Continuous or motion-only recordingthe trade-off that mostly stopped existing

There is a clean way to think about this and it removes all the guesswork. The storage saving from motion-only recording is the inverse of the motion duty cycle. If something is moving in shot a fifth of the time, you store about a fifth as much, so your retention stretches roughly five times. If something is moving half the time, you halve it.

Two published worked examples bracket the sensible range. One video management vendor sizes its own storage guidance on motion occurring 20 per cent of the time, which it says reduces load and wear on the storage system by 80 per cent. An independent industry recording guide works its example at a 33 per cent motion factor, needing roughly one third of the storage. Both are design assumptions rather than measurements, and neither is a promise about your driveway.

Which leads to the part nobody puts on a sales page. The same industry guide notes that in constantly busy scenes the savings from motion recording are minimised. Read that with the formula and the conclusion is uncomfortable: motion-only pays best on the camera watching an empty side return, and pays almost nothing on the camera watching the front of the house. You are being sold a saving that shrinks in proportion to how much you needed the camera.

You can see what continuous recording costs on your own setup with our CCTV storage calculator, and our guide to how long CCTV footage lasts covers the other three things that set your retention.

This is not a criticism from outside the industry. Camera manufacturers document these limits openly, in the technical pages installers are supposed to read and homeowners never see.

• Objects approaching in a straight line towards the camera need to move a longer distance before they are detected than objects crossing the view. That is precisely the path someone takes walking up to your front door.
• Standard AI detection does not detect stationary people or vehicles. Someone who stops and stands still, or a car that parks up, stops producing events.
• There is a minimum size. One manufacturer gives a human minimum of 4 per cent of image height, then recommends at least 8 per cent to minimise missed detections. The gap between the stated minimum and the recommended figure is the manufacturer telling you the minimum is not safe.
• Detection needs people roughly upright, and accuracy drops when objects are only partially visible or the scene is crowded with overlapping objects. Crouching, crawling or moving low behind a wall or a parked car is a documented miss.
• Small and distant objects might not be detected at all.
• Heavy rain or snow reduces both detection distance and accuracy.
• Excessive camera vibration causes false alarms rather than missed ones.

None of this means detection is bad. It means detection is a filter, and every filter has a pass band. The question is whether you want your evidence decided by one.

The whole continuous-versus-motion debate assumes that recording a still scene costs the same as recording a busy one. On a modern camera it does not, and that is what has changed.

Smart codecs watch the scene and compress it according to what is happening in it. When nothing is moving, the camera sends the bandwidth-heavy reference frames far less often, and in a completely static scene the effective frame rate between those frames drops to almost nothing, snapping back the instant motion appears. You are still recording continuously. You are simply not spending much to record nothing.

The measured effect is large. One manufacturer publishes per-scene reductions for constant recording ranging from 25 per cent on a well-lit scene with occasional motion, through 73 and 85 and 90 per cent on various real scenes, up to 99.7 per cent on a night-mode scene with very infrequent motion. Those are vendor figures with vendor caveats, and the reduction varies with light, movement and scene detail. Even discounted heavily, they change the decision.

The same manufacturer states the conclusion outright in its own white paper: the technology enables continuous recordings because of the low bitrate used for static scenes, and it reduces the number of missed triggers by allowing longer recording segments for each motion-triggered event without generating excessive data. When the company selling the compression tells you it makes continuous recording viable, the storage argument for motion-only has largely gone.

There is a wrinkle worth knowing if someone quotes you a storage figure. Manufacturer guidance warns that bitrate limits are often set to a very low default, that most storage calculators base their sums on that low limit, and that systems get built with insufficient storage for high quality recording as a result. A quoted retention figure is only as good as the bitrate assumption under it.

No national UK law requires CCTV at all, and none specifies a recording mode. Where a continuous obligation exists, it exists because a licensing authority attached it as a condition to an individual premises licence.

It genuinely varies. Westminster's model conditions require that the CCTV system shall continually record whilst the premises is open for licensable activities and during all times when customers remain on the premises, with recordings stored for a minimum of 31 days. The City of London uses near-identical wording. Essex Police's model CCTV condition, by contrast, requires a recordable system of evidential quality, specified coverage, correct time and date stamping and 31 days of retention, and says nothing at all about recording continually.

All three publish the same warning about their own documents, which is the real lesson here. Westminster says its conditions are not standard conditions and should be used as a guide. Essex Police says they are not standard conditions it seeks to have applied to each premises of a certain type. So if you run a licensed premises, the answer is on your licence, not in an article.

On insurance, be sceptical. We looked for an insurer document, a policy wording or a trade body standard requiring continuous recording and did not find one. Some commercial policies require a working CCTV system as a condition of cover, which is a different thing. If someone tells you your insurer demands continuous recording, ask them to point at the clause.

Pulling the other way, data protection law asks you to record no more than you need, and to keep it no longer than necessary. That genuinely favours motion-only in some settings. It sits against the evidential case, and the thing that resolves the two is the purpose you installed the cameras for, which is also what the regulator says should set it.