Lyft built an internal Metric Semantic Layer (MSL) to solve the problem of inconsistent metric definitions across teams. Implemented as a Python package, MSL uses YAML configurations with Jinja-templated SQL to store canonical metric definitions. Governance is enforced through 'Golden Metrics' selection criteria (metrics with at least two distinct use cases) and a dual-ownership model requiring both a Business Owner and an Operational Owner for every metric. The system exposes metrics via Python APIs, integrates with Lyft's Amundsen data catalog, offers a self-service UI for SQL generation, and supports AI agents through an MCP (Model Context Protocol) integration enabling natural-language metric queries in tools like Claude and Cursor. Lyft is now exploring vendor-managed solutions to extend MSL further into BI tools and AI-driven analysis.
