Manned Guarding vs. CCTV: Why Cameras Record and Guards Prevent

A camera and a guard are usually compared as two ways of buying the same thing, and they are not. A camera is a witness: it converts an event into a record. An officer is an intervention: they change what the event becomes. Those are different mechanisms operating at different moments — one after a decision has been made, one before — and treating them as substitutes is the reason so many sites end up with excellent footage of losses they could have avoided.

Deterrence is the clearest illustration. Someone assessing a building for an opportunistic entry is weighing effort against risk. The presence of a person who will see them, address them, and raise an alarm changes that calculation before they commit. A camera changes nothing at that instant; it will faithfully record a decision that has already been taken. Both outcomes are real, but only one of them prevents the loss, and the incidents that never happen are what a security budget is actually buying.

None of which makes cameras a poor purchase. They are frequently excellent value and there are jobs they do far better than any person. The useful framing is not which one wins but which of six functions you actually need at your site: deterrence, detection, verification, intervention, evidence, and coverage of places nobody can stand. Score a camera and an officer against each of the six and the answer for your building stops being a matter of opinion.

Start with the honest case for CCTV, because a site that buys cameras for the right reasons gets a great deal from them. Cameras do not sleep, do not take leave, do not become bored at four in the morning, and do not need to be anywhere in particular. They can watch a stairwell, a bin store, a loading bay, a lift lobby, and four corners of a car park simultaneously — a task that would need several people and would still be worse.

They are also very hard to beat on evidence. A dispute over what happened, who was present, and in what order is resolved by footage in a way no recollection can match — and recollection degrades, while a recording does not. For insurance claims, employment matters, and any incident whose account is later contested, a camera is frequently the most valuable object on the site. It is also the only security control that works retrospectively, which is a genuinely rare property.

Modern analytics have narrowed part of the gap, and it is worth being precise about which part. Perimeter-line crossing, loitering in a defined zone, and vehicle-in-restricted-area alerts genuinely convert a passive recording into an active notification. That is a real capability and it matters. What analytics do not do is decide what to do next — they raise a hand, and something with judgement still has to look. Which features earn their place is the subject of a separate piece on AI analytics in CCTV.

The first is the one already named: presence changes behaviour before an event, not after it. The second is less discussed and arguably larger. An officer makes decisions in ambiguous situations — the delivery that does not match the day's list, the contractor whose story is plausible but unverified, the resident who wants a stranger let up, the argument at the barrier that is thirty seconds from becoming an incident. Every one of those is resolved by judgement applied in the moment, and none of them is visible on a screen as a problem at all.

The third is physical action that limits harm. Closing a fire door that has been propped open. Shutting a gate. Getting people away from a hazard. Meeting an ambulance at the entrance and taking it to the right floor without a delay nobody can afford. Turning off a running tap in a flooding plant room. These are small interventions with disproportionate consequences, and a camera can only document them not happening.

The fourth is the one that never appears in a proposal: an officer is a channel. Residents and staff tell a person they recognise about the door that will not lock, the stranger who was in the corridor, the delivery that felt wrong. Nobody tells a camera anything. Over a year, that informal reporting stream surfaces more genuine problems on most sites than any technical system does — and it depends entirely on the same officer being there long enough to be known.

Between a camera detecting something and anything changing on the ground, there is a chain: the alert has to be seen, interpreted, judged worth acting on, and passed to something capable of acting. Each link is a place the chain breaks, and the breaks are ordinary rather than exotic. The alert arrives on a phone that is face-down. It is the eleventh false alarm this week and is dismissed by reflex. It is seen and believed, and then there is nobody within twenty minutes of the site.

This is why the most useful question about a camera system has nothing to do with the cameras. Who is watching, at what hours, and what are they empowered to do? A system monitored by a control room with a defined response is a different product from an identical system whose feed is watched by nobody until something has already gone wrong. The hardware on the wall is the same in both cases and the price difference sits entirely in the part you cannot see.

False-alarm rate is the part of this chain that decays fastest and is easiest to measure. Every alert that turns out to be a cat, a plastic bag, or a shadow trains whoever receives it to trust the next one slightly less, and after a few weeks the system has taught its own audience to ignore it. Ask any monitored-camera provider what the alert volume per week looks like on a comparable site and what proportion required action. A provider who tracks that number is managing the problem; one who has never counted it is about to make it yours.

