A SIM card does not upload the video. It saves everything else.

It makes routers that carry a mobile network modem inside. There are desktop formats and there are rail formats to go inside a panel, with direct current power and no fan, and some with two card slots so you can have two operators and switch to the second when the first goes down.

What it brings to security projects is a specific combination: reasonable price, industrial format and remote management that really works over mobile networks, which is where remote connections get complicated. And it is used more than people say: almost any spread-out installation ends up having a point that nobody's line reaches —a hut, a plot, a car park, a site—, and there is no brand debate there.

What it does, and very well: management and alerts. Reaching the recorder from outside to look at a configuration or pull down a piece of recording. Getting the alarm out to the alarm receiving centre even when the fixed line is dead. Sending the alert for an event with an image, or with a few seconds of video, instead of the whole stream. Keeping a barrier, a door entry unit or a point with two cameras online where the video stays recording right there. In all those cases the mobile network does a job nothing else can do.

What it does not do, almost ever: upload the whole video. And it is not the device's fault. The upload direction —from your installation outwards— is the small one, and the video goes in exactly that direction. The medium is shared with everybody in that cell, so what gets measured on a Tuesday at eleven is not what there will be on a Friday at seven. And cameras do not send in bursts: they send the same throughput every hour, before dawn too and in August too.

On top of that comes the money, which is the part nobody budgets for. A sustained stream against a mobile tariff eats any data plan in days, and the worst is not the expected case: it is the day the backup switches on by itself because the fixed line went down, the cameras carry on sending as if nothing had happened and a month's usage goes in one night. That is avoided with rules written into the router, not by trusting that it will not happen.

There are nuances and they get said: with one or two cameras, at reduced quality, sending only when there is an event and with good coverage, the mobile network can carry video. What it does not do is replace a line for a real installation. If somebody has told you it does, ask them for the sum.

On a fixed business line it is normal to have a public address: a number that identifies your installation on the internet and that can be called from outside. On an ordinary mobile card that does not happen: the operator gives a private address, inside its own network and shared with many more people. Your router can get out without any trouble, but from outside there is no way of calling it. There is no port to open, and it is exactly the moment when somebody proposes publishing the camera on the internet with its factory password.

There are three ways out and it is worth deciding which one before buying anything. Asking the operator for a card with a fixed, reachable address, which exists and is paid for. Having the router be the one that calls outwards and sets up an encrypted tunnel to a server, so that it starts the connection and nobody has to call it: that is what we do almost always. Or the manufacturer's own cloud management, which does something similar and works very well for a fleet spread around.

When this is not decided beforehand the same thing always happens: the card is in the quotation, the access is not, and on commissioning day it turns out the recorder cannot be reached from the alarm receiving centre.

The coverage is measured where the antenna is going to be, not where the technician is standing with the phone. And the antenna usually ends up inside a metal cabinet, which is about the worst thing you can do to a radio signal, or in a basement, or behind a reinforced concrete wall.

The usual solution is an outdoor antenna on the façade or on the roof, with the cable running down to the cabinet. And there is the detail that gets forgotten: that cable loses too, and it loses more the longer and the thinner it is. A badly chosen run eats up the gain from having put the antenna up there. It is calculated, the cable is chosen and it is measured after fitting it.

Two more things we always look at: which operator has coverage at that exact point —it is tested there, with different cards and at different hours, not on a map— and whether the place fills with people at certain moments, because that is where the capacity collapses on exactly the days that matter.

It is configured from the browser, and the configuration can be exported to a file and loaded back, which is what turns commissioning twelve identical points into doing one and copying. The manufacturer's cloud management can be done without on a single point and, on fifteen, it is the difference between maintaining them and forgetting them.

What we always configure and almost nobody asks for: the usage alert before the limit is reached, the rules on what can go out over the mobile network, and the backup with its conditions for coming back. A backup that switches on by itself and does not come back by itself when the fixed line is repaired is a surprise bill at the end of the month.

It fits as backup for the fixed line in any installation where raising the alert is the critical thing: while the operator repairs its fault, the alarm still gets out and you can still get in to look. It fits at isolated points where the video is recorded right there and only alerts and the odd query travel over the mobile network. It fits in temporary installations —a site, an event, a plot while the line is being arranged—. And it fits for keeping online devices that need little: a barrier, a door entry unit, an alarm panel.

We would not put it as the main path for the video, for everything the second section says. We would not put it to save a trench that can in fact be dug: a cable costs once and a tariff costs every month, and the cable does not fill with people on a Friday afternoon. And we would not put it without having measured at the exact point and with the particular operator, because a mobile link that runs tight on the day it is installed is never going to get any better.

An IT team runs this without help: the console is simple and the logic is that of any router. Changing a card, looking at usage, restarting remotely, updating, adding a rule: normal work. We hand over their own account, the configuration in a file and the inventory of which card is at which point and on which contract, which is the piece of information that is always missing when it is time to renew.

The first caution is the design of the tunnel and of who can reach what: it is the piece through which the installation is entered from outside, and a change made in a hurry there does not leave the network down, it leaves the network open. The second is the backup: when it switches on, what it is allowed to pass and when it comes back is the balance between still raising alerts and not spending a year of data in one night, and it is handed over in writing.

And if your IT team already has its router brand and its way of building tunnels, we adapt without arguing: the particular device is the least important thing on this page. What matters to us is that the path your house opens allows what the security installation needs —alarms getting out with priority and controlled management access— and that the usage rules are in place. If that is met with what you already have, fine.