extractloadtransform

ExtractLoadTransform is a data integration pattern used in modern analytics pipelines. It describes a workflow where data is first extracted from source systems, then loaded into a central destination such as a data warehouse or data lake, and finally transformed within the destination environment to make it suitable for analysis. The pattern aligns with the ELT approach, in which transformation occurs after loading, taking advantage of the target platform's compute resources.

Process and components: Data is extracted from databases, SaaS applications, logs, and other sources. The extracted data is loaded into a landing or staging area inside the destination system. Transformations—implemented via SQL, Spark, or platform-native tools—shape, cleanse, normalize, enrich, and aggregate the data, producing curated tables or views. Metadata and lineage tracking help document source-to-target mappings, transformation rules, and data quality checks.

Advantages and considerations: The approach leverages scalable compute in the destination, often simplifying data pipelines and enabling rapid iteration. It can reduce data movement by avoiding early-stage transformations outside the target. However, it requires strong governance, well-defined transformation logic, and robust monitoring to manage data quality, latency, and cost, since heavy transformations may run in the warehouse or lakehouse.

Use cases and tooling: Common targets include cloud data warehouses and data lakes (for example Snowflake, BigQuery, or Redshift). Transformations are executed through SQL or Spark jobs, orchestrated by tools such as Airflow or Dagster. This pattern is well-suited to data-rich environments with scalable compute and evolving analytical requirements.

Related topics include ETL, ELT, data warehouse, and data lake.