Skip to main content

ETL And ELT

Summary: ETL and ELT are two patterns for moving data from source systems into a data warehouse, differing in when the raw data gets transformed relative to when it's loaded.

At a glance​

QuestionAnswer
What is it?Patterns for consolidating scattered data into one warehouse
Why does it matter?It's the layer that actually moves and prepares data before any BI tool can query it
When do I use it?Data lives across multiple source systems (app DB, SaaS tools, APIs, logs) and needs to be centralized

How it works​

  • Extract: pull raw data from source systems (app databases, SaaS tools, APIs, logs).
  • Transform: clean, reshape, join, deduplicate, standardize the data.
  • Load: write the transformed data into the destination, usually a data warehouse.
  • ELT reorders this: load raw data first, then transform it inside the warehouse itself (commonly via SQL/dbt).
  • ELT became common as warehouses like Snowflake and BigQuery got cheap and powerful enough to handle transformation themselves.

Example​

An ETL/ELT tool (Fivetran, Airbyte) pulls data from Stripe, Salesforce, and an app database into one warehouse (Snowflake, BigQuery, Redshift). dbt transforms it inside the warehouse. A BI tool (Superset, Metabase) then connects to that single warehouse to query and visualize the result.

Remember​

  • ETL/ELT tools are what actually consolidate scattered data into one place.
  • BI tools query already-consolidated data — they don't move or merge it themselves.