Google BigQuery ML Contribution Analysis is now generally available, automating the discovery of key change drivers in large-scale multidimensional data
The feature helps users quickly identify the main reasons behind metric changes, reducing the need for manual querying and visualization
Apriori support’s automated support tuning allows users to specify the number of insights (top-k) they want, letting the model set the min_apriori_support threshold automatically
This automated tuning reduces query latency by returning only the most important insights, rather than millions of possible combinations
The new pruning_method option improves insight readability by removing redundant or duplicate insights, ensuring only the most descriptive segments are shown
Expanded metric support now includes the “summable by category” metric, enabling analysis of totals normalized by categorical variables
This new metric helps correct for outliers and compare groups with varying data availability, such as revenue per month across different years
Contribution analysis can be used to investigate business questions, such as understanding declines in sales per user by segmenting data and identifying key contributors
Users define models with MODEL_TYPE=’CONTRIBUTION_ANALYSIS’, set dimension columns, test columns, and contribution metrics for targeted analysis
Output is automatically sorted by contribution, highlighting actionable insights (e.g., sales declines due to referred traffic)