How we count people on your CCTV
What the numbers in your 48-hour analysis mean, how we produce them, how we check them, and where they are less reliable.
By Stella Nurina Susanto · October 2026What we count
- Visitors. People crossing a line we draw across each entrance, in one direction.
- Busy hours. The same entries, grouped by hour, for each day of footage.
- Heatmap and time in view. Where people spend time inside each camera's view, and how long they stay in it. We report the median, so one person who stays all afternoon doesn't skew it. We don't follow anyone from one camera to the next.
- Queues. How many people stand inside a queue zone at your tills, minute by minute, and how long they wait.
How it works
The software finds each person in every frame as an anonymous shape and follows that shape for as long as it stays in view. When the shape crosses an entrance line, enters a zone or joins a queue, it adds to a count. Each person in view gets a temporary number that's thrown away when they leave the camera. Nobody is named, matched across cameras or recognised by their face.
We only count your opening hours. Footage outside them is ignored.
How we check it
Before you see any numbers, a person on our team counts short samples of your own footage by hand. We take two 15-minute samples per camera, one quiet and one busy, and compare our hand count with the system's. Your results show that comparison next to the numbers, camera by camera.
We don't publish a general accuracy figure, because it depends on your cameras. The check on your own footage is the one that matters.
Where it works less well
- Crowds. When people overlap heavily, for example at a sale, counts can run low.
- Steep or very high cameras. Looking straight down makes people small and harder to separate.
- Glare and low light. Strong sun on a glass entrance or a dark corridor reduces accuracy.
- Groups and children. People walking close together can count as fewer; small children may be missed.
- Staff. Staff are counted like anyone else unless they stay in an area we can leave out, such as behind a counter.
If one of your cameras has one of these problems, the hand check shows it, and we tell you which numbers to rely on.
What we never do
- Facial recognition, or matching people across visits.
- Storing images of people. We keep counts, not pictures.
- Keeping your raw footage: it's deleted 30 days after upload.