Your operator has been staring at the same screen for 22 minutes. He’s already missed 95%.

The most-cited study on human CCTV surveillance compresses the answer to one line: after 12 minutes of continuous monitoring, the operator misses up to 45% of screen activity; at 22 minutes, up to 95%.[1][2] This isn’t a training failure — it’s biology of the attentional system.

The degradation compounds when you add cameras. Tickner & Poulton (1973) measured correct event detection falling from 84% on 9 monitors to 64% on 16.[3][4] More feeds isn’t more eyes; it’s divided attention that performs worse than fewer feeds watched properly.

( five decades of cognitive and CCTV research — Mackworth 1948 through NIJ/Sandia 1999 — synthesized by Ambient.ai, 2025 )

The hardware recorded. Nobody watched. That’s what happens when a system is evaluated by megapixel resolution instead of sustainable minutes of human attention.

Every time I review a deployment for tv4agent, the metric nobody calculates is: how many real minutes of sustained attention does my operator have before the system is recording for no one? Today, 20 minutes per shift, with scheduled breaks, is what the evidence supports. Without a system to compensate, what you have is an expensive DVR with a corporate logo.

P.D. How many cameras do you ask one operator to watch at once? If the answer is “more than 9” or “no limit,” today is a good day to revisit the design.

📰 Source: Ainsworth 2002 (Security Oz) cited in Dadashi 2008 + Tickner & Poulton 1973 + Mackworth 1948, synthesized by Ambient.ai 2025 [1][2][3][4].

Sources

[1] Dadashi 2008 — Automatic Surveillance and CCTV Operator Workload (citing Ainsworth 2002, Security Oz) — https://people.cs.nott.ac.uk/pszcmg/G64IDS/isd-dissertations-08/nxd07m.pdf [2] IPVM 2013 — How Many Camera Views Can Be Effectively Monitored? — https://ipvm.com/discussions/video-operator-viewing-effectiveness [3] Ambient.ai 2025 — The Human Ceiling: Limits of Live Video Monitoring — https://ambient.ai/whitepaper/human-ceiling-live-video-monitoring [4] Alga et al. 2013 — Semi-automated CCTV surveillance: Effects of system confidence (Applied Ergonomics) — https://www.sciencedirect.com/science/article/abs/pii/S0003687012000695