Why Your Security Cameras Are Not Helping When Something Actually Happens

Something happens. A door forced overnight, a pallet gone from the yard, a car keyed in the lot. You go to the recorder, find the right time, and the footage is useless.

It is almost never because the cameras were off. It is because of decisions made on the day they went in, by somebody thinking about coverage rather than evidence. Those are not the same goal, and the gap between them only shows up on the day you need the video.

What coverage gets you, and what it does not

A camera that can see a doorway is covering that doorway. Whether it can tell you who walked through it is a separate question, and the answer comes down to how many pixels land on a face.

There is a standard for this. IEC 62676-4 grades a camera view by pixel density, the number of pixels spread across a metre of the scene at the distance you care about. Its 2025 revision puts the level where face-recognition software can reliably work at around 500 pixels per metre. You do not need the number. You need the principle underneath it: the wider you point a camera, the fewer pixels land on any one person, and past a certain distance you are recording that somebody was there rather than who they were.

Which is why 4K on the box settles nothing. A 4K camera aimed across a whole parking lot spreads those pixels thin. The same camera aimed at one doorway is excellent.

Five reasons the footage fails

The camera is mounted too high

The instinct is to put cameras out of reach. Higher is harder to tamper with and harder to knock out of alignment. It is also a worse angle for faces, because the steeper the view, the more of it is hat, hood and shoulders.

The fix is not to drop every camera to head height where it will get covered or sprayed. It is to stop asking one camera to do two jobs. A high camera gives you the overview of a space. A second one at a chokepoint, near head height and looking along the direction people walk, gives you the face.

Nothing is watching the chokepoints

Most buildings have three or four places everybody has to pass through. The customer entrance, the staff door, the corridor to the back office, the yard gate. Those views are worth more than the open floor.

Walk the route somebody would actually take, from the boundary to whatever is worth stealing, and count how many identification-grade views they pass through on the way. In a lot of buildings the honest answer is none, because the cameras were placed to look at valuable things, and nobody mapped the ways in.

The doorway is backlit

Point a camera at a glass entrance in daylight and you have aimed it at the brightest thing in the building. The sensor exposes for that bright background and everybody walking in becomes a silhouette. It looks fine on the monitor at two in the afternoon in December and fails completely in June.

Three ways out of it: angle the camera along the wall so the bright glass sits off to one side, switch on the camera’s wide dynamic range if it has any, or put more light inside so the gap between indoors and outdoors narrows. Check the result in the worst hour of the day, never at midday.

The night performance was never tested

Infrared range on a spec sheet is measured against a reflective target in ideal conditions. Over wet asphalt in a real yard, that figure drops a long way. Colour night modes need genuine light on the scene and give you noise without it.

You can test this without any equipment. Walk the site at the darkest hour that actually matters for your building and watch the live view on your phone while somebody stands where an intruder would.

The footage was already gone

This is the one that catches most people. Recorders overwrite the oldest footage to make room for new. How far back you can go depends on the number of cameras, their resolution, the frame rate, the size of the disk, and whether they record constantly or only on motion.

Plenty of commercial systems hold a week or less. A theft discovered on the Monday of a long weekend can already be over the edge. Worse, a disk that has been failing for months may be holding a fraction of what it was specified to hold, and nothing on the front end announces that.

A twenty minute self-audit

Set aside twenty minutes. You need access to the recorder and somebody willing to walk around the building for you.

• Have them stand at each entrance, facing the camera, while you watch the live view. Could you pick that person out of a line-up if you did not already know them? If not, the view is coverage, not identification.
• Note the height and angle of every camera pointed at a door. Anything looking steeply down on a doorway is handing you the top of a head.
• Look at those same doorway views twice, once at the brightest part of the day and once after dark. A silhouette or a wash of grey noise is a failed view.
• Find the retention setting, then go and open the oldest footage you supposedly still hold. Open it. Do not trust the setting.
• Check the recorder’s disk health page. A drive reporting errors is a retention problem that has not surfaced yet.
• Count the cameras that are offline or sitting on a frozen frame. On a system nobody audits, that number is rarely zero.

Write down what failed. That list is the brief for fixing it, and it is worth more than any brochure.

What to fix first

Order matters, because the budget runs out before the wish-list does.

Retention comes first. It is usually the cheapest thing to change and it decides whether anything else you record is still there when you go looking. Chokepoints are next, one identification-grade view on each way in. Then lighting, on whichever views failed after dark. Replacing cameras wholesale comes last, and often turns out to be unnecessary: a large share of the systems we get called to look at have adequate cameras in the wrong places.

If you want the detail on what we install and how a camera gets specified for a particular view, that is on our security cameras page. If the cameras are really part of a bigger question, and the alarm, the door readers and the monitoring all have to work together, start with commercial security systems.

Or call 719-888-7180 and we will come and walk it with you.

Frequently asked questions

How many cameras does a commercial building need?

It is the wrong question, and it is the one most quotes answer. Two cameras on the right chokepoints beat eight spread evenly around a building. Start from the routes somebody would take through your site, then work out how many identification-grade views those routes need.

Is a 4K camera enough on its own?

Resolution only helps if the pixels land where you need them. A 4K camera covering a wide yard can easily put fewer pixels on a face than a 1080p camera aimed at one door. Field of view and mounting position decide what the resolution is worth.

How long should we keep footage?

Long enough to survive the gap between something happening and somebody noticing. For most businesses that means covering a long weekend with room to spare, so thirty days is a common target. Confirm what your recorder actually holds. The number it was sold on is not the number you have.

Can our existing cameras be moved instead of replaced?

Often, yes. Repositioning, re-aiming and changing the lens or field of view fixes a good proportion of the problems in this post at a fraction of the cost of new hardware. Cameras genuinely worth replacing are usually the ones with no wide dynamic range on a backlit entrance, or analogue units past any useful resolution.

What do the police need from our footage?

Ask your recorder two things before you ever need them: can it export a specific clip to a file somebody else can open, and is its clock right. An export locked to a proprietary player, or a timestamp forty minutes out, makes good footage far less useful than it should be.