Case Study: Delivery Hero achieves faster, scalable voucher fraud detection model management with Looker

Serving up data across the company to ease decision-making with Google Cloud

Delivery Hero, the global local delivery platform, needed a faster way to manage and redeploy more than 20 voucher fraud detection models across 70+ countries and multiple regional clusters. The Incentive Fraud team was challenged by frequent model updates, strict latency requirements, and varying data protection rules, and they used Looker’s broader analytics ecosystem alongside Vertex AI and GitHub-based MLOps workflows to keep model development and deployment under control.

Looker-enabled reporting and model visibility helped Delivery Hero’s team implement an automated CI/CD process around Vertex AI Model Registry, GitHub Actions, and custom tooling for tracking datasets, experiments, and model aliases. With Looker-style visibility into metrics and lineage, the team cut model release time from days to about one hour, reduced the end-to-end ML cycle to four hours, enabled 20+ fully automated deployments, and achieved a 70% reduction in voucher fraud.

Delivery Hero

Mathias Nitzsche

Vice President Engineering Global Data