Metric Semantic Layer: How Lyft Governs and Scales Key Data Definitions

This post details the implementation of Lyft's internal Metric Semantic Layer (MSL) as a Python package. It elaborates on the use of YAML configurations for metric definitions and Jinja SQL templates for dynamic SQL generation, emphasizing DRY principles. The post outlines the governance model, defining 'Golden Metrics' and establishing dual Business and Operational Owner roles for metric health and data health respectively. It also describes access methods including integration with Amundsen for discoverability, a self-service UI, and its application as a knowledge base for AI Agents and Skills, including the use of an MCP for natural language metric querying.