Records leave a raw data lake and pass through five live quality gates — completeness, accuracy, consistency, timeliness and uniqueness — before settling into bronze, silver or gold storage tiers based on a composite Data Quality Score.
Each record carries five independent quality dimensions that combine into a single score. Gates flash green or red per dimension, and the final score decides whether a record is curated (gold), validated (silver), or quarantined.
Raise the defect rate to simulate messier source data, adjust the ingestion rate, and move the gold threshold to see how the split between gold, silver and quarantine shifts.
Bronze/silver/gold is a common tiered data-lake pattern: bronze holds raw data, silver holds cleaned records, and gold holds curated, ML-ready datasets.