A company's philosophy is not declared. It is checked.

The text on this page until now said that our values are commitment, integrity, respect and responsibility, among others. It is not a lie. The problem is that it is not information either: that same sentence sits, word for word, on the website of the company that installs badly and on the website of the one that installs well, which is why it is no use for telling them apart.

What separates two companies is not the values they pick, because everybody picks the same ones. It is the decisions they take on the day the value costs money: when you have to say the project cannot be done the way the client has asked for it, or turn down a contract because it is going to go wrong.

So this is a list of decisions, not of adjectives. There are eight, and they are all verified the same way: ask about one, ask for a case, and see whether the answer comes back with the names of things inside it —a camera, a height, a date— or goes off to the big words again.

A drawing tells you where the walls are. It does not tell you that at six in the evening the sun comes straight in through that window and the camera spends forty minutes seeing nothing, or that the planned column has a lorry parked over it, or that cats cross the fence at the back all night and that is six alerts a shift. It only shows up if somebody goes and looks.

That is why the first step is always the same, and it costs nothing: a visit, or to start with a single image from a camera you already have. That is where the first answer comes from. Sometimes it is yes and here is how; sometimes it is that the camera is not good enough for what you want and it has to be moved; and sometimes it is that what is needed is not ours.

The consequence is that we take longer to give you a number than somebody who gives it over the phone. A price given before looking is a price that gets corrected on site, and correcting it then is almost always paid for by the client.

Every project has a list of things that will not be possible: that camera will not read number plates from there, that corridor cannot be covered with one detector and needs two, that area needs light at night and there is none, with that bandwidth you cannot pull live video off all forty at once. That list always exists. The only difference between suppliers is when it turns up.

If it turns up before signing, it is a decision: you change the scope, you change the quotation, or you accept the limit knowing it is there. If it turns up at commissioning, it is a problem and somebody has to be found to carry the blame. And if it turns up on the day of the incident, it is no longer a technical problem: it is the difference between having the image that was needed and not having it.

That is why the limits go in the project document, with names: what each camera covers, what it does not cover, what can be seen at night and in the rain, and what has to be looked at again in a year. Reading it is uncomfortable and it is the most useful part of what we hand over.

A system that looks at images gets things wrong. It gets them wrong more or less depending on where the camera is, how the light is, what is in front of it and how the rule is tuned, but it gets them wrong. Anyone who tells you theirs never fails has either never put one into production or is telling you a laboratory result as if it were the result in your unit.

What you can do is measure: how many alerts came out, how many were real, at what times it failed and why. That is what you tune with, and the conversation stops being whether the system is good and becomes whether, in your scene, it is already where it needs to be.

And when something is measured by somebody from outside, we show it. People counting is approved by the Centro Español de Metrología: it is stated by a body that does not sell the product, and the exact scope of that approval is written down. When something is not measured by a third party we say that too, and then the only honest thing is a small trial on your installation.

A security system is not bought once: it is maintained for years. On handover day everything works, and that tells you nothing — what decides whether it is still useful three years later is whether somebody inside uses it, checks it and changes it when the work changes. The supplier cannot hold that part up, however good they are.

So we ask for it explicitly, and it is the only thing we ask for: a person with a name who decides what is watched and who gets alerted, who lets us in to inspect when it is due and who tells us the month the operation changes, not two years later.

When there is nobody, the ending is always the same: the alerts go to an inbox nobody opens, somebody silences them so they stop being a nuisance, the following year the system is a line of spending nobody defends in the committee, and at renewal it is swapped for a cheaper one that will end up exactly the same way.

Saying it in the first meeting saves both of us time. And when the answer is that we are not the right people, it is said right there, with the reason.

When a security installation is bought on the price per camera, what is being compared is the equipment, which is the part that costs the same everywhere because the same manufacturer makes it. What does not show up in that figure is the work that makes the equipment useful: the project done beforehand, the tuning done afterwards and the maintenance done for years.

That work can be taken out of the quotation. It comes out cheaper and the installation switches on just the same. The difference is not visible on day one: it is visible the day an image has to be recovered and it turns out that area was not covered, or that the recording days everyone assumed were not real, or that the annual inspection was a one-line report saying «installation correct».

We do not do that version. If the budget has to come in at a fixed number, we reduce the scope —fewer cameras, well placed— and not the work on the ones that are left.

The normal thing in this trade is for those to be two companies, each answering for its own half. The day the system does not see what it was supposed to see, both are right: the installer says the camera meets the specification and the software company says that with that image you cannot do what is being asked. Both are true and the problem is still there.

When they are under one roof, the order of the questions changes. Before a camera position is chosen, what will have to be understood from there is already on the table: if one day a number plate has to be read, people counted or a hard hat checked, that governs the height, the angle and the lens, and it comes out at the site visit.

What you gain is not speed: it is not having to go back. Ten degrees of tilt decided properly on day one are free. Discovered two years later, with the work done and the suspended ceiling closed up, they are a new job.

There is a very widespread way of tying a client in, and it is to leave nothing written down. The passwords are held by the technician who built it, the network diagram is in his head, the rules were configured on a Tuesday and nobody wrote down why, and the drawing of which camera covers what does not exist. The client is not satisfied: the client is trapped, which is a different thing and lasts less.

