What is ELT (Extract, Load, Transform)

Definition

ELT (Extract, Load, Transform) is a cloud-era integration approach that lands raw source data directly into a scalable store such as a lakehouse or warehouse, then performs transformations in place using the platform's elastic compute, decoupling ingestion speed from transformation logic.

« Back to Glossary Index

Loads data fast by deferring transformation, shrinking the window between source change and availability
Lets analysts re-transform the same raw data repeatedly without re-ingesting from source systems
Exploits elastic warehouse compute, scaling transformation horizontally during peak processing
Preserves an untouched raw layer that supports replay, auditing, and future use cases

Real World Example

A logistics firm streams raw IoT telemetry into a Microsoft Fabric Lakehouse and then runs Spark notebooks against the landed data to derive route-efficiency and fuel-usage metrics on demand.

FAQs

Why has ELT become popular?

Cheap cloud storage and on-demand warehouse compute make it practical to load everything first and transform later, which shortens ingestion time and keeps a reusable raw copy.

Does ELT compromise data governance?

Not inherently, but because raw sensitive data lands first, ELT requires strong access controls and masking policies on the landing layer.

Which platforms are built for ELT?

Snowflake, BigQuery, Databricks, and Microsoft Fabric are designed around loading raw data and transforming it with their native engines.