Use case in demand forecasting
summary
DataLab builds a cloud pipeline on Google Cloud using BigQuery, Vertex AI and Python to produce 5 million forecasts/day for a 90-day horizon across ~200 fresh SKUs and 600 stores. A single feature-rich model (horizon as a feature plus weather, competition, promotions, lagged sales and season start flags) replaces many per-horizon models, runs daily, is A/B tested from a 13-store pilot, and shows lower error, higher order acceptance and improvements in revenue, out-of-stock and waste.
AI-written summary · about
chapters (7)
- 0:00Start
- 5:05Many areas of Fortenova Group have potential to be transformed through Data, we are starting with high impact use cases in retail
- 7:07Demand patterns and drivers
- 13:18Multi-step time series forecasting with machine learning
- 15:39Machine learning is an iterative process: it fits well the agile methodology
- 17:58Evaluation of an iteration: 1 improvements of accuracy 2 is there a visible difference in forecasts?
- 22:38Monitoring model performance is key part of the process - Data Studio enables easy sharing with all business stakeholders
related talks
28mDemand Forecasting: Accuracy, Stability, Automation
28mMastering Demand Forecasting: Lessons from Europe's Largest Retailer
18mUsing Transformer Models to Predict Demand at Scale in Production
30mAI in Retail - Advancing Promo Demand Forecasting
28mHow Human Expertise and AI drive Scalable Forecasting Success at
35mBuilding a successful inventory planning solution with Amazon Forecast
24mData Driven Transformation Supporting Demand Forecasting
40mMulti-Series Forecasting at Scale with StatsForecast