21 interview questions
Data analyst interview questions and answers
Data analyst interviews test whether you can turn a business question into the right data, the right method and a clear recommendation. Expect a SQL exercise, some statistics, a case question about a metric that moved, and questions about how you explain findings to people who do not read charts all day.
Updated
SQL and data handling
01Write a query that returns monthly active users for the last 12 months.
What they want to hear: Practical SQL with dates, distinct counts and a clear definition of the metric.
Answer outline
- Confirm the definition of active: logged in, made a purchase, or any event.
- Truncate the event date to the month and COUNT(DISTINCT user_id) per month.
- Filter to the last 12 complete months, and decide whether to include the current partial month.
- Mention time zones and test or internal accounts that should be excluded.
02How do you handle missing or dirty data?
What they want to hear: Judgment: knowing that the right treatment depends on why the data is missing.
Answer outline
- First find out why it is missing or wrong: tracking bug, optional field, data entry error.
- Measure how much is affected and whether it is random or concentrated in one group.
- Choose a treatment: exclude, fill with a sensible value, flag it, or fix it at the source.
- Document the choice and show how it affects the result.
03You join two tables and the totals are higher than expected. What happened?
What they want to hear: A very common real bug; they want to see that you check the grain of your data.
Answer outline
- Most likely a one-to-many join duplicated rows, so sums are counted more than once.
- Check row counts before and after the join and look for duplicate keys.
- Fix by aggregating before joining, or joining on the full key.
- Add a reconciliation check against a known total.
04How would you find the top five customers by revenue in each region?
What they want to hear: Top-N per group with window functions, a pattern used constantly in reporting.
Answer outline
- Aggregate revenue per customer and region.
- Rank with ROW_NUMBER() or RANK() OVER (PARTITION BY region ORDER BY revenue DESC).
- Filter rank <= 5 in an outer query or CTE.
- State how ties are handled.
05Which tools do you use for analysis, and how do you choose between them?
What they want to hear: Whether you pick tools for the task rather than forcing everything into one.
Answer outline
- SQL for pulling and aggregating data where it lives.
- Spreadsheets for quick checks and sharing small results with non-technical people.
- Python or R for cleaning, statistics and repeatable analysis.
- A BI tool for dashboards that people use regularly.
Statistics and experiments
06Explain the difference between mean and median, and when you would use each.
What they want to hear: Basic statistics applied with judgment, especially for skewed business data.
Answer outline
- The mean is the average; the median is the middle value.
- With skewed data such as income or order value, a few large values pull the mean up.
- Use the median to describe a typical case in skewed data, the mean when totals matter.
- Often report both, plus a distribution chart.
07How would you set up and analyse an A/B test?
What they want to hear: Sound experiment design, not just reading a dashboard that says 'significant'.
Answer outline
- Define the hypothesis, one primary metric and guardrail metrics before starting.
- Randomise users, not sessions, and calculate the sample size needed to detect a meaningful effect.
- Run for the planned duration, covering full weekly cycles; avoid stopping early when it looks good.
- Check that groups are balanced, then report the effect with a confidence interval, not just a p-value.
08What is a p-value, in plain words?
What they want to hear: Whether you understand it well enough to avoid the common misreading.
Answer outline
- Assuming there is no real effect, it is the probability of seeing a result at least as extreme as the one observed.
- It is not the probability that the hypothesis is true.
- A small p-value says the data would be surprising if there were no effect; it does not say the effect is large or important.
- Pair it with effect size and a confidence interval.
09What is the difference between correlation and causation? Give an example.
What they want to hear: Protection against the most common analytical mistake in business decisions.
Answer outline
- Correlation means two things move together; causation means one changes the other.
- A third factor can drive both: users who use a feature may simply be more engaged already.
- To show causation, use an experiment, or carefully designed comparisons when you cannot run one.
- Say how confident you are and why.
10Your A/B test shows a positive result on the main metric but a drop in a guardrail metric. What do you recommend?
What they want to hear: Balanced judgment about trade-offs, not a mechanical 'ship it'.
Answer outline
- Check the result first: sample size, novelty effect, a bug in one variant.
- Quantify both effects in business terms where possible.
- Look at segments to see who gains and who loses.
- Recommend a decision with the trade-off stated, or a follow-up test.
Metrics and business cases
11Daily active users dropped 10% yesterday. How do you investigate?
What they want to hear: A structured approach to a vague, urgent question.
Answer outline
- Check whether the data is right: tracking changes, pipeline delays, a broken dashboard.
- Check for known causes: holidays, outages, releases, marketing changes.
- Segment the drop by platform, country, app version, new versus returning users.
- Narrow to the segment that explains most of the drop and share findings early with what is still unknown.
12How would you define success metrics for a new feature?
What they want to hear: Linking metrics to the goal of the feature, not tracking everything.
Answer outline
- Start from what the feature is supposed to change for users and the business.
- Choose one primary metric, such as adoption or task completion.
- Add guardrails that should not get worse, such as retention or support tickets.
