ETL Extract, Transform, Load
The process that pulls data out of source systems, reshapes it and loads it into a data warehouse.
The transform step is where the work lives: deduplicating, harmonising formats, merging customer records where the same person is spelled three ways, converting currencies. Historically it ran overnight in batch because source systems could not take the load during the day.
Cloud warehouses inverted the order into ELT — load raw data first, transform afterwards with SQL inside the warehouse, since compute became cheap. dbt built an entire discipline around that idea. Usually orchestrated with Airflow or Dagster.