spot-check-features

To validate ML feature store data quality before training, apply SQL-based checks to BigQuery tables tracking is_quality_ready, quality_alert_level, and feature_count completeness. This surfaces data quality issues across daily batches to ensure models train on clean data.

Track per-date quality metrics such as matchup_quality_pct, vegas_quality_pct, and feature_count completeness against game_date slices. These metrics detect data quality regressions and readiness across recent batches to guide remediation.

Yes, you can execute SQL queries against BigQuery tables to check feature store readiness for the last 7 to 14 days. Validate is_quality_ready flags, quality_alert_level, and feature_count to surface actionable remediation guidance for your data warehouse.

The best way to detect data quality regressions in an ML feature store is running structured SQL evaluations on daily batches. Comparing per-day quality metrics and alert levels across 7 to 14 day slices identifies readiness drops and guides immediate remediation.

Yes, ML data pipeline validation requires specific BigQuery fields including is_quality_ready, quality_alert_level, feature_count, feature_quality_score, and game_date. These fields are necessary to produce structured evaluations and remediation guidance for daily feature batches.