If your application has a very small and stable schema, or if analytics use cases are extremely simple (e.g., a handful of aggregated metrics), a full semantic layer may be overkill.

Teams that rely exclusively on direct SQL and have a single, well‑defined group of analysts might not need the abstraction.

However, even smaller teams often benefit from standardized metrics and easy natural language querying.

For ad hoc exploratory analysis or R&D environments where schemas change daily, maintaining a semantic layer might slow experimentation.

MageMetrics allows you to adopt the semantic layer incrementally: you can start by exposing core metrics and still run raw SQL when needed.