19 questions
Scoring rubric 1–5

Data Analyst Interview Questions to Ask Candidates

Q: What interview questions should I ask a data analyst candidate?

Beyond the technical test: what they did before starting a vague request, a time the data was wrong and what they checked now, a surprising result and what they did, a finding the audience did not want, and an analysis that changed a decision. Those four areas — framing, data quality, communication, impact — are where analysts succeed or fail.

Q: Should I ask SQL questions in the interview?

Note this trivia. A live exercise on a realistic dataset, where they talk through their approach, tells you far more than questions about joins. Keep the interview for judgment and communication, which no exercise can measure.

Most data analyst interviews test SQL and skip the part that determines whether the analysis will be used: how the candidate frames a vague question, what they do with bad data, how they present a result to someone who does not want to hear it, and whether anything changed because of their work. Test the technical skills with a short exercise; use the interview for judgment and communication.

Try Voicit free Book a demo

4.8 out of 5 350+ reviewsUpdated: August 2026

Chosen by over 1,000 companies to document their interviews

Technical screens tell you whether a candidate can write the query. They do not tell you whether they will ask what the query is for, notice that the data is wrong, or explain the result in a way a manager will act on.

Those are the failures that make analysts frustrating to work with, and the 19 questions below are built to find them before you hire. Each one says what a strong answer contains and what should worry you.

19interview questions

4question sets by competency

1–5scoring rubric for every answer

Before you ask: separate the skills test from the interview

Give the technical assessment its own slot — a take-home or a live exercise on a realistic dataset, scored with a rubric. Then use the interview for the four things a test cannot measure. The table shows how we split it.

SQL, Python/R, BI tools, statistics

Best toolA 60-90 minute exercise on your own kind of data

What to look forCorrect, readable, and they state assumptions

Framing a vague question

Best toolInterview

What to look forThey ask what decision the answer is for before touching data

Data quality and judgment

Best toolInterview + exercise

What to look forThey noticed something wrong and said so

Communication

Best toolInterview + a 5-minute presentation of their exercise

What to look forOne clear finding, caveats included, no wall of charts

Impact

Best toolInterview

What to look forSomething changed because of an analysis they did

Framing the question

Tell me about a request that came to you as «can you pull some numbers on X». What did you do before you started?

What a strong answer shows: They asked what decision the numbers were for, who would use them, and by when — and can say how that changed the analysis.
Red flag: They pulled the numbers.

Describe an analysis where the question you were asked was not the question that needed answering.

What a strong answer shows: A specific case, how they noticed, the conversation with the requester, and what they answered instead.
Red flag: They have never disagreed with a request.

How do you decide how much rigour a question deserves? Give me an example of a quick answer and one of a deep one.

What a strong answer shows: A criterion (reversibility, cost of being wrong, how the answer will be used) and two real examples.
Red flag: Everything gets the full treatment, or nothing does.

Tell me about a time you had to say «we cannot answer that with the data we have». What happened?

What a strong answer shows: They said it clearly, proposed what could be answered or what data to collect, and the requester accepted it.
Red flag: They produced an answer anyway and buried the caveat.

Data quality and judgment

Tell me about a time the data was wrong and you did not notice until later. What happened, and what do you check now?

What a strong answer shows: A real miss, its consequence, and a specific check they added.
Red flag: They have never shipped an analysis with bad data. Everyone has.

Describe a result that looked too good or too surprising. What did you do?

What a strong answer shows: They distrusted it, checked the pipeline and the definition, and either found the bug or confirmed it with a second method.
Red flag: They presented it. Surprising results are the ones that get shared.

How do you handle a metric that two teams define differently?

What a strong answer shows: They surfaced the conflict, got the definitions written down, and pushed for one owner — with a real example.
Red flag: They used whichever definition the requester preferred.

Walk me through an analysis where you had to make assumptions because the data was incomplete. What were they, and did you state them?

What a strong answer shows: Explicit assumptions, stated in the deliverable, with a sense of how much the answer depended on them.
Red flag: The assumptions were made and not written down.

Communication and stakeholders

Tell me about presenting a result that the audience did not want to hear. How did you do it, and what happened?

What a strong answer shows: They led with the finding, brought the evidence, acknowledged the discomfort, and the result was still used — or they can say why it was not.
Red flag: They softened it until it was not really a finding.

Show me — describe — the last dashboard or report you built. Who used it, and how do you know?

What a strong answer shows: A specific audience, evidence of use (a decision, a recurring meeting, usage data), and something they removed because nobody used it.
Red flag: They built it and assumed it was used.

How do you explain uncertainty to a non-technical stakeholder? Give me an example.

What a strong answer shows: A concrete technique (ranges, plain-language confidence, what would change the answer) and a real conversation.
Red flag: They present point estimates as facts, or drown the stakeholder in caveats.

Describe a time a stakeholder kept asking for more cuts of the data instead of deciding. What did you do?

What a strong answer shows: They named the pattern, asked what decision was pending, and set a limit.
Red flag: They kept producing cuts.

Tell me about feedback on your communication that changed how you present.

What a strong answer shows: Specific feedback, quoted, and a visible change.
Red flag: Nothing comes to mind.

Impact and ways of working

Tell me about an analysis that changed a decision. What was decided differently because of it?

