Launch prices end Friday-d --h --m --s left

Data analyst interview questions

22 questions · 5 answer outlines · Updated Oct 6, 2026

Data analyst interviews usually include a SQL test, live or as a take-home, questions on statistics and experiments, a business case where you turn a vague question into an analysis, and a behavioural round. They want to see that you get the numbers right and can explain them to people who don't work with data.

Five questions to prepare first

These come up in most interviews for this role. For each one: what the interviewer is checking, and the points a strong answer covers, in order.

1

What is a window function, and when would you use one?

What they're checking

Whether you know window functions beyond the name, since they solve many analyst problems neatly.

A strong answer covers

  1. Define it as a calculation across related rows that keeps every row, unlike GROUP BY, which collapses them.
  2. Describe the syntax, a function followed by OVER with PARTITION BY and ORDER BY.
  3. Give two or three uses, such as running totals, ranking with ROW_NUMBER or RANK, and comparing a row with the previous one using LAG.
  4. Mention a query you wrote with one and what it replaced, such as a self-join.
2

How would you explain a p-value to a product manager?

What they're checking

Whether you can explain statistics simply without saying something wrong.

A strong answer covers

  1. Give the plain meaning, how surprising a result at least this large would be if the change made no difference.
  2. Say what it is not, the probability that the change works.
  3. Explain the threshold, often 0.05, agreed before the test rather than after looking at the data.
  4. Add that a significant result still needs a meaningful effect size before anyone acts on it.
3

How would you design an A/B test for a new checkout button?

What they're checking

Whether you can plan an experiment that gives a trustworthy answer before any data comes in.

A strong answer covers

  1. State the hypothesis and pick one primary metric, such as checkout completion, plus guardrail metrics like refunds.
  2. Decide the unit of randomisation, usually the user, and check that the groups split as expected.
  3. Work out the sample size from the smallest effect worth detecting, and run whole weeks to cover weekly patterns.
  4. Agree the stopping rule and decision criteria in advance, so nobody stops the test when it first looks good.
4

Weekly active users dropped 10% last week. How would you investigate?

What they're checking

Whether you investigate methodically instead of guessing at causes.

A strong answer covers

  1. Check the drop is real, such as a broken tracking event, a pipeline delay or a changed metric definition.
  2. Segment it by platform, country, new versus returning users and acquisition channel to see where it is concentrated.
  3. Line the timing up against releases, campaigns, outages and seasonal patterns such as holidays.
  4. Form a hypothesis, test it with the data, and report what you found along with what you would do next.
5

Tell me about an analysis that changed a decision.

What they're checking

Whether your work has real impact, and how you get findings acted on.

A strong answer covers

  1. Set out the question the business was asking and why it mattered.
  2. Explain what you analysed and the one finding that changed the picture.
  3. Describe how you presented it, including any pushback.
  4. Say what was decided and what it led to, in numbers if you can.

SQL

Often tested live. Talk through your query before running it, and check the edge cases.

  1. 1What is the difference between an INNER JOIN and a LEFT JOIN?
  2. 2Write a query that returns the second-highest salary in a table.
  3. 3How would you find duplicate rows in a table?
  4. 4What is a window function, and when would you use one?Answer outline
  5. 5What is the difference between WHERE and HAVING?
  6. 6How would you calculate a seven-day rolling average of daily sales?

Statistics and experiments

You don't need to derive formulas, but you do need to use them correctly and explain them plainly.

  1. 7How would you explain a p-value to a product manager?Answer outline
  2. 8How would you design an A/B test for a new checkout button?Answer outline
  3. 9An A/B test shows a significant result after two days. Would you stop it?
  4. 10What is the difference between correlation and causation? Give an example.
  5. 11When would you use the median instead of the mean?
  6. 12What is selection bias, and how can it creep into an analysis?

Business cases

Open-ended on purpose. Structure your approach out loud before you go into detail.

  1. 13Weekly active users dropped 10% last week. How would you investigate?Answer outline
  2. 14Which metrics would you track for a subscription app?
  3. 15How would you tell whether a marketing campaign worked?
  4. 16A stakeholder asks for a dashboard. What do you ask before building it?
  5. 17How would you estimate the size of a market with very little data?

Communication and behavioural

Analysts are judged on the decisions their work changes. Bring examples with outcomes.

  1. 18Tell me about an analysis that changed a decision.Answer outline
  2. 19Tell me about a time your numbers turned out to be wrong. What did you do?
  3. 20How do you present findings to people who don't work with data?
  4. 21What do you do when the data you need doesn't exist yet?
  5. 22Which tools do you use for analysis and visualisation, and why those?

Data analyst interview tips

  • Practise SQL by writing queries out loud on paper or a plain editor. Live tests rarely give you autocomplete.
  • In case questions, ask clarifying questions first, such as how a metric is defined and over what period.
  • Prepare two stories where your analysis changed a decision, with the numbers before and after.
  • Learn to explain p-values, confidence intervals and A/B tests in one or two plain sentences each.

Data analyst interview FAQ

Most cover joins, aggregation with GROUP BY and HAVING, subqueries or CTEs, and often window functions. Senior roles may add performance questions or messier data, such as duplicates and nulls.

It depends on the team. Many analyst roles only test SQL and spreadsheets, while others expect Python or R for cleaning and analysis. Check the job description, and mention it if you have it.

Two or three projects that start from a real question, show your cleaning and analysis, and end with a clear recommendation. The reasoning matters more than the charts.

Related roles

All interview questionsHow a mock interview works

Say these answers out loud before the real interview.

A free mock interview for this role asks five questions you haven't seen, scores each answer and shows a stronger one.

Start with a free mock interview
Start a free mock interview