Small Retailer
Empowered by Vision AI
Advances in AI models, wide-angle imaging, and edge processing now allow defined behaviours to be detected and converted into measurable operational metrics across entire environments.
Behaviour, Not Just Movement
Earlier systems tracked motion. Now, AI models detect defined behaviours within defined zones:
- Queue forming vs dispersing
- Table occupied vs vacant
- Task completed vs missed (e.g. packing, cleaning)
This is not open-ended interpretation.
Wide Coverage Economics
Fisheye and wide-angle optics enable full-area visibility:
- Entire queue zones
- Seating areas
- Kitchen stations
Instead of multiple narrow cameras,one device can cover a full operational zone (~3,000–4,000 sq ft).
Real-World Conditions Are Now Solvable
Previous systems failed in real environments:
- Occlusion in crowds
- People carrying trays / objects
- Children, groups, irregular movement
Modern models are trained on dense, real-world scenarios, enabling:
- Stable queue measurement
- Reliable tracking in cluttered scenes
- Differentiation of roles (e.g. staff vs customers via uniform cues)
Edge + CCTV Reuse Makes It Deployable
Two key enablers remove deployment friction:
Edge Processing
- Data processed locally
- No heavy cloud dependency
- Lower latency, controlled bandwidth
CCTV Reuse
- Existing cameras can be analysed
- Supplement only where needed
From Observation to Score
All detected states are converted into structured metrics:
- Queue length → time-based distribution
- Table occupancy → utilisation rate
- Task completion → compliance score
These are not raw signals.They become standardised variables that can be:
- Compared across stores
- Compared across time (hour / day / campaign)
- Compared across teams
What was invisible is now measurable.
Operations are no longer inferred - they are measured. Execution is no longer sampled - it is continuously visible. Performance is no longer subjective - it is comparable across locations, teams, and time.