Huru

Data Scientist mock interview questions

Data science interviews test statistics and modelling, but the deciding question is usually whether you can turn analysis into a decision someone acted on, and explain it to people who do not work with data.

What interviewers look for

Statistical judgement
Knowing which method fits, and its limits.
Business sense
Starting from the decision the analysis supports.
Experimentation
Designing tests that answer the real question.
Communication
Explaining results to non-technical people.

The questions to practise

Grouped by type. The note beside each says what the interviewer is really testing, which is what your answer has to show.

Behavioural

  • Tell me about an analysis that changed a decision.

    What it tests
    Impact, not just output.
  • Describe a model that did not perform as expected.

    What it tests
    Debugging and honesty.
  • Tell me about explaining a result to a non-technical audience.

    What it tests
    Communication.
  • Describe a time the data was messy or incomplete.

    What it tests
    Pragmatism.

Technical

  • How would you design an A/B test for a new feature?

    What it tests
    Experiment design and power.
  • Explain overfitting and how you guard against it.

    What it tests
    Fundamentals.
  • How do you choose between precision and recall?

    What it tests
    Linking metrics to costs.
  • A test result is significant but tiny. Do you ship?

    What it tests
    Practical versus statistical significance.
  • How would you predict customer churn?

    What it tests
    End-to-end problem framing.

Situational

  • A stakeholder wants the analysis to support a conclusion already made. What do you do?

  • Your model works offline but hurts a metric in production. What next?

  • You have a week to answer a question that needs a month of data. How do you approach it?

For the questions every interview asks, whatever the role, see the full list of mock interview questions.

An example answer, told with STAR

Situation, task, action, result. The structure is what to copy; the story should be yours.

“Tell me about an analysis that changed a decision.”

  1. Situation

    Marketing planned to raise spend on the channel with the most sign-ups.

  2. Task

    Check whether those sign-ups were worth the spend.

  3. Action

    I joined sign-up data to 90-day retention by channel and showed that one smaller channel retained users at twice the rate.

  4. Result

    Budget shifted towards the smaller channel, and paid conversions rose the next quarter.

Mistakes to avoid

  • Describing the model in detail but not the decision it served
  • Quoting a metric without its baseline or uncertainty
  • Answering technical questions without stating assumptions

Practise it in Huru

  1. Choose what to practise

    Pick your role from 242+ career roles, or paste the job title and description to build a custom interview for that job.

  2. Answer on camera

    Huru asks the questions and records your answers, so you rehearse saying them, not reading them.

  3. Fix one thing and go again

    Every answer is scored out of 100, with what worked, the one thing to fix and a stronger version.

Start practising free

Data Scientist interview: questions

What questions are asked in a data scientist interview?
Statistics and probability, machine learning concepts, experiment design, a case or take-home problem, SQL or coding, and behavioural questions about impact and communication.
How do I practise for a data science interview?
Practise explaining your projects out loud in business terms: the question, the method, the result and the decision it changed. Huru scores each spoken answer on relevance, structure, specifics, impact and articulation.

Practise your data scientist interview out loud

Answer on camera and get every answer scored, with the one thing to fix and a stronger version. Huru is free for everyone.