Key takeaways
- Solution: Develop a demand forecasting machine learning model.
- Used Data: 2 years of sales history + Weather data (public API) + Local calendar (holidays/events).
- Waste: -25% (direct savings of tens of thousands of dirhams).
- Availability: +15% sales increase due to decreased shortages (-80% on flagship products).
The Challenge: Forecasting Demand
'Délices du Souss' suffered from a major problem: the inability to forecast fresh product needs, leading to stock imbalance oscillating between overproduction (waste) and shortages.
AI wasn't used for an abstract technological feat, but to solve this critical logistics issue.
- Solution: Develop a demand forecasting machine learning model.
- Used Data: 2 years of sales history + Weather data (public API) + Local calendar (holidays/events).
Results and Lessons Learned
After six months, this approach radically changed the company management.
- Waste: -25% (direct savings of tens of thousands of dirhams).
- Availability: +15% sales increase due to decreased shortages (-80% on flagship products).
- Managerial Impact: Reduced stress for the supply team.
Field Lessons
- Clean Data: Success is conditioned by the quality and cleanliness of the two-year history.
- Frugal approach: No need for complex neural networks; a simple model (Gradient Boosting) sufficed.
- Human aspect: AI proposes a recommendation, the manager decides (the model stays under control).
AH
Author
AI HUB Editorial
Research Desk
