The stores already had cameras, but nobody had time to watch them. Losses were found weeks later in stock counts, long after anything could be done about them.
Case study — Retail
Loss-prevention vision for store floors.
How a retailer’s existing cameras began flagging the moments that matter — without asking staff to watch more screens.
Client
Retail chain (name withheld)
Disciplines
Computer vision, edge deployment
Engagement
Feasibility study, pilot, rollout
Delivered
Event-detection models, alerting service, store dashboard
Fig. 1 — A store aisle seen from the ceiling camera the model works from.
01The problem
02Our approach
We tested the approach in a feasibility study on recorded footage from a few stores, tuning for each store’s layout and lighting rather than relying on one generic model. A small pilot followed, with alerts sent to store managers’ phones so we could measure how many were actually useful.
“It tells us where to look. That’s all we needed.”
Store operations manager — quote to be confirmed
ALERTS THIS WEEK · 6 PILOT STORES
−31%
unexplained stock loss
Fig. 2 — The store-alerts dashboard (figures shown are illustrative).
03What we delivered
Event-detection models running on in-store hardware, an alerting service that sends a short clip to the right person, and a dashboard showing alerts and outcomes store by store.
04The result
Store teams now respond to flagged events on the day instead of finding losses at the next count. Replace this paragraph with the measured outcome once the client has approved the figures.