Ahold Delhaize trains ML models for personalization and inventory forecasting across brands

Overview

Ahold Delhaize, one of the world's largest food retail groups (7,452+ stores across multiple countries), built a self-service data platform on Databricks Lakeflow Jobs and Auto Loader that lets data engineers across its brands build pipelines for AI/ML needs. Etos uses the platform for machine learning and model training for personalization and inventory forecasting; Albert Heijn's ML engineering team uses Lakeflow Jobs to orchestrate feature engineering and continuous ML model training/retraining pipelines. The platform runs about 1,165 data ingestion jobs daily, cut deployment time 4.5x (1.5 hours to 20 minutes), and reduced costs by 50% through cluster reuse.

The challenge

Ahold Delhaize needed to modernize a complex legacy architecture with multiple triggers, external orchestrators like Azure Data Factory, and dependent pipelines across its global brands, which made building and deploying data pipelines a laborious, unmanaged process.

The solution

Ahold Delhaize built a self-service data platform using Databricks Lakeflow Jobs and Auto Loader, streaming raw store, online and merchandising data from Kafka into a Delta Lake medallion (Bronze/Silver) architecture; brand teams like Etos use it for machine learning and model training for personalization and inventory forecasting, while Albert Heijn's ML engineering team uses Lakeflow Jobs to orchestrate feature engineering and continuous model retraining pipelines.

Reported business value

The company now runs about 1,165 data ingestion jobs daily, cut full deployment time 4.5x from 1.5 hours to about 20 minutes, and achieved cost savings of more than 50% through cluster reuse.