CareerGuide 1 of 2
Career guide · Guide 1 of 2
Data Engineering Interview Trade-offs: A Practical Guide
The trade-offs interviewers expect Data Engineers to reason about, with how to frame each one: latency vs cost, consistency vs availability, normalisation and more.
Senior interviewers rarely want “the right tool”. They want to hear you name a trade-off, pick a side for this scenario, and say what you give up. Use this structure:
“We could do A or B. A gives us X but costs Y. Because the requirement here is Z, I would choose A, and I would mitigate Y by W.”
The trade-offs that come up most
| Trade-off | Lean one way when | Lean the other way when |
|---|---|---|
| Batch vs streaming | Freshness of hours is acceptable; simplicity and cost matter | Decisions depend on data within seconds or minutes |
| ETL vs ELT | Data must be masked before storage; target cannot compute | Warehouse compute scales; you need to reprocess from raw |
| Normalised vs dimensional | Write-heavy operational data | Read-heavy analytics with simple queries |
| Full refresh vs incremental | Small tables; simplicity | Large tables; cost and runtime matter |
| Exactly-once vs at-least-once + idempotency | Within a system that supports transactions end to end | Across systems (the usual case) |
| Partition by date vs by key | Most queries filter by time | Most queries look up by key (then cluster instead) |
| Managed service vs self-hosted | Small team, speed matters | Strict control, special requirements, very large scale |
| Strong vs eventual consistency | Financial correctness, deduplication | Dashboards that tolerate seconds of delay |
| Precompute vs compute on read | Repeated heavy queries, strict latency | Ad-hoc, rarely used queries |
| Wide denormalised tables vs star schema | One consumer, simple access | Many consumers, shared dimensions |
How to sound senior
- Tie every choice to a requirement stated by the interviewer, or one you clarified.
- Quantify roughly: “50 GB a day is fine for a single daily batch.”
- Name the cost of your choice and how you would mitigate it.
- Say when you would revisit: “If the business later needs 5-minute freshness, I would move ingestion to streaming but keep modelling in batch.”
Common mistakes
- Picking the newest technology without a requirement.
- Listing pros and cons without deciding.
- Absolute statements (“always use streaming”).
Key takeaway
State the options, choose for the scenario’s requirement, name what you give up and how you mitigate it, and say what would change your mind.