Course contentsShow
← Data Science
Lesson 1815 of 2,14543. Data Pipelines and OrchestrationFree lesson

What is ETL? Extract, Transform, Load Explained

Define ETL, its three stages, and the traditional workflow of transforming data before loading it into storage.

What is ETL? Extract, Transform, Load Explained

You'll learn the fundamental ETL workflow and how data moves from raw sources into analytics-ready storage systems.

The ETL Concept

ETL stands for Extract, Transform, Load—a traditional data integration pattern that moves data from source systems into a data warehouse or analytics platform. Think of it like meal preparation: you gather ingredients from different stores (extract), prep and cook them (transform), then arrange the finished dishes on serving plates (load).

The Three Stages

Extract

Pull raw data from various source systems—databases, APIs, flat files, streaming feeds, or third-party services. This stage focuses on reading data without modifying it, handling connection protocols, authentication, and incremental vs full extractions.

Transform

Apply business logic, data cleansing, and structural changes before loading. This is where the heavy lifting happens: filtering rows, joining datasets, aggregating metrics, standardizing formats, handling nulls, converting data types, and applying business rules. Transformations occur in a staging area or processing engine separate from both source and destination.

Load

Write the cleaned, transformed data into the target system—typically a data warehouse optimized for analytics. The data arrives query-ready, pre-aggregated, and conformed to your star or snowflake schema.

Traditional vs Modern

The "T before L" sequence is ETL's hallmark: transform data before storing it. This approach made sense when storage was expensive and compute power was concentrated in ETL servers. It ensures only clean, validated data enters your warehouse, but requires upfront schema decisions and can be inflexible when business needs change.

Key Takeaway: ETL is a workflow where data is extracted from sources, transformed into analytics-ready format in an intermediate layer, then loaded into a data warehouse—optimizing for storage efficiency and query performance over flexibility.