Data Analyst Interview Questions: What to Ask and What a Good Answer Contains
A data analyst interview should find out whether the candidate asks what a number is for before computing it, and whether they can tell when a number is wrong. Tool fluency matters less than either; it can be taught in a month.
- Ask about work the candidate owned, not opinions: every question below asks for a system, a deal, a campaign or a decision that actually happened, with what they did and what came of it.
- Score each answer against the note beside it. The notes describe observable things a strong data analyst answer contains, so two interviewers write down the same evidence and reach the same score.
- Use the scorecard: question framing; data scepticism; statistical judgement, and two more. Five criteria, scored 1 to 4, with a sentence of evidence each.
- Ask five questions well rather than eight in a rush. The first one, about something the candidate owned end to end, sets the tone: specifics, not generalities.
- Watch for the red flags listed below, which are patterns rather than single answers: no analysis they can name that changed anything, for one.
The questions
1. Tell me about an analysis that changed a decision. What was the question, what did you find, and what happened?
What a strong answer contains: A real decision-maker, a real question, a finding stated in plain words, and a consequence. The best answers include what the stakeholder originally asked for and how the candidate reframed it.
2. A dashboard number dropped 30% overnight. Walk me through what you do.
What a strong answer contains: Checks the pipeline before the business: did the data arrive, did a definition change, is a source late. Then segments. Strong candidates say they would tell the stakeholder "I am checking" before telling them anything else.
3. How do you decide whether a difference between two groups is real?
What a strong answer contains: Sample sizes, variance, whether the groups were comparable to begin with, and whether the difference matters even if it is real. Knows what a p-value does and does not say.
4. Describe a metric definition that caused an argument. How was it resolved?
What a strong answer contains: A specific metric (active user, churn, conversion), the two readings, and a resolution that was written down somewhere. Someone who has never seen this argument has not worked with metrics long.
5. Write, or talk through, a SQL query for: the top five products by revenue in each region last month.
What a strong answer contains: Window function or a correlated approach, correct handling of ties, and a question about what "last month" and "revenue" mean before writing. Fluency here is expected for the role; the questions are the differentiator.
6. How do you present a finding to someone who has ten minutes and no statistics?
What a strong answer contains: One chart, one sentence, the caveat that matters, and the decision they are being asked to make. Look for empathy with the reader and an example.
7. Tell me about a time your analysis was wrong. How did you find out and what did you change?
What a strong answer contains: Owns it, names the mechanism (a join that duplicated rows, a filter applied twice, a definition mismatch), and has a check that would catch it now.
8. What do you do when the data to answer a question does not exist?
What a strong answer contains: Proxies with stated limitations, a proposal to start collecting it, and honesty with the stakeholder about what can and cannot be known. A candidate who would "estimate" without saying so is a risk.
Scorecard
Score each 1 to 4 with one line of evidence. A total is less useful than a pattern.
- Question framing: asks what the number is for before computing it
- Data scepticism: checks the pipeline and the definitions before the business story
- Statistical judgement: knows when a difference is real and when it matters
- Communication: one chart, one sentence, the caveat that matters
- SQL fluency: correct, readable, with the right questions asked first
Red flags
- No analysis they can name that changed anything
- Trusts a dashboard number without asking where it came from
- Statistical vocabulary without the ability to explain it plainly
- Has never been wrong
How to run the interview
Give the SQL question as a written exercise before the interview and discuss it live. You want to hear the questions they asked about the schema and the definitions, not watch them type. If they wrote a query without asking what revenue means, ask them what it means.
The dropped-dashboard question is the best predictor in the list. Analysts who check the pipeline first are the ones who will not send a panic to the CEO over a late batch job. Score the order of their checks, not their conclusion.
End with the "data does not exist" question. Every analyst faces it weekly, and the honest answer, that some things cannot be known yet and here is how we start knowing them, is the mark of someone leadership can trust.
Frequently asked questions
Should we ask about specific tools like Tableau or dbt?
Ask what they have used and for what; do not score the list. A strong analyst learns a new BI tool in weeks. A weak one with five tools on their CV is still weak.
How technical should the SQL exercise be?
As technical as the job. If the role writes production models, test that. If it mostly reads from a warehouse, a clean query with a window function is enough.
How do we assess business sense in a junior candidate?
Ask about a decision from their own life or studies that they made with data, and about a number in the news they distrust. Judgement shows before experience does.
Related guides
Screening data analyst CVs this week?
Paste the job description into WadeCV, check the rubric it drafts, and upload the CVs. Each one is read against the rubric and scored with an explanation. Ten screenings are free; there is no subscription.
Screen your first CVs free