We hand over the opposite. The diagram of what is there and how it is connected. Which camera covers which area and with what limits. The active rules, what they trigger and who they alert. The retention periods configured and how much they take up. The users and what each of them can do. And the passwords, which are yours.

That means you can change supplier without starting from scratch, and yes, that you can change us. It is the right price to pay: we would rather you stayed because the system works than because leaving is too expensive. And the day an inspection, an insurer or an audit asks for that documentation, it exists.

We sell systems that look at people and draw conclusions from what they see, and that has real consequences: an alert can send somebody to have a word with a worker, keep a door shut or put a recording into a personnel file. Treating it as a technical detail is what makes people distrust this technology, and rightly so.

The system detects, measures and alerts. It does not sanction, it does not accuse and it does not decide anything about anybody. What arrives at the other end is a situation with its image, its time and its context, and the person who decides what is done is a person with the responsibility to do it.

It is not a legal precaution: it is how the systems are built. When a client asks for a detection to trigger a consequence for somebody on its own —a report opened, a misconduct entry recorded— that is where the conversation stops and we talk about who validates it and on what criteria. Because the system is going to get it wrong at some point, and if there is nobody in between, the error goes through whole.

In a demonstration it is very easy to show what the system does well. What decides whether a project turns out well is the other thing: under what conditions it fails, what it confuses with what, what happens to it in fog, in backlight, with a camera too far away or with a field of view that is not the one it needs.

So that goes first and it goes in the proposal. If a detection only holds up within a range of distances, the range is stated. If it needs the camera positioned differently, that is said before it is sold. And if what is being asked for is not solved today, that is said.

A success rate that does not say in which scene, with which camera, over how many hours of video and against what it was compared, is not a figure: it is a decoration. And it is a dangerous decoration, because the client uses it to decide and afterwards the number looks nothing like the one that comes out of their installation.

Our way of giving a number is always the same: measured in your scene, over a period, with the camera in the position it is going to stay in. It takes longer than showing a catalogue figure and it is the only one that holds up afterwards. And when a measurement has been made by a third party, it is quoted with its exact scope and no further.

A detection needs to see in order to work, but it almost never needs to keep everything it sees. Counting how many people go along a corridor does not require knowing who they were. Checking that somebody is wearing a hard hat does not require a named record of who was not. Measuring how long a lorry spends at the loading bay does not require identifying the driver.

The rule is that what is needed for the function requested is processed and kept, and that the rest is not stored. It is decided at the design stage, one by one, and it is a data protection conversation as much as a technical one: that is why it has a page of its own.

If a head of security cannot explain to their board why the system raised an alert, the system is no use for deciding anything. So the rule has to be tellable in one sentence: «if somebody enters this area between these hours, alert this shift». And the alert arrives with the image, the time, the camera and the rule that triggered it, not with «an anomaly has been detected», which is what makes nobody trust it by the third time.

The teams that develop and the teams that operate these systems are trained continuously in the privacy practices and standards that apply to them. It is a commitment that was already in the previous text on this page and it stays, because it is the one holding up all the others: a well designed and badly operated system ends up doing exactly what had been decided against.

And the client's training too: training the people who will watch the screens and handle the alerts is part of the project, not something left to a PDF manual.

This took up half a page in the previous text, with the biggest words available. Underneath there were specific things, so they stay and they are told without those words: they are written commitments, most of them verifiable in a document you can ask for.

The commitments to the team are in writing and these are they: a safe working environment with risk prevention measures kept up to date, professional development with equal opportunities, fair pay and permanent contracts wherever possible, support for balancing work and family life, no discrimination on any personal, physical or social grounds, eradication of forced and child labour and respect for the rights of minorities.

Equality is built in as a policy, with training for staff and flexibility for balancing work and family life. And there is a practical reason as well as the right one: a team that looks too much like itself takes worse decisions about a system that is going to be looking at all kinds of people.

Workplace safety, on top of that, is certified: ISO 45001 requires this to be a system with records rather than a declaration, and requires workers to take part. Somebody from outside checks it.

The small part counts and gets done: saving water, recycling paper and plastic, and improving the premises so they use less. It is not spectacular and it is what is in our hands every day.

The less small part: planting trees to offset the emissions of the business, and the commitment that the energy our products depend on comes from renewable sources. The previous text on this page put a percentage there; here it goes without one, because a figure like that is evidenced with a supply contract and not with a paragraph on a website.

Local suppliers are given priority where possible. And anyone taking part in choosing a supplier has an obligation to act impartially, applying criteria of quality and cost, so that their personal interest does not come into the decision. It is in the code of conduct and it is the line most often crossed in this industry.

In the product it shows in one thing: we work with several international manufacturers and we are not married to any of them. When an installer only proposes one brand, it is worth asking whether it is the best one for that case or the one that gives them margin.

None of the above is worth anything if it cannot be verified, so there are three ways. One: ask for the certificates. They are audited by bodies that are not us, and the scope and the validity are on the document those bodies issue.

Two: put a problem like yours to us and listen for whether the answer has a camera, a height and a date inside it, or stays general. Three: send an image from one of your cameras and see what we reply. It costs nothing and it is the quickest test.