AI for retail: till video checks, footfall and demand forecasts
We build AI for retail chains on the cameras and tills you already have. Every receipt is matched with video, footfall is counted per entrance and zone, purchasing gets demand forecasts, and the data flows into your ERP.
Typical tasks
- 01Every receipt matched with video
POS events are matched with footage from the camera above the till. The system finds refunds with no customer, voids after payment and sales without a receipt. We run this for a bakery chain under NDA.
- 02Footfall counting
Entrance cameras count visitors by hour and day. Combined with receipts, you see conversion for each store. Staff rotas can follow real traffic.
- 03Store heatmaps
The system shows where people stop and which shelves they skip. Merchandising tests layout changes on data, not on gut feeling.
- 04Demand and order forecasts
The model forecasts demand per product for each store. Purchasing gets suggested orders that account for season, promotions and stock. Less waste and fewer empty shelves.
- 05ERP and inventory sync
Sales, stock, prices and orders sync with SAP, Odoo or your inventory system. AI data lands in the reports finance and purchasing already use.
- 06Morning report
The owner gets a summary every morning: revenue, footfall, conversion and till incidents per store.
Data and access requirements
- POS integration is read-only: we take fiscal events through API and never change the till software.
- GDPR for video: signage, no facial recognition, limited retention of clips and a DPIA. The EU AI Act restricts biometric identification, so we do not use it.
- Integration with SAP, Odoo or your inventory system through API or file exchange.
- We use existing cameras over RTSP or ONVIF, with processing on our or your servers.
What we build for this industry
Case
- // RETAIL · CVBakery chain · NDA
Till control in a bakery chain
TaskCash mismatches and sales without a receipt.
What we didA camera over the till matched against receipts. Mismatches go into the report with video.
Questions
What do you need from us to start?
Access to the till and entrance cameras, POS API access and an export from your ERP. During discovery we pick a few stores for the pilot and check video quality.
How much does retail video analytics cost?
From $5,000 for a turnkey project. The price depends on the number of stores, cameras and scenarios. You get a fixed estimate 2-3 days after discovery.
How soon will we see results?
The build takes 4-8 weeks with weekly demos. We start on part of the stores, compare with your current reports, and then roll out across the chain.
Can it work without facial recognition?
Yes, that is our default. Footfall counting and till checks do not need to identify anyone. The system counts silhouettes and actions, not identities.
What if a camera has a poor view?
AI sometimes errs in bad light or with a blocked frame. We show which cameras should be moved, and we draw no conclusions from footage where nothing can be seen.
Describe the task. The first call is with the founder.
We reply within one business day. 30 minutes to tell you whether it is worth doing and what it costs.