ETL transforms data before loading it, usually on a separate compute cluster. ELT loads raw data into the warehouse first, then transforms it there with SQL. The shift happened because cloud warehouses made storage cheap and compute elastic, so keeping raw data and transforming in place is faster to iterate on, easier to audit, and lets you reprocess history without re extracting from source systems.
Why interviewers ask this
This checks whether you understand the economics behind modern data stacks rather than just the tool names. Interviewers want you to mention that ELT preserves raw data so transformations become reproducible, and that it is not universally better: heavy unstructured processing, tight privacy requirements, or expensive per query billing can all push you back toward transforming before load.
How to structure your answer
- Define both by where the transform runs.
- Explain the economics that drove the shift to ELT.
- Emphasize keeping raw data for replay and auditability.
- Give the cases where ETL still wins.
- Connect it to how you structure layers in the warehouse.
Example answer
The difference is just where the transform happens. ETL transforms in flight, so what lands in the warehouse is already modeled. ELT lands raw and models it inside the warehouse with SQL. The reason ELT took over is that storage got cheap and warehouse compute got elastic, so there is no longer a good reason to throw away the raw payload. That matters more than people expect. When we found a bug in our revenue logic, we rebuilt eighteen months of marts from raw in about forty minutes, without going back to the source APIs, some of which only retain ninety days. It also keeps transformation logic in version control as SQL rather than in a proprietary tool. That said, I would still transform before load for things like decoding large binary files, or when I need to drop or tokenize personal data before it ever touches the warehouse, because you cannot un load sensitive data once it is there.
Walking into this interview soon? GhostPilot listens to your live call, spots the question the moment it is asked, and puts a structured answer on your screen in real time. Try it on your next mock, or grab a $29 Session Pass, no subscription, for the real thing.
See how it worksFollow-up questions to expect
- Where would you mask personal data in an ELT pipeline?
- How do you structure raw, staging, and mart layers?
- When would per query billing change your answer?
Related data engineer questions
Your interviewer will ask their own version of this. Paste your actual job description into the free Question Predictor and get the 20 questions that role is most likely to ask, with what each one is really probing.
Predict my questions