Computer vision · Retail
See the empty shelfbefore the customer does.
Retail video analytics uses in-store cameras to flag a shelf that is running empty, a checkout queue that is getting longer or the zones most visitors walk through. Everything is aggregated and anonymous: we count and measure, without recognising customers and without building profiles.
The customer sees the empty shelf before the team does
In a busy store, staff cannot watch every shelf and every checkout at the same time. Stock gaps, queues and zones customers avoid are noticed late, and decisions about restocking and staffing are based on impressions.
- Missing products are discovered during shelf-check rounds, not when they run out
- Opening an extra checkout is decided once the queue is already long
- There is no data on which zones get more or fewer visitors
- Visitor data must respect customer privacy, so identifying people is not a suitable option
Who it is for
- Store chains and supermarkets that want to react faster to empty shelves
- Store managers who decide how many checkouts to open and when
- Merchandising teams that want to understand how visitors move through zones
- Convenience stores with small teams, where no one can check the shelves all the time
What it can flag and measure in the store
- 01
Shelf gaps
Detects empty spaces on monitored shelves and sends the team a notification with an image of the shelf.
- 02
Queue length
Estimates how many people are waiting at the checkout and flags when the queue exceeds the threshold set by the store.
- 03
Aggregated traffic maps
Shows which zones of the store get more visitors, by hourly interval, without saving individual paths.
- 04
Entry counting
Counts entries anonymously, so traffic can be compared with the number of receipts issued in the same interval.
- 05
Visible shelf layout
Can flag visible differences from the agreed layout, where the camera angle allows it.
How it works
From store cameras to team action
The steps marked in yellow are approved by a person.
- Input
- AI
- Action
- Human in the loop
- Result
- 01
Input
In-store cameras
Compatible cameras capture the monitored shelves, the checkouts and the entrance.
- 02
AI
Anonymous analysis
The model detects shelf gaps and counts people, with no facial analysis.
- 03
Action
Aggregation by zone
The data becomes totals by zone and interval, with no images of people kept.
- 04
Action
Team notification
Gaps and queues above the threshold reach staff phones or tablets.
- 05
Human in the loop
Staff decide
An employee checks the shelf, and the manager decides whether to open a checkout.
- 06
Result
Store report
Trends in traffic, queues and restocking, for planning.
What data and cameras the store needs
- Real images of the monitored shelves, full and empty, at different times and in different lighting
- The planogram or product list for the monitored shelves, if layout checks are wanted
- Thresholds set by the store: queue length, how empty a shelf must be to trigger an alert
- Cameras with a clear view of the shelf or checkout area, with no permanent obstructions in the frame
- Sales data aggregated by interval, if correlation with traffic is wanted
Possible integrations
- IP cameras with an RTSP stream
- ONVIF-compatible cameras
- NVRs and VMS systems
- POS systems, for aggregated sales data
- ERP or stock management system
- Notifications in Microsoft Teams or Slack
- Staff mobile apps, via API
- Power BI or Looker Studio dashboards
- REST API and webhooks
- IP cameras with an RTSP stream
- ONVIF-compatible cameras
- NVRs and VMS systems
- POS systems, for aggregated sales data
- ERP or stock management system
- Notifications in Microsoft Teams or Slack
- Staff mobile apps, via API
- Power BI or Looker Studio dashboards
- REST API and webhooks
Each integration's compatibility is checked during the analysis phase, for the systems and versions you use.
Control and transparency
What stays with people
- Physically checking the shelf before placing a restocking order
- Opening checkouts and allocating staff, which remain the store manager's decision
- Approving the zones analysed and how customers are informed, for example through signage at the entrance
- Regularly reviewing that the system keeps no more data than necessary
Limits and operating conditions
- Works best on shelves with a clear front view; partial gaps, products pushed to the back or very deep shelves are hard to assess from a single angle
- Queue estimates become less accurate when people overlap or stand in groups
- Changes to the layout, shelving or lighting require the zones to be reconfigured and, sometimes, new data
- The analysis does not say who customers are or why they chose a product; it only provides aggregated data
How we evaluate shelf alerts and queue estimates
We agree the indicators together before launch and track them afterwards. The values depend on the process, the data and the volume.
- Confirmed empty-shelf alerts
- The share of notifications confirmed by staff when checking the shelf, recorded in the app
- Queue estimate error
- Estimates compared with manual counts of people at the checkout, at randomly chosen moments
- Response time
- The interval between the notification and confirmation of restocking, calculated from the log
- Shelf coverage
- The share of monitored shelves with a clear enough view for analysis, checked during set-up
How we implement video analytics in the store
- 01
We choose the shelves and zones
We start with a few key shelves, the checkouts and the entrance, plus the alert rules.
- 02
We check the cameras and customer information
We review the existing angles and streams and agree how customers are informed about the processing.
- 03
We configure and calibrate
We draw the zones, adapt the model to the store's images and compare against manual checks.
- 04
Pilot in a single store
We run it in one location, fine-tune it with the team and only then discuss scaling up.
Use cases
The scenarios below are conceptual examples, designed to illustrate realistic applications. They are not implementations delivered by WEB-AZ.
- Supermarkets
Restocking the dairy shelf
The camera facing the shelf flags gaps, and the employee on shift receives the image on a tablet and checks stock in the storeroom.
Conceptual example - Convenience stores
When to open a second checkout
When the estimated queue exceeds the set threshold, the manager receives a notification and decides whether to open a second checkout.
Conceptual example - DIY and home improvement
Aisles visitors avoid
Aggregated traffic maps show rarely visited aisles, and the merchandising team tests a different product layout.
Conceptual example
Useful guides
Before you decide, read
Frequently asked questions
What is video analytics for retail?
It is the use of in-store cameras, together with a computer vision model, for three practical things: flagging empty shelves, estimating checkout queues and producing aggregated traffic maps by zone. Staff receive the notification and decide, and the data stays anonymous and aggregated.
Does the system recognise returning customers?
No. The solution does not use facial recognition and does not build profiles. People are only counted in zones, and the results are aggregated totals.
What data is kept?
Usually only totals and events, for example “empty shelf at time X”, plus images of shelves without people, where possible. Retention periods, access and customer information are set in line with your video surveillance policy and applicable law.
What happens if the system is not sure the shelf is empty?
Uncertain cases can be sent as “to check” rather than as a firm alert. Staff confirm at the shelf, and their responses help us tune the thresholds.
Can I use the store's security cameras?
Sometimes. Security cameras are usually installed to cover checkouts and entrances, not shelves. We check what they actually show and, if needed, recommend additional cameras pointed at the shelves.
Can it be linked to sales data?
Yes, at an aggregated level, if your checkout system allows exports. We can place traffic by interval alongside the number of receipts, with no data about individual shoppers.
Can’t find the answer? Write to us and we’ll reply directly.
Want to know about an empty shelf before your customers do?
Tell us which shelves and checkouts matter most. We review your existing cameras and what data can stay strictly anonymous.
Related solutions
- People and object countingAnonymous counting of crossings, zone occupancy or the items in a batch, without identifying anyone.
- AI assistant for e-commerceGuides shoppers to the right product using your catalogue, answers questions about availability and orders, and prepares cart recovery messages only where consent exists.
- AI agent for data analysisYou ask a question about your company's data in natural language, and the agent writes the query, runs it on authorised sources and explains the result, showing how it got there.