belongs to you and that somebody is answerable for, and then it is an addition: when it goes down, the house still arms, detects and sounds locally. Notification to the outside then needs its own independent path, and that is a separate question to ask. The difference shows up on the day the manufacturer switches the service off – a local system simply does not notice that day.
Does an alarm have to be planned before construction, or can it be added later?
The alarm itself can be added later, wirelessly too, and on an ordinary house that is a sensible route. Before the plastering, though, decisions are made that cannot be reversed without chasing walls: cable routes, sensors ordered with the joinery, power outdoors, and running the alarm separately from the automation. The more of that you want, the earlier the subject has to enter the conversation with your designer.
Are an alarm and a smart home not the same thing these days?
No, and it is as well they are not. Automation is certified on whether the devices talk to each other properly. Resistance to sabotage is tested elsewhere, different things are asked of it, and different people service it. Integration makes every sense, but at one defined point of contact. A shared bus turns the alarm into one participant among many, and that is the difference.
Which security grade should I choose for a house?
I do not settle that in an article and I will not do it over the phone. The grades in the standard describe not so much the equipment as the expected opponent – his knowledge and his tools – so the answer depends on the risk profile, and that is assessed by a security consultant: somebody other than the alarm company, and somebody other than me. I answer for the house being built the way we agreed. Who might come and what for is a question for someone who does that for a living. It is worth knowing, though, that the grade is a shared language for the client, the designer, the installer and the insurer, and that it is declared for the whole installation, rather than read off the brochure of a single device.
How much does an alarm system in a residence cost?
I do not price it by the square metre or by the number of sensors, because the cost comes from something else: how many layers, how much detection outdoors, what response, and whether the installation is to be independent. The biggest difference, though, comes not from the equipment but from the moment of the decision: a system planned before the plastering costs differently from one added afterwards. A concrete conversation about cost starts once we have established what we are defending against.
Does detection at the fence line make sense on a large plot?
It does, provided somebody has worked out what happens once it is triggered, and provided you are ready for false alarms – because outdoors there are more of them and that is how it will stay. The deer, the wind and the growing branch are real participants in this system. Detection at the boundary makes sense when it converts into time that somebody will actually use.
This content is educational and stops at the level of decisions and requirements; execution stays outside the text. Every installation requires its own risk analysis and design. Answers to common questions about privacy and discretion are collected in the Questions section.
If you are thinking about a home where security is designed in from the start, that conversation belongs before the installation design. Let’s talk about your project.