Launch prices end Friday-d --h --m --s left

Data scientist interview questions

22 questions · 5 answer outlines · Updated Oct 6, 2026

Data scientist interviews usually cover statistics and probability, machine learning concepts, coding in Python or SQL, and a case study or take-home where you frame a problem and defend your modelling choices. Many add a round on explaining results to people outside the team, because a model nobody trusts doesn't get used.

Five questions to prepare first

These come up in most interviews for this role. For each one: what the interviewer is checking, and the points a strong answer covers, in order.

1

Explain the bias-variance trade-off.

What they're checking

Whether you understand why models fail on new data, which underlies most modelling decisions.

A strong answer covers

  1. Define bias as error from over-simple assumptions, which leads to underfitting.
  2. Define variance as sensitivity to the particular training data, which leads to overfitting.
  3. Explain that more complex models usually lower bias and raise variance, and the aim is the lowest error on unseen data.
  4. Name how you manage it in practice, with cross-validation, regularisation, more data or ensembles.
2

How can you tell a model is overfitting, and what do you do about it?

What they're checking

Whether you validate models properly and know the standard fixes.

A strong answer covers

  1. Give the signal, strong performance on training data and much weaker performance on validation data.
  2. Explain how you check, with a held-out set or cross-validation, split by time if the data has a time order.
  3. List fixes, simplifying the model, regularisation, more data, fewer features or early stopping.
  4. Mention a case where you found and fixed overfitting, and how much the validation score changed.
3

Which metric would you use for a classification problem where only 1% of cases are positive?

What they're checking

Whether you choose metrics from the cost of mistakes, rather than defaulting to accuracy.

A strong answer covers

  1. Point out that accuracy misleads here, since predicting no positives at all scores 99%.
  2. Use precision, recall or their balance in F1, and the precision-recall curve rather than ROC when positives are this rare.
  3. Tie the choice to the business cost, such as missing fraud versus annoying a genuine customer.
  4. Add that you would tune the decision threshold for that trade-off, not leave it at 0.5.
4

How would you build a model to predict customer churn?

What they're checking

Your end-to-end judgement, from defining the target to using the model.

A strong answer covers

  1. Define churn and the prediction window precisely, such as no purchase in the next 60 days.
  2. Build features from data available before the prediction date only, such as usage trends, support contacts and billing history.
  3. Start with a simple, explainable model, then try gradient boosting, validating on a split by time.
  4. Explain how it would be used, such as which customers to contact, and how you would measure the effect of acting on it.
5

Tell me about a model you built that ended up in production.

What they're checking

Whether your work reaches real use, and what you learned getting it there.

A strong answer covers

  1. Describe the problem and why a model was the right tool.
  2. Explain your key choices, the data, the features that mattered and the model, briefly.
  3. Cover deployment and monitoring, and how you worked with engineers to ship it.
  4. Give the impact in numbers, and one thing you would do differently.

Statistics and probability

The foundations. Expect follow-up questions that check you understand, not just remember.

  1. 1Explain the bias-variance trade-off.Answer outline
  2. 2What is the central limit theorem, and why does it matter?
  3. 3How would you test whether a coin is fair?
  4. 4What is the difference between a Type I and a Type II error?
  5. 5How do you deal with missing data?

Machine learning

From choosing a model to knowing when it is lying to you.

  1. 6How can you tell a model is overfitting, and what do you do about it?Answer outline
  2. 7Which metric would you use for a classification problem where only 1% of cases are positive?Answer outline
  3. 8Explain how a random forest works.
  4. 9When would you use logistic regression instead of a gradient boosted model?
  5. 10What is regularisation, and how do L1 and L2 differ?
  6. 11What is data leakage? Give an example.
  7. 12How would you explain a model's predictions to a stakeholder?

Case studies

Usually open-ended. Interviewers care about how you frame the problem as much as the model you pick.

  1. 13How would you build a model to predict customer churn?Answer outline
  2. 14How would you measure whether a new recommendation algorithm is better?
  3. 15A model performs well offline but poorly in production. What could be wrong?
  4. 16How would you detect fraudulent transactions?
  5. 17How do you decide whether a problem needs machine learning at all?

Behavioural and stakeholders

Senior interviewers often weigh this round heavily. Bring examples where your work was used.

  1. 18Tell me about a model you built that ended up in production.Answer outline
  2. 19Tell me about a time your analysis was wrong.
  3. 20How do you explain uncertainty in your results to decision-makers?
  4. 21How do you prioritise between requests from several teams?
  5. 22How do you work with engineers to get a model deployed?

Data scientist interview tips

  • Be ready to write Python or SQL live, including data cleaning, not only modelling code.
  • For case studies, spend the first minutes on the target and the data available at prediction time. Leakage is a common trap.
  • Prepare one project you can explain at three depths, a one-line summary, a five-minute version and a deep dive.
  • Practise explaining a model to someone non-technical; many interviews include a stakeholder round.

Data scientist interview FAQ

Usually probability, statistics and the intuition behind common models, rather than proofs. Research roles go deeper into linear algebra and optimisation.

Yes, many companies use one. Treat it like work you would ship: state your assumptions, keep the code readable, and finish with a short summary of results and limitations.

Data scientist interviews put more weight on modelling, experimentation and coding, while analyst interviews focus on SQL, metrics and business questions. The two overlap more at some companies than others.

Related roles

All interview questionsHow a mock interview works

Say these answers out loud before the real interview.

A free mock interview for this role asks five questions you haven't seen, scores each answer and shows a stronger one.

Start with a free mock interview
Start a free mock interview