Fifteen common data architect interview questions
- Warehouse, data lake or lakehouse — how do you choose?
- Design a data platform for a company moving off spreadsheets.
- How do you decide between batch and streaming pipelines?
- What modelling approach would you use for analytics, and why?
- How do you handle slowly changing dimensions?
- What is a data mesh, and when does it make sense?
- How do you enforce data quality across pipelines?
- What does data governance mean in practice?
- How do you handle personal data and regulatory requirements?
- How do you design access control for a shared data platform?
- How do you control cloud data platform costs?
- How would you migrate a legacy on-premises warehouse to the cloud?
- How do you support both BI and machine learning on one platform?
- How do you document and catalogue data?
- Tell me about an architecture decision you'd reverse.
Choosing the platform shape
| Architecture | Strengths | Watch-outs |
|---|---|---|
| Data warehouse | Fast SQL analytics, strong governance, BI-friendly | Less suited to raw, unstructured or ML data |
| Data lake | Cheap storage for any format, ML-friendly | Becomes a swamp without catalogue and quality controls |
| Lakehouse | Open table formats add transactions and schema to lake storage | Operational maturity varies by stack |
| Data mesh | Domain teams own data products | Needs strong platform and governance standards; heavy for small orgs |
A strong answer starts from requirements — users, latency, data types, regulations, team skills and budget — rather than a favoured tool, and often lands on a hybrid. See data warehouse interview questions.
Modelling and pipelines
For analytics, dimensional models with clearly declared grain remain the most usable for BI, often built in layers: raw ingested data, cleaned and conformed data, and business-ready marts. Choose batch for most reporting (simpler, cheaper, easier to backfill) and streaming where decisions depend on fresh data — fraud, operational dashboards, personalisation — accepting more operational complexity. Change data capture from operational databases bridges the two. See data modeling interview questions.
Governance, quality and security
- Ownership: every dataset has a named owner and a documented contract — schema, freshness and quality expectations.
- Quality: automated tests in pipelines for nulls, uniqueness, referential integrity and freshness, with alerts and blocking on critical failures.
- Catalogue and lineage: discoverable datasets with descriptions and column-level lineage so impact of changes is visible.
- Security: role-based and attribute-based access, column masking for personal data, encryption, audit logs.
- Regulation: data minimisation, retention policies, and the ability to locate and delete a person's data on request.
Cost and migrations
Cloud data costs grow quietly. Control them with workload isolation and auto-suspending compute, partitioning and clustering to reduce scanned data, materialising expensive transformations incrementally, tagging spend by team, and reviewing the most expensive queries regularly. For migrations, inventory and prioritise workloads, migrate in slices with parallel runs and reconciliation, retire the legacy system deliberately, and communicate timelines to every consuming team.
Stakeholders
Expect a question about conflicting requirements — finance wants consistency, product wants speed. Good answers separate concerns (certified datasets versus exploratory sandboxes), make trade-offs explicit with owners, and document decisions in architecture decision records.
Practising
Reading answers is not the same as saying them. Pick ten of the questions above and answer them aloud against a timer; a free AI interview assistant like CoPilot Interview can ask data architecture questions in a mock interview and flag answers that ran long or lacked an example. Prepare one platform you designed as a diagram you can redraw in two minutes, with three decisions and their trade-offs. See also data engineer interview questions.
Disclosure: CoPilot Interview is our product, so weigh what we say about it accordingly. Everything else on this page applies whichever tool you use, or none.
Frequently asked questions
What does a data architect interview cover?
Platform architecture choices, data modelling, batch versus streaming, governance, quality, security and privacy, access control, cost management, migrations and stakeholder trade-offs, usually with a deep dive into a past design.
What is the difference between a data lake and a lakehouse?
A data lake stores raw files cheaply in any format. A lakehouse adds open table formats on top, bringing transactions, schema enforcement and performance features closer to a warehouse.
When should you use streaming instead of batch?
When decisions depend on fresh data, such as fraud detection or operational dashboards. Batch is simpler, cheaper and easier to backfill for most reporting.
What is a data mesh?
An organisational approach where domain teams own and publish data as products, supported by a self-service platform and shared governance standards.
Practise on real interview questions
Run 145 real interview questions across 10 tracks, answer out loud and see a structured answer to compare with yours. Free, no signup.
Try a practice question