An officer on site collapses that entire chain into one step, because detection, verification, judgement, and action all live in the same person standing forty metres away. That is the structural reason a modest camera system supporting a competent officer usually outperforms an excellent camera system supporting nobody. It is not that the technology is worse; it is that the chain is shorter, and every link removed is a failure mode removed.

Camera systems degrade silently, which is their least discussed property. A lens fogs. A cable corrodes. A recorder fills and stops overwriting correctly. A night-vision illuminator fails and the image looks fine at midday. A camera is nudged by a delivery van and now covers a wall. A firmware update changes a retention setting. None of these announce themselves, and the site discovers the state of its system at the exact moment it needs footage.

The counter is unglamorous and cheap: a documented health check on a fixed interval, with someone named to do it. Every camera viewed live, day and night. Recording confirmed present for the full retention window rather than assumed. Timestamps checked for drift, because footage with a wrong clock is considerably weaker in any dispute. On a staffed site this is a natural task for the guarding team, which is one of the ways the two layers pay for each other.

There is a second, subtler decay on the human side and it deserves equal attention. Officers drift too: patrol routes settle into a pattern, visitor checks shorten, log entries get briefer. The difference is that human drift is detectable by a supervisor walking in at two in the morning, whereas a failed camera looks exactly like a working one from the outside. Both layers need a maintenance regime; only one of them can be asked how its night went.

Cameras carry an obligation that guards mostly do not: they record people who have not consented to anything, continuously, and store the result. In a residential building this is the single most common source of conflict between a security committee and the residents it serves, and it is entirely avoidable with rules agreed in advance rather than after a complaint.

Four decisions cover most of it. Where cameras may and may not point — common areas and entrances yes, anything overlooking a private balcony, window, or apartment door no. How long footage is retained, stated as a number of days rather than left to whatever the recorder happens to do. Who can view it, named by role rather than by person, and whether viewing is itself logged. And the process by which a resident may request footage of an incident involving them.

Because rules of this kind have legal as well as practical dimensions, and because the applicable requirements can change, put the policy in front of a lawyer before you publish it to residents rather than after. What is worth saying plainly here is that a written, agreed camera policy is not a bureaucratic tax on the system — it is what stops the system from being switched off by the people it was installed to protect, which is the most common way a camera investment is actually lost.

The practical sequencing matters more than the philosophy. Decide where an officer stands, what they must be able to see, who they must be able to stop, and where they physically cannot be. Those four answers place every camera in the design automatically — they tell you which blind spots need a lens, which approaches need early warning, and which areas need a record because nobody can be there.

Do it in the other order and the result is the pattern visible on sites all over the country: sixteen cameras covering a car park that is already overlooked by four apartments, and nothing at all on the service entrance where every delivery, contractor, and unfamiliar face actually arrives. That layout is not a failure of the installer. It is what happens when coverage is specified by where it is easy to mount a bracket rather than by where a decision has to be made.

The layers should also be wired to each other rather than merely coexisting. The officer should be able to see the camera feeds relevant to their post. Analytics alerts should reach the person who can walk there in ninety seconds. Access-control events should be visible to whoever decides whether to intervene. A site where the guard, the cameras, and the access system each report to a different party has three security systems and no security posture, which is a design problem rather than a budget one — and the subject of electronic security and access control as a discipline.

Most real decisions are made under a ceiling, so the honest question is sequencing rather than preference. As a general rule, if the site has decisions to make — people arriving, deliveries, contractors, disputes, an entrance that has to be controlled — the person comes first, because a decision needs someone to make it and no amount of coverage substitutes. Cameras then extend that person's reach into the places they cannot stand.

The inversion is equally honest. If the site has no decisions to make — a locked, empty warehouse at night, a remote yard nobody visits, a plant room — then cameras with a monitored, verified alarm response are frequently the better buy, and paying a person to sit in an empty building is paying for reassurance rather than protection. The two cases genuinely differ, and a provider who recommends the same answer for both is not assessing your site.

The layered answer is not a compromise between the two; it is what a designed system looks like. Cameras record and extend, access control decides who reaches which door, alarms raise a hand, and a trained officer supplies the judgement and the physical action that none of the others can. Specify in that order, connect the layers to each other, and maintain both on a fixed rhythm. If you are working out which mix your building needs, the guarding and CCTV service pages set out what each layer actually covers.