Data Glossary 🧠

Search

Search IconIcon to open search

What is ETL?

Last updated Oct 25, 2022 - Edit Source

ETL (Extract, Transform, and Load) is a paradigm for moving data from one system to another. It involves reading data (Extract) from one system, modifying the data (Transform), and then sending it (Load) to a destination system. The ETL paradigm emerged in the 1970s.

A key feature of ETL is that data is transformed before being sent to the destination, as demonstrated in the following image:

However in recent years, the preferred data movement paradigm has shifted to ELT (Extract, Load, and Transform). The ELT philosophy dictates that data should be untouched – apart from minor cleaning and filtering – as it moves through the extraction and loading stages so that the raw data is always accessible in the destination Data Warehouse. See ETL vs ELT for a comparison of these approaches.

# ETL is Changing

The way we do ETL is changing. For a long time ETL was done with tools such as Informatica, IBM Datastage, Cognos, AbInitio, or Microsoft SSIS. Today we use more programmatic or configuration-driven platforms like Airflow, Dagster, and Temporal.

Historically ETL was once preferred over ELT for the following no-longer-valid reasons: