Interview100x

Interview prep

How to prepare for data and analytics interviews

Data scientist, data analyst and data engineer interviews share a core: write correct SQL quickly, choose the right measure for a question, and explain what the result means for a decision. Beyond that the roles split, with scientists tested on experiments and statistics and engineers on modelling and pipelines. Many loops also include a take-home or a case on realistic data.

See the typical interview loop for this family

Last reviewed November 12, 2026

What these interviews usually test

General guidance for the role family, not a description of any one company's process.
SQL under time pressure
Expect joins, aggregations, window functions and date logic on an unfamiliar schema, often without autocomplete. Read the schema carefully, state the grain of each table, say your plan before you type, and check for duplicates and nulls. Finish by saying in one sentence why the query answers the business question.
Metric choice and explaining a change
You will be asked how to measure success for a product or why a number moved. Define the metric precisely, order a few metrics into a primary measure and guardrails, and name how each could mislead. To diagnose a change, rule out logging and definition problems first, then segment before proposing causes.
Experiment design and causal reasoning
Data science loops commonly test A/B testing in depth. Be ready to choose the randomisation unit, size the test, pick guardrail metrics, and spot problems such as novelty effects or interference between users. Know what you would do when randomisation is not possible, and remember that a p-value informs a decision rather than making it.
Statistics and modelling with a baseline
Modelling questions reward framing before technique. Start from the problem and a simple baseline, choose an evaluation metric that matches the real cost of errors, and check for leakage and untested assumptions. Expect follow-ups on the method behind any tool you name.
Explaining a result to a non-technical partner
Most loops include a round with someone outside data. Lead with the recommendation, then the evidence, then the uncertainty, and drop the jargon. Practise explaining the same analysis at two depths, one for a peer and one for a busy executive.
Take-home hygiene
Read the instructions twice, including any rule on AI use, and time-box the work. Write down your assumptions and how you resolved ambiguity instead of resolving it silently, check data quality before analysing, and end with a clear recommendation. Be ready to defend every line and chart in the follow-up discussion.
Data engineering: models, pipelines and quality
Data engineer interviews add data modelling, pipeline design and reliability. State the grain of each table, explain how a pipeline handles late, duplicate or bad data, and say who consumes the output and what breaks for them when it fails. Explain trade-offs between approaches rather than listing tool names.