ResumeSkip

Interview questions

Data Scientist interview questions

Data science interviews span statistics, a coding or SQL screen, a machine-learning discussion, and a case study on framing a business problem. Teams want rigor and the judgment to know when a simple answer beats a sophisticated one.

What interviewers evaluate

Statistical reasoning, experiment design, SQL and coding fluency, modeling judgment, and the ability to turn a fuzzy business question into a measurable one.

Behavioral questions

Tell me about an analysis that changed a decision. How did you make it land?

What they're really testing: Impact and communication matter as much as the model; insight nobody acts on is worthless.

Describe a time your first hypothesis was wrong. What did the data tell you?

What they're really testing: Intellectual honesty and following the evidence are core to the craft.

How do you explain a model's limitations to a non-technical stakeholder?

What they're really testing: Overclaiming erodes trust; they want someone who communicates uncertainty well.

Role-specific questions

How would you design an A/B test for this feature, and how do you know it worked?

Why they ask it: Experiment design, power, and pitfalls like peeking are bread-and-butter here.

Explain the bias-variance trade-off with an example from your work.

Why they ask it: A fundamental that reveals whether modeling intuition is real or memorized.

Write a SQL query to find the second-highest value per group.

Why they ask it: SQL screens filter fast; window functions and joins are expected fluency.

How do you handle imbalanced classes or missing data in a model?

Why they ask it: Practical data problems separate people who ship models from those who only study them.

A metric dropped 15% overnight. How do you investigate?

Why they ask it: A diagnostic case tests structured thinking and knowledge of confounders.

When would you choose a simple model over a complex one?

Why they ask it: Judgment about interpretability, maintenance, and marginal gains signals seniority.

How to prepare

  • Drill SQL window functions, joins, and aggregation until they're automatic under time pressure.
  • Review hypothesis testing, p-values, confidence intervals, and common A/B-test traps.
  • Prepare one project story framed as business problem → approach → measurable impact.
  • Be ready to reason about metrics: how you'd define success and detect a broken metric.
  • Practice explaining a model to a skeptical, non-technical stakeholder in plain language.

Frequently asked questions

How much coding versus statistics should I expect?

Most loops include a SQL or Python screen and at least one statistics or experiment-design round, plus a case study. The mix leans analytical, not pure software engineering.

Do I need deep machine-learning theory?

You should understand core algorithms, evaluation metrics, and trade-offs well enough to justify choices. Many roles value strong statistics and product sense over cutting-edge ML.

What's the case study round really testing?

Whether you can turn an open-ended business question into a measurable problem, choose the right method, and communicate what the answer means for a decision.

Prep pairs with tailoring: How to Tailor a Resume to a Job Description.

Get questions built from your real resume and the actual posting, free