Vision · Retail
Store video analytics for retail
In retail the real loss comes from empty shelves: the product is in the stockroom but not on the shelf. Zero-Shot builds vision systems that measure shelf availability, planogram compliance and queue density. The setup is anonymous — no facial recognition, no identity matching, only counts and zone data.
What we build
- 01Shelf availability detection with out-of-stock alerts
- 02Planogram compliance checks — right product, right facing
- 03Checkout queue length and waiting time estimates
- 04Anonymous density and space utilisation maps
- 05Alerts delivered to store staff in real time
How it runs
01
Privacy setup first
What is processed, and what is never recorded, is written down. Frames are turned into numbers and not retained.
02
Pilot in one store
It starts on a few aisles in one store, and we measure whether staff actually act on out-of-stock alerts.
03
Match against sales data
Alerts are compared against lost sales. That is where the value shows; alert counts on their own mean nothing.
04
Roll out to the chain
Once the gain is confirmed it spreads, with thresholds tuned per store.
What we measure
- Out-of-stock detection accuracy and detection time
- Time to restock after an alert — whether alerts change behaviour
- Movement in planogram compliance
- Gap between queue estimates and real waiting time
- Change in estimated lost sales from empty shelves
When this is the wrong fit
Facial recognition, identity matching, customer profiling and person-level tracking are out of scope — we do not build them. If camera angles do not cover the shelves, the system cannot work and camera placement has to be discussed first.
Common questions
Do you use facial recognition?
No. There is no identity detection, facial recognition or person matching in the setup. The system produces anonymous counts and zone data, and frames are not retained after processing. That is the boundary we prefer, for privacy law and for customer trust.
Are our existing security cameras enough?
Usually enough for queues and footfall, often not for shelf availability — security cameras point at the aisle, not at the shelf. The initial assessment separates what current hardware can do.
How many stores does this need to make sense?
Even a single store can gain from out-of-stock alerts, but the real value appears across a chain, where comparison shows which store's problem is a process problem. We pilot in one store and let the measurement decide.
Other services
What we've written on this
Last updated: 2026-09-05
Let's talk about this one
Tell us what you want to build and within two days we'll tell you who takes it and how we'd measure it.