Short term, restore service: map the new name in the staging layer, rerun the affected partitions, and confirm downstream numbers reconcile. Long term, stop relying on goodwill. Introduce a data contract with the source team, validate schema at ingestion so failures are caught at the boundary with a clear message, and land raw payloads untyped so a rename does not destroy the ingestion path itself.
Why interviewers ask this
Interviewers want to see both incident response and systemic thinking, plus how you work across team boundaries. The strongest answers avoid blaming the upstream team and instead build defenses: schema validation at the edge, contracts with alerting, and a raw layer resilient to change. They also want you to check whether other pipelines share the dependency.
How to structure your answer
- Fix forward in the staging layer and rerun affected partitions.
- Check who else consumes that source and warn them.
- Move schema validation to the ingestion boundary.
- Propose a data contract with versioning and notice periods.
- Make the raw layer tolerant so ingestion never fails on a rename.
Example answer
Immediately it is a small fix: alias the new column in the staging model, rerun the affected partitions, and reconcile a couple of trusted aggregates so I know nothing silently changed meaning as well as name. I also check the catalog for other consumers of that table, because if it broke us it probably broke someone else who has not noticed yet. Then the systemic part. I would rather the failure happen at ingestion with a message saying expected column customer_id, found customer_identifier than three models deep with a null column. So schema validation moves to the boundary, and the raw layer stores the payload as is, which means ingestion keeps working and only the typed layer complains. And I would go and talk to the source team, not to complain but to set up a contract: here are the fields we depend on, here is the alerting, please give us a release notice. We did that at my last place and it converted a recurring 6am problem into a Slack message a week ahead.
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
- What belongs in a data contract, and who enforces it?
- How would you detect a semantic change where the name stays the same?
- How do you version a table so consumers can migrate gradually?
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