A loss at a warehouse is discovered on Monday, and it apparently happened ten days earlier. The recorder keeps seven days of footage. The cameras captured everything — and overwrote everything before anyone went looking. This is one of the most common CCTV failures, and it has nothing to do with the cameras. It is a storage calculation that was done wrong.
Archive size is not a question of how many disks to buy. It is the result of four decisions: what quality you record, how many hours a day, how many days you keep, and how many cameras. Below is the formula, plus the factors most often overlooked when designing a video surveillance system.
The formula that explains everything
Storage (GB) = bitrate (Mbit/s) × recording seconds per day × number of days × number of cameras ÷ 8 ÷ 1000. Dividing by eight converts bits to bytes; dividing by a thousand converts megabytes to gigabytes. There are 86,400 seconds in a day, so one camera recording continuously at 4 Mbit/s uses about 43 GB per day.
For example, 30 cameras at 4 Mbit/s each, continuous recording, 30 days of retention: 43.2 GB × 30 cameras × 30 days ≈ 39 TB of usable capacity. That is the net figure — RAID overhead and free-space headroom come on top, as discussed below. If the same system moves from 30 to 60 days of retention, storage doubles exactly — the relationship is linear.
What drives bitrate
Bitrate is the most variable part of the formula and the one most often underestimated. Manufacturer calculators often assume an ideal, quiet scene. In reality it depends on:
- Resolution: a 4K camera produces considerably more data than a 1080p one. High resolution is justified only where you genuinely need detail — to identify a face or a number plate, or to cover a wide area.
- Frame rate: 10–15 frames per second is often enough for general observation; 25–30 is needed for fast movement, such as at a checkout or on a road.
- Codec: H.265 typically needs noticeably less bitrate than H.264 at the same quality; manufacturers' smart codecs reduce volume further on static scenes.
- Scene complexity: a crowded hall, trees in the wind, rain and moving traffic all push bitrate up.
- Night mode: in low light, image noise sharply increases bitrate — night footage is often heavier than daytime footage.
- Bitrate mode: variable bitrate (VBR) adapts to the scene; constant bitrate (CBR) is predictable but wastes space on quiet scenes.
The most reliable approach: run one or two cameras at the real location for a few days and measure average bitrate across day and night. That beats any calculator. If that is not possible, add generous headroom to the calculator figure — overestimating is far cheaper than an archive that turns out too short when it matters.
How long should footage be kept?
Retention is a business and legal decision, not a technical one. The key question is: how much time usually passes between an incident and its discovery? In some cases — a shortfall found during a stock count in retail or a warehouse, for example — that gap can be weeks. Retention should cover that gap, with margin. Also allow for the time it takes to find and export footage — plan retention so that there is still time left to investigate.
On the other hand, longer is not automatically better. Footage showing people is personal data. The retention period should be set by your internal policy and the applicable personal data protection requirements, and keeping footage longer than necessary can itself be a compliance risk. Taking legal advice on this point is sensible.
Continuous or event-based recording
Recording on motion or events saves significant storage where the scene is still most of the time: a corridor at night, a storeroom, an office closed at weekends. But it carries risks: a poorly tuned detector misses slow or distant movement, while trees, shadows and headlights create false recordings. A common compromise is dual streaming: continuous recording at lower quality plus high-quality recording on events.
When sizing an event-based system, replace recording seconds per day in the formula with a realistic activity percentage — and estimate it conservatively, especially for outdoor cameras.
Storage architecture: RAID, disks and headroom
Converting net capacity into physical disks needs a few corrections. RAID 5 uses one disk's worth of capacity for parity, RAID 6 uses two; RAID 6 is safer for arrays of large disks because it survives a second disk failure during the long rebuild of the first. Add free-space headroom and, ideally, a hot-spare disk. Use disks designed for continuous surveillance recording, not ordinary desktop models.
The network: the part people forget
The same bitrate you use to size storage also loads the network. The stream into the recorder is the sum of all camera bitrates: 30 cameras at 4 Mbit/s continuous is about 120 Mbit/s, and more with dual streams and remote viewing. Switch uplinks must carry this, and camera switches need an adequate PoE budget. CCTV is usually placed in its own virtual network — a security and traffic-management question that should be part of the network infrastructure design.
Before you order a recorder or server
- Define each camera's purpose — overview, detection or identification — and set resolution and frame rate accordingly.
- Measure or conservatively estimate average bitrate in day and night conditions.
- Agree the retention period with management and, where needed, legal counsel.
- Choose the recording mode for each camera: continuous, event-based or combined.
- Calculate net capacity with the formula, then add RAID overhead, free-space headroom and growth for new cameras.
- Check network bandwidth and PoE budget.
- Decide who exports important footage, how, and where it is stored.
A few weeks after go-live, check how many days are actually being retained. If the archive is shorter than planned, you want to find out now — not after an incident.
If you are planning a CCTV system or your current archive is running short, write to us — we will check the calculation together and find a sensible way forward.