FranceRetail and e-commerce2019 – ongoing

zest HACCP

Automatic reading of food labels at goods reception

zest HACCP halved data entry time per delivery from 30 to 15 seconds by reading food labels with Cloud Vision

WorkingVendor case studyVendor ↗

Problem

French supermarkets must record details of every delivered product, such as lot number and allergens, and keep the label data for six months. Staff did this by hand or filed the paper label in a folder.

Approach

zest HACCP, which sells a food-safety record-keeping platform to retailers, built a label-reading tool with partner Cirruseo. A store operative photographs the label, Google Cloud Vision extracts the text, and a scoring system built on App Engine decides which text belongs in which field, such as supplier, allergen code or batch number. Cirruseo used a rule-based scoring system because no data set was available to train a machine learning model. The operative then checks the pre-filled form. The tool was built in five weeks and was in production in a number of stores.

Results

15 secondsdata entry time per delivered unit, down from 30 secondsVendor case study[1]

85%accuracy of the stored records, up from 70% before, as stated by Cirruseo's head of dataVendor case study[1]

Takeaway

The tool pre-fills the form and the operative only validates it, so the time saved falls on the person receiving the goods.

Sources

zest HACCP: Developing a fresh approach to processing food labels with machine learning
Google Cloud5 Jul 2019Vendor publication

Source summaries

[1]Google Cloud5 Jul 2019

Google Cloud's customer story describes how zest HACCP and its partner Cirruseo built a tool that reads food labels with Cloud Vision and sorts the text into record fields. It reports data entry time falling from 30 to 15 seconds per unit and record accuracy rising from 70% to 85%. The page carries no date; the Internet Archive first saved it on July 5, 2019.

Full details are in the original sources.

Added 4 Nov 2026 · Updated 4 Nov 2026