Case study ·
10 million forecasts a day: Monoprix entrusts the stocking of its shelves to AI
Original title, in French: “10 millions de prévisions par jour : Monoprix confie le remplissage de ses rayons à l'IA”
- Publication
- Alliancy (France)
- Author
- Coline Lafargue
- Published on
- 2 October 2026 · read in the page structured data
Summary
Over the past two and a half years, Monoprix has deployed a machine learning system, built on the Dataiku platform, that calculates around 10 million sales forecasts every day, fourteen days ahead, by store and by product, to avoid shelf stockouts without overstocking. Presented at the Dataiku Summit by Valérie Latchoumanin, Head of Data Science at Monoprix, the system combines time series and a LightGBM model, incorporates weather, calendar and commercial operations, and had to learn to distinguish products sensitive to cold, to heat or unaffected by weather. It replaces a manually corrected moving average and improves reliability by 5 points on average. Deployed in all Monoprix stores in France and extended to Naturalia, it remains managed by the business teams and has yet to cover fruit and vegetables. Generative AI is being considered as a complement to detect unreported local events.
Summary based on the full text of the article.
Read the article on Alliancy →Figures cited
-
24,000 employees
Monoprix workforce
Monoprix, France
« qui revendique plus de 24 000 salariés et 6 400 points de vente en France »
-
6,400 stores
Number of Monoprix points of sale
Monoprix, France
« qui revendique plus de 24 000 salariés et 6 400 points de vente en France »
-
10,000,000 forecasts
Number of sales forecasts calculated by the tool each day
All Monoprix stores in France
« Soit environ 10 millions de chiffres à produire chaque jour. »
-
5 points
Forecast reliability gain compared with the previous system
Monoprix sales forecasting system, average observed
« le nouveau système gagne en moyenne 5 points de fiabilité par rapport à l'ancien et davantage lors des fêtes, des vacances scolaires ou des événements particuliers »
-
4 weeks
Development time for the first version of the tool
Monoprix
« La première version de l'outil a été bouclée en quatre semaines. »
Classification
- Country
- France
- Company size
- Large companies
- Industries
- Retail and distribution, e-commerce
- Companies
- Naturalia (deploys AI), Casino (cited), Dataiku (supplier), Monoprix (deploys AI), Snowflake (supplier)
- Keywords
- sales forecasting, replenishment, large-scale retail, machine learning, stockout, Dataiku
Entry written by Marsanne Studio from the cited article, with the help of an AI model (claude-sonnet-5). The rights to the article and its illustrations belong to Alliancy and its author. Entry translated from French by Claude Sonnet 5.5.