Glossary/Semantic Layer & Metrics

Metrics Store

A metrics store is a centralized repository that persists metric definitions, calculated values, and metadata, enabling fast access and governance of business metrics across the analytics platform.

A metrics store is where metric definitions and their pre-computed or cached results live. Unlike the metric layer, which is conceptual (the abstraction and logic), the metrics store is the physical system: a database, data warehouse table, or specialized service that maintains actual metric records. It stores the metric name, formula, dimensions, time series values, ownership, lineage, and historical changes.

The metrics store solves latency and consistency problems. Recalculating revenue across billions of transactions every time someone opens a dashboard is expensive. A metrics store pre-computes metrics at defined grain (daily, hourly) and stores results, making retrieval instant. It also provides a queryable registry: teams can discover what metrics exist, who owns them, and what they mean, reducing redundant metric creation.

Modern metrics stores integrate with BI tools, APIs, and SQL engines. Some are purpose-built systems (like dbt Cloud's metrics interface or specialized metrics platforms). Others are implemented as tables in the data warehouse with governance overlays. A metrics store includes not just final values but metadata: when the metric was last updated, data freshness guarantees, associated dimensions, and change history.

Key Characteristics

  • Persists metric definitions and calculated values
  • Enables fast metric retrieval without recalculation
  • Maintains metric metadata and ownership
  • Tracks metric history and lineage
  • Provides discovery interface for metric inventory
  • Supports dimension and filter combinations efficiently

Why It Matters

  • Performance: Cached metrics eliminate expensive recalculations
  • Availability: Metric values always queryable, supporting SLOs
  • Discovery: Single source to find metrics, definitions, owners
  • Compliance: Audit trail of metric calculations and changes
  • Reliability: Separate from transactional systems; governed updates

Example

A financial services company pre-computes daily metrics (customer lifetime value, churn rate, account balance) into a metrics store. Dashboards query pre-calculated values (microsecond response), while a metadata API tells analysts that "churn rate" includes only customers with 90+ days tenure, last updated 2 hours ago.

Coginiti Perspective

Coginiti's approach to a metrics store is definition-first rather than pre-computed. SMDL files store measure definitions as code, and Semantic SQL computes metric values on demand against the underlying platform. For cases requiring pre-computation, CoginitiScript publication can materialize metric results as tables, views, Parquet, or Iceberg on any supported platform. The Analytics Catalog provides the organizational layer, with workspaces for personal development, shared collaboration, and production publication of metric definitions.

See Semantic Intelligence in Action

Coginiti operationalizes business meaning across your entire data estate.