- Set a baseline and an expected change before launch.
13How would you calculate customer retention?
What they want to hear: Knowing standard definitions and the choices that change the result.
Answer outline
- Group users into cohorts by signup month.
- For each cohort, measure the share still active in month 1, 2, 3 and so on.
- Be explicit about the definition of active and whether retention is exact-month or rolling.
- Show it as a cohort table or retention curves and compare cohorts over time.
14Revenue is up but profit is down. Where would you look?
What they want to hear: Business sense and the ability to break a metric into its parts.
Answer outline
- Break profit into revenue minus costs and see which costs grew.
- Check the mix: lower-margin products, regions or customers growing faster.
- Look at discounts, refunds and acquisition costs.
- Summarise the main driver with numbers and a recommendation.
Visualisation and communication
15How do you choose the right chart?
What they want to hear: Whether your charts answer a question quickly instead of decorating a slide.
Answer outline
- Start from the message: trend, comparison, distribution or relationship.
- Line charts for trends over time, bar charts for comparisons, histograms for distributions, scatter plots for relationships.
- Avoid pie charts with many slices and dual axes that mislead.
- Put the takeaway in the title so the chart reads without you.
16How do you build a dashboard people actually use?
What they want to hear: Focus on decisions and users rather than packing in every metric.
Answer outline
- Ask who uses it, how often, and which decisions it supports.
- Show a few key metrics at the top, with detail and filters below.
- Keep definitions consistent and documented.
- Check usage after launch and remove what nobody looks at.
17How would you explain a complex finding to a non-technical audience?
What they want to hear: Communication that leads to action, which is what the job is for.
Answer outline
- Lead with the answer and the recommended action.
- Support it with one or two clear charts and plain numbers.
- Explain limitations and confidence in simple terms.
- Leave the method in an appendix for those who want it.
Working as an analyst
18Tell me about an analysis that changed a decision.
What they want to hear: Real impact: that your work leads to action, not just reports.
Answer outline (STAR)
- The question and the decision at stake.
- Your approach and the key finding.
- How you presented it and to whom.
- The decision that changed and the result.
19Describe a time your analysis showed something a stakeholder did not want to hear.
What they want to hear: Integrity and tact when the data contradicts expectations.
Answer outline (STAR)
- The expectation and what the data showed.
- How you checked your work before presenting.
- How you delivered the message and handled pushback.
- What happened next.
20Tell me about a time you had to work with a vague request.
What they want to hear: Whether you clarify the real question before spending days on the wrong analysis.
Answer outline (STAR)
- The request as it was given.
- The questions you asked to find the decision behind it.
- How you scoped the work and agreed a deadline.
- What you delivered and how it was used.
21How do you prioritise many data requests from different teams?
What they want to hear: Managing workload with impact in mind and keeping stakeholders informed.
Answer outline
- Ask for the decision and deadline behind each request.
- Prioritise by business impact and urgency, and agree the order with your manager.
- Offer quick answers or self-serve options for simple requests.
- Tell people when to expect results, and update them if it changes.
How to prepare
- Practise SQL with joins, aggregations and window functions until you can write them without looking anything up.
- Prepare a portfolio project or two that start from a business question and end with a recommendation, not just charts.
- For case questions, always check the data first, then segment, then form hypotheses.
- Learn the basics of A/B testing well: sample size, significance, confidence intervals and common pitfalls.
- Practise explaining one of your analyses in two minutes to someone outside data.
Practice these questions out loud with NORDON
Reading answers is not the same as saying them. NORDON's interview practice asks you questions written for the role, listens to your spoken answer and shows you how to make it stronger.
- Upload your CV in the Knowledge base, so questions and feedback use your real background.
- Open Practice, enter the role (for example data analyst), paste the job ad if you have it, and choose 3, 5, 8 or 10 questions and the answer language.
- Listen to each question and answer out loud, as you would in the room. You can also type an answer.
- Get a score from 1 to 10, what worked, what to improve and a stronger answer to rehearse. It uses only what you said and your documents, and leaves a [placeholder] where a fact about you is missing.
- Try the question again, or move on. For a dress rehearsal, run a mock call where an interviewer voice asks the questions out loud.
The free plan includes 10 scored practice answers a month and mock calls of up to 10 minutes, no card needed. Paid plans include more practice and mock calls of up to 30 minutes. How practice works
Practice is for preparing. Use live help in a real interview only where the interviewer allows AI assistance, never in an interview, test or assessment whose rules forbid it.
Preparing for your data analyst interview
What does a data analyst interview process usually include?
Often a recruiter call, a SQL or technical screen, a case or take-home analysis, and interviews with the team and stakeholders. The mix varies by company, so ask what each stage covers.
How should I approach a take-home analysis task?
Restate the question, list your assumptions, check data quality first, and keep the analysis focused. Present a clear recommendation up front, with the method and limitations after it.
Do I need Python or R for a data analyst role?
Not always. Many analyst roles focus on SQL, spreadsheets and a BI tool. Python or R is a plus and is required in some teams, so check the job description.