ELT (Extract, Load, Transform)

ELT — extract, load, transform — loads raw data into the warehouse or lakehouse first and transforms it there, in SQL, using the destination’s own compute. It became the default when cloud warehouses made storage and compute cheap and elastic: land everything, keep the raw history, and let transformation be versioned SQL (the dbt model) that analysts can read and change.

Keeping raw data means any table can be rebuilt by replay — the same instinct as the bronze layer of a medallion architecture. The trade-off versus ETL is that ungoverned raw data and compute bills both accumulate; ELT shifts the discipline problem downstream rather than eliminating it.

Go deeper: ETL vs ELT: The Same Three Letters in a Different Order