Data
Updated 2026-10-18
Practice data analyst interview questions for screen, manager, SQL and metrics, behavioral, and final rounds.
Analyst interviews test whether you can turn a messy question into a figure someone can use. A strong answer names the decision, the source, the definition you argued for, and the mistake you almost shipped. A tour of every chart you have ever built does not do that work.
Keep a few projects ready. For each one, know the audience, the grain of the rows, the filter that changed the result, and what the partner did after they saw it. If the work was a class file or a take-home, say so before you describe the chart. People can hear when a dashboard never had a user.
When you practice out loud, talk as if the listener has not seen the picture. Define the metric in one sentence before you defend it. If you would check nulls, duplicates, or a timezone before you trusted the figure, say that check out loud. That habit is the job.
Try a free spoken mock for this job: 3 questions, feedback that quotes your own answers. No card.
Start the free mockThe recruiter is sorting for tools, domain, and whether you want this kind of seat. Be specific about SQL, a spreadsheet, a BI tool, and the questions you have actually answered. You do not need to perform a query yet. You do need to sound like you have sat with someone who wanted an answer before a Thursday meeting.
1. What questions have you answered with data, and who used the result?
What they are checking: They want a real audience and a decision, not a list of product names.
Follow-up: Where did the heavy lifting happen, and which part did you still finish by hand?
2. How comfortable are you writing SQL to join, filter, and aggregate, and where do you still look things up?
What they are checking: They are placing you against the team's daily work and listening for an honest gap.
Follow-up: Tell me about a query that was wrong the first time. What was the bug?
3. Why this analyst seat, and which kind of work do you want less of?
What they are checking: They want to hear that you understand ad hoc requests, standing dashboards, and partner work.
Follow-up: If the first months were mostly repairing broken definitions, would you still want the job?
Managers have been burned by a pretty chart with a quiet definition problem. Show that you push back when the number and the question do not match, and that you can say what you do not know yet. Bring one story where you slowed a decision down on purpose.
1. People were quoting a metric whose definition was shaky. What did you do about it?
What they are checking: They want the definition, who was misled, and a correction that did not embarrass the room for sport.
Follow-up: What did you change in the chart or the note so the same mix-up would be harder later?
2. A partner needs an answer today and the clean version will take longer. How do you respond?
What they are checking: They are listening for a scoped first cut, a caveat you say out loud, and a promise you can keep.
Follow-up: What would you refuse to include in that early figure?
3. You and a product or operations partner disagreed about what to measure. How did that end?
What they are checking: They want the decision the metric was supposed to serve and how you landed on one definition.
Follow-up: Where did you write that definition so a new teammate could reuse it?
Expect a table, a SQL prompt, or a critique of a metric. Narrate the grain, the join keys, and the filter before you talk about the chart. If you forget a function name, describe the operation in words. If you are ignoring an edge case, say so. The point is the logic, not a performance of syntax.
1. How would you find customers who show up more than once when they should be a single person?
What they are checking: They want keys, messy fields such as email and name, and a plan for the leftover rows.
Follow-up: How would you show the effect of your rule before anyone merged records for real?
2. A leader asks whether a recent change helped. What do you need before you will answer?
What they are checking: They want the outcome, the comparison, the time window, and the other shifts you would look for.
Follow-up: What would you say if the only file you had mixed old behavior with new behavior?
3. A daily number dropped and someone wants the reason before standup. How do you investigate?
What they are checking: They are listening for a sequence: tracking, the pipeline, a mix shift, then a real product change.
Follow-up: Which check do you run first so you do not explain a bug as if people had changed?
These questions look soft. They are not. Analysts sit between teams that want different stories from the same table. Practice a story where you were late or wrong, and where you repaired trust with a clearer artifact rather than a quieter one.
1. A report you owed went out late. What did you tell the requester, and what did you change about scoping?
What they are checking: They want a warning sent before the meeting, a smaller deliverable, and a habit you kept.
Follow-up: Which signal do you now treat as a reason to cut the ask?
2. Your analysis was wrong after people had already used it. How did you correct the record?
What they are checking: They want a fast correction to the same audience, not a quiet edit and a hope that nobody noticed.
Follow-up: Which check would have caught the error before you sent the file?
3. A stakeholder keeps asking for slices that do not change the decision. How do you handle the third request?
What they are checking: They want a polite redirect back to the decision, plus the one cut you are still willing to run.
Follow-up: How do you leave a note so the next analyst is not starting from zero?
They are deciding whether you will be careful with their numbers. Ask how definitions are stored, who can change a metric, and what the first messy dataset actually is. Say how you like to work without turning it into a demand list.
1. How would you spend the first weeks learning which numbers this team actually trusts?
What they are checking: They want you in the warehouse, in the notes, and in conversation with the people who use the figures.
Follow-up: What would you leave alone until someone asked, and what would you flag right away?
2. What do you still need to know about the data, the tools, or the partners before you would accept?
What they are checking: They are listening for a practical question about access, definitions, or how fast requests arrive.
Follow-up: If the stack were older than you hoped, what would you need in order to do careful work anyway?
Talk through the joins, the filters, and the grain in sentences. Name the function if you remember it, and describe the step if you do not. Offer to write the query if they hand you a table. Do not recite syntax you cannot explain when they change the question.
Say that before you describe the file. Then cover the question, the rows you were given, the assumption you had to make, and what you would have asked a real partner. Pretending a class file was a company dashboard is an easy way to lose the room.
Ask what decision the metric is for, restate the definition in your own words, and confirm it. You are graded on whether you clarify. You are not graded on already knowing their nickname for a revenue line.
Bring one or two examples you can explain without opening a laptop, and be ready to share a link if they ask. Walk through the question, the picture, and the decision. A folder of dashboards you cannot narrate will not help.
Separate the figure from the story they told. Explain the definition you would defend, who you told, and what you changed in the chart or the note beside it. Do not mock the leader. Show that you protect the meaning of the number.
Practice questions written by Round Rehearsal. Not any employer's actual questions.
Try a free spoken mock for this job: 3 questions, feedback that quotes your own answers. No card.
Start the free mock