BigQuery measures is now GA, unifying BI metrics and knowledge graphs for AI accuracy

Google Cloud has made BigQuery measures generally available, addressing a longstanding gap where business metrics and entity relationships were siloed across separate systems. Historically, organizations relied on disjointed pipelines to reconcile BI dashboards with knowledge graphs, often leading to inconsistent definitions and governance failures. The new capability integrates measure definitions directly into BigQuery Graph, enabling AI agents, SQL queries, and dashboards to operate from a single, trusted semantic model. This eliminates the need for manual reconciliation and reduces discrepancies between teams relying on the same data.

A retail banking example illustrates how BigQuery measures resolve a common analytics challenge. Teams investigating a transaction spike in Prague previously required separate tools to calculate volume and trace account flows, often producing conflicting results. With BigQuery measures, analysts can define semantic models over transaction, account, and district tables, then use GRAPH_EXPAND and AGG() to aggregate measures while preserving the native grain of each entity. This ensures accurate counts and prevents overcounting, such as when a single account has multiple transactions, by locking measures to a fixed granularity.

The generally available release expands measure modeling capabilities to support composite and derived metrics, allowing complex business logic to be defined once and reused across queries. BigQuery validates measure dependencies at schema compilation, catching errors early and enabling consistent calculations at any dimensionality. For instance, unsettled volume can be derived from settled and total volume, with the logic embedded directly in the schema. Additionally, the GRAPH_EXPAND function now includes a start_node argument to prevent chasm traps when joining fact tables, ensuring each measure is evaluated at its native grain before aggregation.

BigQuery measures also introduce richer semantic metadata and tighter integration with Google Cloud’s Knowledge Catalog. Measures can now include plain-language descriptions, domain synonyms, and custom extensions via the OPTIONS clause, clarifying business meaning for analysts and AI agents. Knowledge Catalog automatically ingests this metadata, enabling unified discovery where users can search by business terms, inspect governance tags, and verify SQL definitions. This closes the loop between data governance, semantic context, and machine-readable metrics, supporting consistent interpretation across teams and AI-driven workflows.