Normalized claim
Forecast performance improvement: 43% increase
Vertex AI Forecast provided a 43% performance improvement relative to models trained in-house on a custom VM.
Coop reduces food waste by forecasting with Vertex AI and BigQuery
Use case typeDemand forecastingUpdated Mar 16, 2023
Coop, a Swiss retailer and cooperative, used Google Cloud to operationalize machine-learning forecasting for demand planning based on supply-chain seasonality and expected customer demand. The team moved from an on-premises workstation setup to Vertex AI Workbench and Vertex AI Forecast, with BigQuery and Google Kubernetes Engine supporting the broader data science platform. The goal is to improve forecasting accuracy, support distribution centers, and reduce food waste across Switzerland.
- Organization
- Coop
- Industry
- Retail
- Location
- Switzerland
- Published
- March 2023
Reported outcomes
14.5 WAPE
forecast error on test setQuality & accuracy
+43%forecast performance improvement
Strategic outcomes
Sustainability & ESGReduced food wasteBetter decisions & insightBetter support for distribution-center planningInnovation & cultureCreated a standard ML platform for other teams
Why do we believe this?Outcome claims, sources, and evidence checks
Normalized claim
Forecast error on test set: 14.5 WAPE decrease
On the test set in the POC, the team reached 14.5 WAPE (Weighted Average Percentage Error)
Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks
- Customer
- Coop
- Provider
- GCP
- Maturity
- Pilot
- Linked source
- Google Cloud Blog
The pilot is intended to support one distribution center and potentially scale across Switzerland
Customer identity supportedSource describes one deploymentMaturity supported
Primary read
Use case focus
Showing 3 of 3
- 1Demand forecasting
- 2Supply chain planning
- 3Operations optimization
- Coop's on-premises forecasting environment could not scale to support model tuning and larger CPU or GPU workloads.
- The team needed better demand planning for distribution centers based on seasonality and expected customer demand.
- Reducing uncertainty was important to avoid over-ordering and food waste while keeping products available for customers.
- Coop ingested data from its data pipelines and SAP systems.
- The team used Vertex AI Workbench to develop forecasting workflows and Vertex AI Forecast to operationalize models.
- BigQuery and Google Kubernetes Engine were used to create a more efficient ML platform, with forecasting insights planned to stream back to SAP.
- Vertex AI Forecast delivered 43% performance improvement relative to models trained in-house on a custom VM.
- The team achieved 14.5 WAPE in the proof of concept.
- The pilot is intended to support one distribution center and potentially scale across Switzerland.
Architecture
Coop ingested data from pipelines and SAP systems into a Google Cloud ML workflow. The team used Vertex AI Workbench for development, Vertex AI Forecast for operationalized forecasting models, BigQuery as a pre-stage for Vertex AI, and Google Kubernetes Engine for the wider ML platform. Forecasting insights were designed to stream back to SAP for downstream planning.
Sources & evidence1
Evidence: Medium50/100Evidence strength
- Customer explicitly identified
- Deployment status explicitly supported
- Quantified outcome available
- Technical implementation details available
Type: Blog PostPublished: Mar 16, 2023Publisher: Google CloudEvidence: VendorConfidence: Medium
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Customer evidence
Provider evidence
Browse the catalog
Was this useful?
Community
Comments
No published comments yet.