Medallion Architecture Accelerator: Medallion Architecture Guide

Medallion Architecture Guide Medallion Architecture Accelerator by Datanest Digital Overview The medallion architecture organises a data lakehouse into three logical layers, each providing progressively higher data quality and business value: Layer Purpose Data State Write Pattern Bronze Raw ingestion, source-of-truth archive As-is from source Append / Streaming Silver Cleansed, conformed, deduplicated Type-safe, validated Merge (Upsert) Gold Business aggregations, KPIs, models Aggregated / Dimensional Overwrite / Merge Bronze Layer -- Patterns What belongs here Exact copies of source data with ingestion metadata. Rescue columns for malformed records. Schema evolution history (Auto Loader). Recommended metadata columns Column Description _source_system Upstream system identifier _source_name Logical source name _source_file Input file path or URI _ingestion_timestamp When the row was ingested _ingestion_date Date partition key _row_hash SHA-256 hash for change detection _batch_id Batch