Google releases BigQuery measures, one model for metrics and graphs

Google tells that BigQuery measures are now open to all customers. A measure is a business metric, such as revenue at risk, that you write one time. It sits on BigQuery Graph, which holds the links between things such as accounts and transactions. Google tells that an agent then sees the number and the cause of it, and guesses less. It gave no test results.

How it works

You write measures on the graph, with no data move and no second system. Dashboards, SQL queries and agents use the same definition. Each measure counts at its own level, so an account with thousands of transactions counts one time. A new argument lets one query use many fact tables with no double count. You can add business meaning to a measure, so an agent can tell metrics with a close name apart.

The proof

Amanda Reelick of Woolworths NZ tells that measures give dashboards, agents and data team queries the same logic. The source gives no numbers for precision, speed or price.

What is not known

Google tells that an agent without both views guesses the remaining part. It gave no data on how much the wrong answers decrease.

Insight: Google builds the layer where a metric gets its meaning

In October 2026, Google put Looker metric formulas in its data catalog for agents. Now it puts measures and graphs in BigQuery. This reads as a race to own the layer where a metric gets its meaning, because the owner of the definition selects what an agent answers. A definition kept in one vendor layer is hard to move.