What is ETL (Extract, Transform, Load)

Definition

ETL (Extract, Transform, Load) is a data integration pattern in which data is pulled from source systems, reshaped and cleansed in a dedicated staging layer, and only then written into a target warehouse, so that downstream consumers receive analytics-ready records rather than raw operational data.

« Back to Glossary Index

Delivers conformed, business-ready tables because transformation happens before the load step, not after
Suits regulated environments where sensitive fields must be masked or dropped prior to landing in the warehouse
Keeps warehouse compute lean since heavy reshaping runs on a separate transformation engine
Produces predictable, auditable batch outputs that map cleanly to scheduled reporting cycles

Glossary: ELT (Extract, Load, Transform)

Real World Example

A retail enterprise extracts POS, inventory, and CRM feeds nightly, conforms currencies and product codes in a staging tier, and loads the cleaned result into Azure Synapse to drive executive dashboards and demand forecasts.

FAQs

How is ETL different from ELT?

In ETL the data is transformed before it lands in the warehouse, whereas ELT loads raw data first and transforms it in place using the warehouse's own compute.

When is ETL the better choice?

ETL fits cases with strict pre-load governance, limited target compute, or complex transformations that are easier to manage outside the warehouse.

What tools commonly run ETL workloads?

Engines such as Informatica, Azure Data Factory mapping data flows, Talend, and SSIS are widely used to orchestrate extract, transform, and load stages.