ETL is an abbreviation for extraction, transformation and loading of data, and denotes the process by which data gets from source systems into a target store in a usable form. Extraction means obtaining data from an interface, database or file; transformation means cleaning it, unifying formats, recalculating and supplementing it; loading means writing it to the target. A common variant is the order in which data is first loaded and only transformed in the target. For a company, it is essential that this flow be repeatable, scheduled and monitored – without an alert on failure, reports look correct even though they contain old figures. The design should therefore include error handling, the ability to rerun without duplicates, and a record of when the data last passed through.
See also: Data Warehouse, Data quality in analytics, Record deduplication.
