Shelf photos that turn themselves into a report
Field managers were already photographing shelves across cities and chains. A second group of managers then read every photo by hand to build the report.
- Client
- Meat products manufacturer
- Scope
- Shelf photo → recognition → report
- Surface
- Telegram bot with a WebApp
- Role
- Solo — design, build, hand-over
Every shelf photo read twice by a human
Tracking your own brand and your competitors' across chains meant a lot of photographs, and a lot of manual reading.
Managers photographed store shelves at different locations and in different cities to track the presence of their own brand alongside competitors' across retail chains.
Other managers then reviewed all of those photos manually and compiled them into a report — a second full pass over work that had already been done in the field.
A Telegram bot that does the reading
Delivered where the field team already works, so there was no new app to adopt.
Photograph the shelf
fieldThe manager photographs the shelf at a location, exactly as before.
Send it to the bot
telegramThe photo is uploaded to the Telegram bot — no separate system to log into.
Recognise and report
visionThe agent recognises the products, their prices and the number of facings, and generates the report.
Confirm
checkpointThe manager reviews the recognised result and confirms it before it counts.
Python
PostgreSQL