Why Google Is Extending the Semantic Layer with Business Relationships for Agents

From "What Happened" to "Why It Happened"

Historically, the semantic layer focused on defining metric calculations and unifying definitions across BI tools. In the Agent era, the problem shifts to explaining why a metric changes and how that change propagates through business relationships. Google Cloud’s recent updates to BigQuery Graph with Measures place governed Measures and Graph in a single model, extending the semantic layer from pure metrics to entities, relationships, and business context.

Graph Finds Paths, Measures Ensure Numbers

BigQuery Graph with Measures does more than combine graph queries with SQL aggregation. Because graph traversal can duplicate rows, ordinary SUM or AVG would over‑count. Google defines MEASURE objects (SUM, AVG, COUNT, COUNT DISTINCT, MIN, MAX) on node or edge properties, binding aggregation to a unique key. This guarantees correct results even after expansion. Queries use GRAPH_EXPAND to materialize the graph as a table and then apply the predefined Measure. The official example shows how a Measure on the User node avoids double‑counting customers when counting distinct users served by each distribution center.

Semantic Layer Boundary: From Metrics to Business Structure

Google’s retail case study illustrates the need for multi‑hop relationship tracing: a 12% drop in Seattle jacket sales can be linked to downstream suppliers affected by a storm. The analysis pipeline consists of three layers—Metadata Grounding (identifying available data), Business Metrics (calculating measures), and Relationship Mapping (using the graph to trace causality). Without the graph, agents would only see the sales dip and might suggest price cuts, missing the supply‑chain root cause.

Google’s Real Focus: The Agent Business‑Context Layer

Beyond BigQuery, Google’s Knowledge Catalog (announced April 22) aims to become an enterprise Context Engine, aggregating Measures, LookML, data products, and metadata from various platforms. It enriches entities, relationships, and business vocabularies, addressing the “semantic gap” that causes hallucinations in agents. Integration with Looker allows LookML to reference Graph definitions and vice‑versa, ensuring a single source of truth for KPIs across dashboards and agents.

The overall direction is a shift from the classic "Metrics + Dimensions + Joins" model to "Metrics + Entities + Relationships + Business Context". While the semantic layer, Knowledge Graph, and Ontology remain distinct products, Google is progressively incorporating business relationships and context into the governed data model, enabling agents to move from answering single‑point queries to root‑cause analysis and cross‑domain decision making.