BigQuery Introduces New Augmented Analytics Table-Valued Functions

New Augmented Analytics Capabilities in BigQuery

Google has introduced a suite of augmented analytics Table-Valued Functions (TVFs) within BigQuery. These new functions are designed to automate complex data analysis at scale by combining AI, machine learning, and statistical methods. The goal is to facilitate insight discovery and pattern explanation directly within the data platform.

Functionality and Integration

The TVFs allow users to diagnose reasons for metric changes, identify underlying trends and relationships in data, and determine the true impact of business decisions. They operate directly where the data resides, which helps to accelerate analysis and minimizes the need to export data to external tools. Their structured SQL outputs also enable easy integration as skills for AI agents, supporting automated and conversational data investigation workflows.

Specific Analytical Challenges Addressed

Six new augmented analytics functions have been introduced, each targeting a specific analytical challenge. These functions can be chained together, allowing the output of one function, such as a detected time window, to parameterize the subsequent analytical step. An example provided uses ML.DETECT_CHANGE_POINTS, AI.KEY_DRIVERS, and AI.CAUSAL_EFFECT to diagnose the cause and measure the business lift from a metric shift using the Austin Bikeshare dataset.

Example: Detecting Change Points

One of the new functions, ML.DETECT_CHANGE_POINTS, automatically identifies statistically significant structural shifts or level changes in time-series data. This function is scalable and can run across millions of individual time series. An example query demonstrates its use to find shifts in daily baseline data, identifying exact time intervals where baselines have changed over time, such as a shift in February 2018 aligning with the Austin City Council's "Dockless Mobility Pilot Program".