ETL or ELT for this platform?
Whether the candidate treats this as a question about where compute lives and what history is retained, rather than as a fashion.
The situation behind the question
Interviewers ask this because it happened to them.
A team is designing ingestion for a new source. One engineer wants to clean and conform the data before it lands; another wants to land it raw and transform it in the warehouse. Both have done it their way before and both worked.
A strong answer
Flags
Green flags
- Starts from what the consumer needs and what must be reprocessable, not from which acronym is current.
- Separates "where does compute run" from "what history is retained", and knows the second is the consequential one.
- Names privacy and minimisation as the legitimate reason to transform before landing.
- Preserves raw and replayable data as the default position and can defend the exceptions.
Red flags
- Argues for one based on which is modern, with no reference to reprocessing or retention.
- Cannot say what is lost when data is cleaned on the way in.
- Believes ELT means no transformation discipline, rather than transformation in a different place.
- Ignores that landing raw has a privacy and retention consequence that has to be designed for.
Follow-ups
Where the conversation goes if the first answer holds up.
- The source contains free-text fields customers type into. Does that change where you transform?
- You land raw and the transformation is wrong for six months. Walk me through the repair.
- What does keeping raw cost you, and how would you decide the retention period?