← Back to the glossary
Architecture models
ETL / ELT
Extract, Transform, Load
ETL transforms data before loading it into the destination. ELT loads first and uses the destination platform to transform it afterward.
- When it applies
- They apply when data from operational systems must reach a platform, warehouse, or analytical layer through a controlled model.
- What it is confused with
- They do not define frequency, latency, or quality by themselves. A pipeline may be batch or frequent and still use either approach.
- Architecture consequence
- The decision determines where transformation rules live, who operates them, and which version of the data can be considered reusable.
Related concepts
- Data Prep
- Kafka
- Reverse ETL
- CDW