ETL and ELT

ETL means extract, transform, load; ELT means extract, load, transform. The distinction is the position of the main transformation relative to loading a chosen destination. Name that destination: the same platform may transform before one store and after another. These labels do not determine whether execution is batch or streaming.

Suppose fields prohibited in an analytical warehouse are removed before delivery, while sales aggregation runs inside the warehouse afterward. That path combines pre-load transformation with ELT-style analytical work. ELT need not mean copying every source field unchanged, and ETL can retain a permitted raw copy elsewhere.

Choose per operation from restrictions, engine capability, data volume and measured cost. SQL is common for warehouse transformation, but language alone does not identify the pattern. Recomputing January after a June fix requires suitable January inputs, code and reference data; neither acronym restores evidence that was never retained.

Reference: AWS ETL and ELT comparison.