13.3 ETL & ELT
ETL and ELT are the two patterns for moving data into an analytics platform. ETL transforms before loading; ELT loads then transforms in-place. The cloud era flipped the default from ETL to ELT. This lesson explains why.
ETL vs ELT Flow
Figure 5.10 - ETL transforms en-route; ELT lets the warehouse do the work.
When Each Wins
| Choose ETL When | Choose ELT When |
|---|---|
| Target store is expensive per query (legacy DW) | Target is elastic and cheap (Snowflake, BigQuery) |
| Regulatory rules forbid raw landing | Raw landing is allowed and useful for audit |
| Sources are tiny and clean | Sources are large or messy |
| You want one tool, one workflow | You can split EL (Fivetran, Airbyte) from T (dbt) |
Modern Stack
The dominant 2026 pattern: **Fivetran or Airbyte for EL, dbt for T, Snowflake or BigQuery as the store, Airflow or Dagster as the orchestrator, Great Expectations or dbt tests for quality.**
Discussion
Loading…