Capability
Loitering detection — theft has a shape, and it starts with standing still
Loitering detection on CCTV catches the part of an incident that happens before the incident. A person who stands at a high-value shelf, a fence line or a loading bay for far longer than the task requires is often worth a closer look — and it is a pattern a busy human easily misses. With PGAK, loitering detection is evaluated per zone on your own cameras; there is no published PGAK demonstration of it yet.
How loitering detection cctv works
Mark the zone
The shelf, the bay, the fence stretch, the parked-vehicle row — anywhere lingering is unusual.
Set a dwell threshold
How long is too long here? Thirty seconds at a jewellery counter; three minutes at a fence.
Exempt the expected
Where face recognition is configured, enrolled staff working in the zone can be exempted. A queue at a till is handled by per-zone thresholds — and queue and legitimate-worker scenarios are part of the test at your site.
Alert discreetly
Often the right response is a quiet nudge to a manager's phone rather than a siren. A staff member walking over is often an effective deterrent.
Where it earns its keep
- High-value retail shelves and jewellery counters
- Warehouse aisles holding fast-moving, high-value SKUs
- Perimeter fence lines where someone is assessing a way in
- ATM lobbies, parking areas and society common areas
What it won’t do
Every AI vendor lists strengths. These are the limits, so you can plan around them instead of discovering them.
- Legitimate lingering exists — a customer genuinely deciding, a technician working. Thresholds and staff exemptions carry the weight here, and they need tuning per zone.
- It flags behaviour, not intent. Every alert deserves a human glance before any action, and the system is designed around that assumption.
Loitering & dwell time — common questions
Won't this flag ordinary customers browsing?
Only if the threshold is set badly. Dwell times are per-zone — thirty seconds at a jewellery counter, three minutes at a fence — and staff are exempt. How rare the alerts become is measured on your own footage during testing, including queue and legitimate-worker scenarios.
Is this the same as motion detection?
It's almost the opposite. Motion detection reacts to movement; loitering detection reacts to a person tracked continuously in one zone without leaving. That requires tracking an identified person across frames, which pixel-difference motion detection cannot do.
What should we do when a loitering alert fires?
Look at the snapshot, and if it warrants it, have someone walk over. Presence is the deterrent. We'd caution against automated sirens on dwell alerts specifically — the false-positive cost is a confronted customer.
Sectors that lean on this: Warehouses and logistics, Retail shops and showrooms, Jewellery showrooms, Pharmacies and medical stores — see every industry PGAK serves.
Calculator · C08
Loitering detection: estimate it for your site
Use your own measurements. The result is the staff time released, shown separately from any cash, using the formula below. It runs in your browser and nothing is sent anywhere.
Result
Enter your own figures, or load the worked example, to see the result.
How it is calculated
hours = alerts × (manual − assisted) ÷ 60 − tuning hours
Worked example (illustrative inputs)
120 alerts; 6 old; 2 new; 1 tuning hour.
staff time released: 7 hours a month.
Formula version 1.0.0. The full standalone tool: false alarm cost.
What this does not tell you
- Hours released are capacity, not cash. They become cash only through a mechanism you state — overtime that stops, a post that is not refilled — entered as the realisation fraction.
- The 'after' values are your assumptions or pilot measurements. PGAK publishes no effectiveness figure for this feature.
Evidence status: Supplier/integration or marketed capability; PGAK availability and site performance require verification.
Discuss your camera setup
Share your phone number, city and camera count. We will clarify the intended use case and agree assessment scope and timing with you. Do not submit camera credentials or footage.