Two Orderings
Moving data from sources into an analytics store involves three steps: extract, transform, and load. The order of the last two defines the pattern.
ETL
In ETL you transform data in a separate processing engine before loading it.
- Data is cleaned and reshaped before it lands in the warehouse.
- Only the final modeled data is stored.
- It suits older warehouses with limited compute or strict schemas.
ELT
In ELT you load raw data first, then transform inside the warehouse using its own compute.
- Raw data lands quickly, then SQL or tools like dbt reshape it in place.
- The warehouse scales transforms with cheap elastic compute.
- Raw history is kept, so you can re transform later when logic changes.
Modern cloud warehouses such as Snowflake or BigQuery make ELT the common default because storage is cheap and compute is elastic.
Key idea
ETL transforms data before loading while ELT loads raw data and transforms inside an elastic warehouse, which is why cloud platforms with cheap storage favor ELT.