What a strong answer shows: A specific decision, the person who made it, and what would have happened otherwise.
Red flag: They cannot name one. Analyses that changed nothing are the norm; you want someone who notices.

Describe an analysis you were proud of that nobody used. Why, and what did you learn?

What a strong answer shows: An honest cause (wrong question, wrong timing, wrong audience) and a change to how they scope work now.
Red flag: It was the stakeholders’ fault.

How do you prioritise when three teams each think their request is urgent?

What a strong answer shows: A criterion, a visible queue, and a conversation where they said no or later.
Red flag: First come, first served, or whoever shouts loudest.

What do you do to make your work reproducible? Give me an example of it paying off.

What a strong answer shows: Version control, documented queries, a notebook someone else could run — and a time it saved them or a colleague.
Red flag: Their work lives in a spreadsheet on their laptop.

What is a tool or technique you learned in the last year, and what did you use it for?

What a strong answer shows: Something specific, applied to real work, with an honest view of its limits.
Red flag: A list of courses with nothing applied.

Tell me about the analysis you would redo if you could. What would you change?

What a strong answer shows: Self-critical, specific, and about the method or the framing rather than the tooling.
Red flag: Nothing they have done needs redoing.

How to score the answers

Score the exercise and the interview separately and do not let a strong SQL test compensate for weak framing or communication — that is the analyst who produces correct answers to the wrong questions. Use the same scale for every candidate and write the evidence next to the score.

1
No example
Talks about tools and methods, no real analysis described
2
Generic example
An analysis, but no framing conversation, no data problem, no decision
3
Solid example
A real request they reframed, a data issue they caught, a result someone used
4
Strong example
Above, plus a miss they own, uncertainty explained well, an analysis nobody used and why
5
Exceptional
Above, plus a decision they can show changed because of their work

FAQ

Frequently asked questions

What interview questions should I ask a data analyst candidate?

Beyond the technical test: what they do before starting a vague request, a time the data was wrong and what they check now, a surprising result and what they did, a finding the audience did not want, and an analysis that changed a decision. Those four areas — framing, data quality, communication, impact — are where analysts succeed or fail.

How do I assess a data analyst’s technical skills if I am not technical?

Use a realistic exercise on your own kind of data with a rubric written by someone technical, or borrow a colleague for that part. Then ask the candidate to present their exercise in five minutes to you. Whether you understood the finding is itself a valid test.

What are red flags when interviewing data analysts?

They pull the numbers without asking what decision they are for; they have never shipped bad data; surprising results get presented rather than checked; findings get softened for the audience; they cannot name a decision their work changed; and their work lives in a spreadsheet on their laptop.

Should I ask SQL questions in the interview?

Not as trivia. A live exercise on a realistic dataset, where they talk through their approach, tells you far more than questions about joins. Keep the interview for judgment and communication, which no exercise can measure.

What is the difference between a data analyst and a data scientist interview?

The judgment and communication questions on this page apply to both. For a data scientist, add questions about model choice, validation, and what happened when a model was deployed — and weight the technical exercise more heavily. For an analyst, weight framing, data quality and stakeholder work.

Keep reading

Interview questions to ask candidates: the complete guide Problem solving interview questions Communication interview questions Interview scorecard template (free Excel, Word and PDF)

Stop writing up interviews by hand.

Voicit records the interview, transcribes it and fills your scorecard against the competencies you defined, so you can listen instead of typing.

Try Voicit free Book a demo

4.8 out of 5 350+ reviews

Data Analyst Interview Questions to Ask Candidates

Before you ask: separate the skills test from the interview

SQL, Python/R, BI tools, statistics

Framing a vague question

Data quality and judgment

Communication

Impact

Framing the question

Tell me about a request that came to you as «can you pull some numbers on X». What did you do before you started?

Describe an analysis where the question you were asked was not the question that needed answering.

How do you decide how much rigour a question deserves? Give me an example of a quick answer and one of a deep one.

Tell me about a time you had to say «we cannot answer that with the data we have». What happened?

Data quality and judgment

Tell me about a time the data was wrong and you did not notice until later. What happened, and what do you check now?

Describe a result that looked too good or too surprising. What did you do?

How do you handle a metric that two teams define differently?

Walk me through an analysis where you had to make assumptions because the data was incomplete. What were they, and did you state them?

Communication and stakeholders

Tell me about presenting a result that the audience did not want to hear. How did you do it, and what happened?

Show me — describe — the last dashboard or report you built. Who used it, and how do you know?

How do you explain uncertainty to a non-technical stakeholder? Give me an example.

Describe a time a stakeholder kept asking for more cuts of the data instead of deciding. What did you do?

Tell me about feedback on your communication that changed how you present.

Impact and ways of working

Tell me about an analysis that changed a decision. What was decided differently because of it?

Describe an analysis you were proud of that nobody used. Why, and what did you learn?

How do you prioritise when three teams each think their request is urgent?

What do you do to make your work reproducible? Give me an example of it paying off.

What is a tool or technique you learned in the last year, and what did you use it for?

Tell me about the analysis you would redo if you could. What would you change?

How to score the answers

Frequently asked questions

Keep reading

Stop writing up interviews by hand.

USE CASES

PRODUCT

RESOURCES

LEGAL