ETL
Abbreviation · Development
Definitions
Extract, Transform, Load — the three steps of moving data between systems. Extract from sources, transform into the desired format, load into the destination. Data engineers spend 80% of their time on the T.
In plain English: The three-step process for moving data: pull it out of the source (extract), clean and reshape it (transform), and put it in the destination (load).
ELT (Extract, Load, Transform) has emerged as a modern alternative where raw data is loaded into a data warehouse first, then transformed using SQL. This shift was enabled by cheap cloud storage and powerful query engines like BigQuery, Snowflake, and Databricks.
Example: 'We switched from ETL to ELT — load everything raw into BigQuery first, then transform with dbt. It's simpler and we can re-transform without re-extracting.'
Source: ELT alternative
Origin Story
The unglamorous plumbing that powers every data warehouse
**ETL** (Extract, Transform, Load) emerged as a concept in the 1970s with the rise of data warehousing, though the acronym was popularized in the 1990s by data warehouse pioneer Ralph Kimball. The idea is straightforward: pull data from source systems, clean and restructure it, then load it into a destination.
ETL became the backbone of business intelligence. Tools like Informatica, Talend, and Microsoft SSIS built multi-billion-dollar businesses around the three-letter acronym. Every company with a data warehouse runs ETL jobs, usually overnight.
The modern trend is **ELT** (Extract, Load, Transform) -- load raw data into a data warehouse first, then transform it using SQL. Cloud warehouses like Snowflake and BigQuery are powerful enough to handle transformations themselves. Tools like dbt (data build tool) embrace ELT, treating SQL as the transformation language.
Coined by: Data warehousing community (Ralph Kimball popularized)
Context: 1970s (concept), 1990s (term popularized)
Fun fact: ETL jobs are notorious for breaking in unexpected ways. Data engineers joke that ETL stands for 'Extremely Tedious Labor.' The field has spawned an entire discipline -- data engineering -- largely because ETL turned out to be far harder than anyone expected.