ETL / ELT

Detailed definition

ETL (Extract, Transform, Load) and ELT (Extract, Load, Transform) are two approaches for moving data from source systems into a data warehouse. In ETL, data is cleaned and transformed in a separate layer before it is loaded into the target; this is typical of classic, mostly on-premise warehouses. In ELT, raw data is loaded first and the transformation runs inside the target using its own compute; this has become the norm in modern cloud warehouses such as BigQuery, Snowflake and Databricks. Which to choose depends on data volume, source variety, compliance/privacy needs and your team's tooling: ELT offers flexibility and scale in high-volume, cloud-first scenarios, while ETL fits cases where transformation must happen before loading. Both approaches require attention to data quality, lineage and reliable, scheduled data pipelines. ETL/ELT flows are often fed by API integrations across systems and turn data from sources like an ERP into something usable for reporting and analytics.