CCTV Fallback Plans: Recording Failure vs the Detection Gap

1. Two different failure modes, usually conflated

When people ask whether their CCTV has a fallback plan, they usually mean: what happens if the recorder fails to save a clip? That is one failure mode. It is real, it happens, and it is worth planning for. It is also the less common of the two.

The other failure mode is more subtle and more frequent. Footage is recorded perfectly. The drive is healthy, the cameras are online, the clock is synced. No alert goes off because there is no alert system. Nobody watches the monitor wall after the first twenty minutes of a shift. Then something happens at 2:14 am. Two weeks later, somebody asks for the clip. The DVR has already overwritten it. No fallback plan in the world saves a clip that was overwritten three days ago.

These two failure modes need two different solutions. Recording redundancy is the answer to the first. Real time detection, plus a documented review protocol, is the answer to the second. Most CCTV operations have one but not both.

2. What recording redundancy actually covers

Recording redundancy is the layer that protects you against a drive dying, a DVR being power cycled mid write, or a camera dropping a feed for a few minutes. The standard toolkit includes:

• Dual drive RAID 1 in the DVR. Two drives writing the same data. A single drive failure is survived. This is the cheapest, easiest fix, and it is still not standard on most installed DVRs.
• UPS with proper runtime. A 10 to 20 minute UPS prevents the most common cause of drive corruption, which is an unclean shutdown during a write cycle. Most surveillance drives will tolerate many power events, but not indefinitely.
• Cloud or off site mirror for critical channels. The two or three cameras that really matter (main entry, parking lot entrance, server room) can be mirrored to an off site recorder or a cloud bucket. This covers the case where the entire DVR is destroyed in a fire or taken in a theft.
• Retention audit. Confirm that the DVR is actually keeping footage for the number of days you think it is. The number in the settings menu and the actual retained days are often different, because resolution or channel count changes eat through storage faster than the documented estimate.
• Camera health monitoring. A camera that has been offline for six weeks is not caught by any recording plan, because the recorder does not see it as a failure. It sees a silent channel. A monthly audit, or a monitoring tool that alerts on stream loss, catches this before the incident arrives.

3. The detection gap: footage nobody looked at

The detection gap is what happens between an event and a human seeing it. In a typical property, that gap is measured in days. The event happens at 2:14 am. The next morning, nobody knows. Maybe a resident mentions something to the front desk at 4 pm. Somebody pulls up footage and scrubs for 90 minutes. Maybe they find it, maybe they do not. Meanwhile the overwrite window is ticking.

The gap is not a personnel failure. It is a workload failure. A human watching a 24 tile monitor wall catches maybe one in ten events after the first twenty minutes. It gets worse over an eight hour shift. This is well documented in every study of guard monitoring, and it is the reason real time detection exists.

Real time detection closes the gap by changing the failure mode from passive (nobody noticed) to active (an alert went out, and the team decided whether to act). Either outcome is better than the alternative, because the team now has a record of the decision.

4. A plan that covers both

A fallback plan that covers recording and detection has four layers:

• Layer 1, recording redundancy. Dual drive RAID 1 in the DVR. UPS with 15 minute runtime. Critical channels mirrored off site for 30 days. Monthly health audit.
• Layer 2, real time detection. A detection layer that watches the feed in real time and triggers alerts on the events you care about: loitering, restricted zone entry, package handling, crowd formation, unauthorized vehicle. Alerts go to a messaging channel the team actually reads.
• Layer 3, documented review protocol. A written protocol for what to do when an alert comes in. Who acknowledges it, what the acceptable response time is, how the event is logged, who escalates if needed. Without this, the alert layer becomes noise.
• Layer 4, quarterly fire drill. Four times a year, the team verifies the plan works. Unplug a camera, note how fast it is caught. Pull a clip from 36 hours ago, note how fast it is produced. Fire a simulated alert, note who responded. Update the protocol based on what failed.

5. How to fire drill your plan

The plan that has never been tested is not a plan. Most properties have never run a fallback drill on their CCTV, and they find out what works when an incident happens. That is an expensive way to learn.

A quarterly drill takes about 30 minutes and covers three things:

• Camera loss test. Unplug one camera during business hours. Time how long until the team notices. If the answer is more than 24 hours, your camera health monitoring is not working.
• Clip retrieval test. Pick a camera and a 5 minute window from 36 hours ago. Ask the team to produce an MP4 export. Time how long it takes. If the answer is more than 15 minutes, your retrieval workflow is too slow for real incidents.
• Detection response test. Walk into a restricted zone during business hours. Time how long until an alert fires, and how long until someone responds. If the answer is measured in hours, your detection layer is decorative.

Log the results. Improvement comes from the gap between what the plan says and what the drill measures.

6. Tooling options, honestly compared

There are several ways to add the detection layer without replacing the DVR or the cameras.

• Cloud VMS with AI analytics. Replace the DVR, replace the cameras, move to a cloud platform. Highest ceiling, highest cost. Good option on a major facility refresh, bad option when the existing infrastructure still works.
• On prem server pulling RTSP or ONVIF streams. Works when the DVR exposes streams cleanly. In practice, many hybrid DVRs from 2012 to 2017 either do not expose RTSP or require credentials that were lost years ago.
• Outsourced guard monitoring service. A human watches your feed for a monthly fee, usually $250 to $600 per camera. Works well on a small number of critical cameras. Does not scale to a 24 camera property without a large monthly bill.
• Edge AI device on the DVR's HDMI output. A small box sits between the DVR and the guard monitor, captures the multiview grid, splits it into tiles, and runs detection on each tile. No camera replacement, no RTSP dependency, no cloud required. Cyrano is one example, at $450 up front and $200 per month. It is not a full VMS replacement, but it closes the detection gap at a fraction of the cost.

The right choice is the one that fits your budget, your infrastructure age, and your operational maturity. The wrong choice is leaving the gap open because the full refresh is too expensive to plan right now.

7. FAQ