Data analyst interview questions
Data analyst interviews test four things: whether you can turn a vague business question into something answerable, whether your SQL and spreadsheet work is sound, whether you notice when data is wrong, and whether you can explain a result to someone who will not read your query.
The last one decides more outcomes than candidates expect.
What is actually assessed?
Technical screening covers SQL, spreadsheets and usually one of Python or a reporting tool. The rest of the interview is about judgement: what you do when the question you were given is not the question that needs answering, and what you do when the numbers look wrong.
| Area | Typical question |
|---|---|
| Scoping | A manager asks why sales dropped last month. What do you do first? |
| SQL | Write a query for the top customers by revenue this quarter, excluding refunds. |
| Data quality | Your daily numbers doubled overnight. How do you investigate? |
| Communication | Explain a result you found to someone with no technical background. |
| Judgement | Tell me about an analysis you got wrong. |
How do you answer the vague question?
Ask what decision the answer will drive before you write any query. Interviewers set these questions deliberately loose to see whether you clarify or start calculating. Rushing to a number without knowing what it is for is the most common failure in the whole interview.
A good answer names the decision, states what you would check first, and says which cut of the data you would look at before the rest. Saying out loud what you would rule out early is a strong signal.
The SQL round
Expect joins, grouping, filtering and window functions, usually against a small schema you are shown on the spot. Talk while you write. An interviewer can follow a correct answer they watched you reason through far more easily than a silent one that appears at the end.
- Confirm what a row means in each table before joining anything.
- State your assumptions about duplicates and nulls out loud.
- Say what you would do differently on real data at scale.
- If you are stuck, say what you are trying to achieve. Interviewers usually help.
How do you handle data that looks wrong?
Describe a real check sequence rather than saying you would investigate. Did the definition change, did a pipeline run twice, did a source system start sending something new, is it real. Analysts who assume the number is right are more expensive to employ than analysts who are slow.
Explaining to a non-technical audience
Lead with what it means and what should change, then support it. Candidates who narrate their method first, and reach the conclusion last, lose the room in the first thirty seconds. This is worth rehearsing out loud because it is the opposite of how analysis is usually written up.
This is the round most often failed by technically strong candidates. Practise saying one finding in two sentences with no jargon.
Tell me about an analysis you got wrong
Answer it honestly with a real case. Say what you concluded, what was wrong, how it was caught, and what you changed about how you work. A candidate who cannot name a single mistake reads as either inexperienced or not paying attention.
Questions to ask them
- Who asks the questions this role answers, and how are they usually phrased?
- How much of the job is building reporting versus one-off analysis?
- How trusted is the data today, and what is the worst part of the stack?
- What would a good first ninety days look like?
Giving the opening round once
The technical rounds differ by company. The opening conversation almost never does. Just One Interview covers that first round once, scores it, and takes it to teams hiring analysts, so you spend your effort on the rounds that actually vary.