Where is data stored in the bronze, silver, and gold layers?

In a typical data lakehouse architecture, data is staged in three progressive layers. The bronze layer holds raw, ingested data in its native format (e.g., JSON, CSV, Parquet) with minimal transformation. It is usually stored in a shared storage system such as an object store (S3, ADLS) or a distributed file system (HDFS). The silver layer contains cleansed, enriched, and semi‑structured data that has undergone basic transformations like schema enforcement, duplicate removal, and type casting. It is often stored as Delta Lake or Iceberg tables, enabling ACID transactions and versioning. The gold layer stores curated, business‑ready datasets that are fully transformed, aggregated, and optimized for analytics. These are typically stored in a data warehouse or a columnar format (Parquet, ORC) within the same lake or in a dedicated warehouse like Snowflake, BigQuery, or Redshift. Each layer builds on the previous one, providing a clear lineage and incremental refinement of data quality.