ETL
Extract Transform Load. The data-integration discipline — to be told apart from (yet often combined with) EDI.
Definition
ETL (Extract Transform Load) covers the practices, tools and products designed to copy data from a source to a destination through a transformation phase (typing, mapping, cleansing, aggregation). Historical tools have been Informatica PowerCenter, IBM DataStage, Talend Open Studio; today the dbt + Airflow / Dagster + cloud warehouse (BigQuery, Snowflake, Redshift) stack dominates the analytics segment.
ETL differs from EDI on three axes:
- Scope: ETL is mostly intra-enterprise (ERP → analytics warehouse). EDI is inter-enterprise by construction.
- Syntax: ETL accepts any file or database format. EDI imposes a normalised syntax (EDIFACT, X12, cXML).
- Timing: ETL is often batch (nightly, hourly). EDI is mostly transactional (the order arrives, we process immediately).
Origin
The "ETL" term appeared in the late 1980s in the context of Inmon/Kimball data warehouses. "ELT" (Extract → Load → Transform, with transformation after loading) arose in the 2010s, reflecting the shift of transformations from the ETL tool to the destination database (SQL inside the warehouse).
Example in context
A company receives an EDI order (ORDERS / 850). An EDI connector translates it to JSON and pushes it into the ERP. Overnight, an ETL pipeline extracts all orders of the day from the ERP, transforms them into a dimensional table, and loads them into BigQuery for sales reporting. EDI handled inter-enterprise; ETL handled intra-enterprise.