ELT: Load Raw Data First, Transform It Later
ELT pipelines load raw data directly into a data lake *before* any transformation. This speeds up ingestion and lets you figure out the data's structure later.
The mental model
Extract, Load, Transform (ELT) is a data pipeline approach that prioritizes speed and flexibility by changing the order of operations. Unlike its predecessor, ETL (Extract, Transform, Load), ELT loads raw data into storage first and worries about transforming it later, at the moment it's needed.
How it works
The process has three stages. First, data is EXTRACTED from its source systems. Second, it is immediately LOADED into a data lake or modern cloud data warehouse in its original, raw format. Third, the TRANSFORMATION happens on-demand when a user or application queries the data. This is often called "schema-on-read," because the structure is applied at query time, not when the data is stored.
When to use it
ELT shines in big data environments that use a data lake. It's ideal when you need to ingest massive volumes of data quickly and when the final use cases or required data structures are not yet known. This gives data scientists and analysts the flexibility to explore the raw data for new insights without waiting for engineering to build a transformation pipeline.
When not to use it
ELT is a poor choice if your data processing engine lacks power. Because transformations run at query time, an under-resourced system will lead to very slow query performance. If you need highly structured, standardized data for consistent reporting and your transformation logic is stable, a traditional ETL approach might be more efficient.
One canonical example
A company wants to analyze user clickstream data from its mobile app. Using an ELT model, they dump the raw JSON event data directly into a cloud data warehouse. Later, an analyst can write a SQL query that parses the JSON, filters for specific events, and joins it with user data on the fly to calculate a new metric. The data was available for querying moments after being generated, not days later.
Interview question
Which scenario is best suited for an ELT data pipeline approach?
- a.An organization requires all incoming data to be rigorously cleaned and standardized into a fixed schema before storage.
- b.A system needs to perform all complex data transformations on a dedicated server before loading to minimize query-time processing.
- c.A business wants to reduce storage costs by only loading highly curated and summarized data into its data warehouse.
- d.A data science team needs to rapidly ingest massive amounts of raw, varied data to explore for unforeseen analytical patterns.Correct
Why? this is the answer
Option D perfectly describes ELT's strengths: rapid ingestion of raw data when use cases are unknown, allowing for flexible exploration (schema-on-read). Option A describes a traditional ETL process where data is transformed and standardized before loading, which is contrary to ELT's 'load raw first' principle.
Just read this? Test yourself on what you have been reading.
Read the original → en.wikipedia.org
- #data engineering
- #data pipeline
- #elt
- #analytics
Put your scrolling time to good use
Learn one idea, try a quiz and save useful cards for revision. Tezvyn makes it easy to learn and stay current in your tech field, a few minutes at a time.
The iPhone app is on the way
We are building it. Until it lands, nothing here is held back from you: every interview card, your saved cards, streaks and the job board all work in Safari, plus hundreds of free practice quizzes of thirty questions each. Sign in and it all carries over to the app the day it arrives.
Want it as an icon? Tap Share at the bottom of Safari, then Add to Home Screen. It opens full screen and the cards you have read stay available offline.
We are hiring for this. Open roles that interview on data engineering — each one lists the topics its interview covers.
See open roles