How BMW Group detects cost anomalies across 14,000 cloud accounts
aws ·
BMW Group has developed Cloud Efficiency Analytics (CLEA), an in-house FinOps system that monitors over 14,000 cloud accounts to detect cost anomalies automatically. The system ingests daily billing data from AWS Cost and Usage Reports and aggregates it into account-service pairs for forecasting. Using Meta’s Prophet library, CLEA generates expected spend baselines based on 365 days of historical data for each pair. To distinguish significant deviations from normal fluctuations, the platform applies a multi-layered filtering process that includes deviation thresholds, account clusters, and service-specific rules. This approach minimizes false positives by requiring alerts to meet both percentage and absolute dollar impact criteria before notifying owners. The entire detection pipeline runs serverlessly on AWS Step Functions and Lambda, processing all accounts in approximately 20 minutes for about $50 per month. Account owners receive detailed email alerts with context and links to self-service dashboards for root cause analysis. Future enhancements include ITSM integration and agentic AI tools to provide automated explanations for spending spikes.
aws