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
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.”
- SSituation
Marketing planned to raise spend on the channel with the most sign-ups.
- TTask
Check whether those sign-ups were worth the spend.
- AAction
I joined sign-up data to 90-day retention by channel and showed that one smaller channel retained users at twice the rate.
- RResult
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
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.
Answer on camera
Huru asks the questions and records your answers, so you rehearse saying them, not reading them.
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.
Data Scientist interview: questions
What questions are asked in a data scientist interview?
How do I practise for a data science interview?
Go deeper on the Huru blog
Mock interview questions for other roles
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.