Mock Interview Copilot Interview questions by role
Data analyst interview questions
Data analyst interviews check whether you can get correct numbers out of messy data and then turn them into a decision someone else can act on. Expect SQL questions, a case about a metric that moved, questions on cleaning and validating data, and at least one question about explaining a finding to people who do not work with data.
How the interview usually runs
A common pattern is a recruiter screen, a SQL test (live, or in an online assessment), and then a case or take-home: here is a dataset or a business question, what would you look at? Final rounds usually include a stakeholder-style conversation with a manager from the business side.
Technical skill gets you through the test, but the case and stakeholder rounds decide most offers. Interviewers want to hear you ask what decision the analysis is for, check the data before trusting it, and say what you found in a sentence a busy manager would understand.
Common questions and what the interviewer is checking
Write a query to find each customer's most recent order.
What they are checking: Practical SQL: joins, grouping, and window functions such as ROW_NUMBER, plus whether you think about ties and customers with no orders.
Sign-ups dropped last week. How would you investigate?
What they are checking: A structured approach: confirm the drop is real (tracking, definitions), segment it (channel, device, region), check for releases or outside events, and form a hypothesis before digging.
How do you clean a dataset you have just been given?
What they are checking: That you check for duplicates, missing values, odd ranges, inconsistent categories and join problems, and that you record what you changed.
How would you define a good metric for customer engagement?
What they are checking: Whether you tie metrics to the business goal, think about how they could be gamed or misread, and pair a headline metric with a guardrail.
Tell me about an analysis that changed a decision.
What they are checking: Impact. They want the question, what you found, how you communicated it and what the business did differently because of it.
How do you explain a finding to someone with no data background?
What they are checking: Plain language, the conclusion first, one clear chart, and honesty about uncertainty without burying the message in caveats.
What is the difference between correlation and causation? Give an example from work.
What they are checking: That you would not tell a manager a feature caused growth just because both went up, and that you know what an experiment or control group adds.
A stakeholder disagrees with your numbers. What do you do?
What they are checking: Calm, evidence-led handling: reconcile definitions and filters first, since most disagreements are about what is being counted, then walk through the method together.
Which tools do you use for dashboards and why?
What they are checking: Not loyalty to a tool, but whether you build dashboards people actually use: clear definitions, few charts, and an owner who maintains them.
How to structure a strong answer
For case questions, think out loud in a fixed order so the interviewer can follow you. For experience questions, lead with the business question and end with the decision it changed, not with the query you wrote.
- Clarify the question and the decision it supports.
- Check the data: definitions, tracking, quality.
- Break the problem down by segment or step in the funnel.
- State the finding in one sentence, then what you would recommend and how sure you are.
Example answer
Sign-ups dropped last week. How would you investigate?
First I would check that the drop is real. I would compare the dashboard with the raw events, and ask whether anything changed in tracking or in how a sign-up is defined. In a previous role a similar drop turned out to be a tag that stopped firing on one browser after a site update.
If the data holds up, I would segment it by channel, device, country and landing page to see whether the drop is everywhere or concentrated. Concentrated drops usually have a specific cause, like a paused campaign or a broken form on mobile. Then I would check the release log and the marketing calendar for that week. I would report back with the size of the drop, where it sits, the most likely cause, and what we would need to confirm it.
Mistakes to avoid
- Jumping into a query before asking what the analysis is for or how the metric is defined.
- Trusting a dataset without checking for duplicates, gaps or tracking changes.
- Presenting ten charts when one chart and a sentence would make the point.
- Claiming that one thing caused another from a correlation alone.
- Describing your tools in detail and never mentioning what the business did with your work.
Frequently asked questions
What questions are asked in a data analyst interview?
Usually SQL questions, a case about a metric that changed, questions on cleaning and validating data, defining metrics, a past analysis that influenced a decision, and how you explain findings to non-technical colleagues.
How hard is the SQL in a data analyst interview?
It varies by company, but joins, aggregation, filtering, subqueries and window functions come up most. Practise writing queries without autocomplete and explaining what each part does.
How do I answer a data analyst case question?
Clarify the goal, check whether the data can be trusted, break the problem into segments or funnel steps, form a hypothesis, and finish with a recommendation and what you would need to confirm it. Thinking out loud is expected.
What if I do not have work experience as a data analyst?
Use projects with real, messy public data, and talk about them the same way: the question, how you cleaned and checked the data, what you found and what someone could decide from it.