What does a data analyst do?
A data analyst collects, cleans and analyzes data, then turns it into reports, dashboards and recommendations for the people who make decisions. Most of the job is answering business questions with numbers: what changed, why, and what to do next. The work usually involves SQL, a spreadsheet or BI tool, and plain-language writing.
US government profiles have no entry called data analyst. A search for "data analyst" on O*NET OnLine returns Data Scientists first and Business Intelligence Analysts (15-2051.01) second. The second profile's tasks include generating "standard or custom reports summarizing business, financial, or economic data" for executives, managers and clients, and maintaining "dashboards, systems, or methods." Its software list includes Microsoft Excel, Microsoft Power BI, SAS and SQL database tools.
The Bureau of Labor Statistics page for data scientists covers the closest group. Its duties include collecting, categorizing and analyzing data, using data visualization software to present findings, and making "business recommendations to stakeholders based on data analysis." BLS gives a median annual wage for data scientists of $120,230 in May 2025, a figure for that group, which analyst roles may not match.
In Canada, Job Bank's page for Data Analyst, Informatics and Systems files the title under NOC 21223. It lists a national median of $40.87 an hour, with wages updated November 19, 2025.
How were these data analyst questions chosen?
They come from the duties in the O*NET and BLS profiles, matched to the question each duty tends to produce. They are drawn from the job's published duties, not from a survey of real interviews. Nothing here claims to be the most asked list, or to show what a named company asks.
| Duty (O*NET and BLS) | Question it tends to produce | What the interviewer is checking |
|---|---|---|
| Determine which data are available and useful | "Where would you get the data to answer this?" | Whether you ask what decision the answer feeds |
| Collect, categorize and analyze data | A SQL exercise with joins and grouping | Correct results, and that you check row counts |
| Clean data before analysis | "What would you do with these odd values?" | Whether you investigate before deleting |
| Generate reports summarizing business data | "Walk me through a report you built" | Who used it and what they changed because of it |
| Maintain dashboards and BI tools | "A dashboard number looks wrong. What now?" | A calm order of checks, from source to chart |
| Make recommendations to stakeholders | "Explain this finding to a sales manager" | Plain words, one main point, and a clear next step |
Which behavioral questions should you expect?
Behavioral questions ask for real past work, best told in STAR order: situation, task, action, result. Interviewers listen for the business question behind your analysis, the choices you made with messy data, and what someone did with your answer. A story that ends with "I built a dashboard" and no use of it tends to score lower.
The STAR method guide shows the structure, and the page on using data to make a decision covers the most direct version of these questions.
1. Tell me about an analysis that changed a decision.
What they listen for: the decision, the data, what you found, and what happened next. If the result was small, say so plainly.
Made-up example answer. Situation: a regional manager believed a weekend promotion was driving sales and wanted to extend it. Task: I was asked to confirm the effect before the next quarter's budget. Action: I pulled two years of daily sales, compared promotion weekends with similar weekends without it, and split the results by store. Result: the lift showed up in only a few stores, so the manager kept the promotion there and moved the rest of the budget elsewhere, and I set up a monthly check to track it.
2. Tell me about a time your numbers were wrong.
What they listen for: how you found the error, how fast you told people, and what check you added. Hiding a mistake is the answer that fails. The page on tell me about a mistake you made has a structure for it.
3. Describe a time a stakeholder disagreed with your findings.
What they listen for: whether you rechecked your work, showed your method, and stayed open to their view. Sometimes they know something the data doesn't.
4. Tell me about a messy dataset you cleaned.
What they listen for: the problems you found, the rules you applied, and how you recorded them so others could repeat the work.
What technical data analyst questions come up?
Technical questions test SQL, basic statistics and how you check your own work. Interviewers watch whether you ask what the tables contain, write a simple query first, and confirm the result makes sense. Saying out loud how you would validate a number often counts as much as the syntax.
5. What is the difference between an inner join and a left join?
What they listen for: an accurate definition and a reason to pick one. The PostgreSQL documentation on joins between tables says a left outer join keeps every row of the left table, and "when outputting a left-table row for which there is no right-table match, empty (null) values are substituted for the right-table columns." An inner join keeps only rows that match on both sides.
Made-up example answer: "An inner join returns only customers who have at least one order. A left join from customers to orders returns every customer, with nulls in the order columns for those who never ordered. If I wanted to count customers with no orders, I'd use the left join and filter where the order ID is null. Then I'd check the total count against the customers table so I know no one was dropped or doubled."
6. Write a query that returns monthly revenue by region.
What they listen for: grouping by month and region, handling refunds or nulls, and checking the total against a known figure.
7. How do you handle outliers?
What they listen for: investigating before removing anything. The NIST/SEMATECH handbook's page on outliers defines one as "an observation that lies an abnormal distance from other values" and says "one should try to understand why they appeared" before considering removal.
8. How would you explain a correlation to someone who wants to treat it as a cause?
What they listen for: plain language, an example of a third factor that could explain both, and a suggestion of how to test the cause, such as a controlled comparison.
9. How do you decide which chart to use?
What they listen for: the question the chart answers. Lines for change over time, bars for comparing groups, and a table when people need exact values.
Which situational questions test judgment?
Situational questions describe a problem you have not faced and ask what you would do first. Interviewers want an order of steps: confirm the problem, find the cause, tell the people who rely on the number, then fix it. Asking what decision the number feeds is often the best first move.
The guide to situational interview questions explains how these differ from behavioral ones.
10. The weekly dashboard shows revenue down 30 percent. What do you do?
What they listen for: checking the data pipeline and filters before raising an alarm, then confirming against a second source.
Made-up example answer: "First I'd check whether the data loaded fully, because a failed load can look like a drop. Then I'd compare the dashboard with the finance system for the same week. If the drop is real, I'd break it down by region and product to find where it sits, and I'd tell the dashboard owner what I'd checked so far. If it's a data problem, I'd put a note on the dashboard until it's fixed."
11. Two teams give you different definitions of "active user."
What they listen for: bringing both teams together, writing one agreed definition down, and noting where the old numbers change.
12. A manager asks for a number by end of day that needs a week to do properly.
What they listen for: a rough answer with its limits stated clearly, and a plan for the full version.
What follow-up questions tend to come next?
Follow-ups test whether you understand your own work. Expect the interviewer to ask how you checked a number, what you would do with more time, or how a stakeholder reacted. Prepare by noting, for each story, one check you ran and one thing you would now do differently.
- "How did you know the data was complete?"
- "What would change if the table had a billion rows?"
- "What did the stakeholder do with your answer?"
- "Which part did you do yourself?"
How should you prepare for a data analyst interview?
Practise SQL on a sample database until joins, grouping and window functions feel routine, and say your checks out loud as you go. Prepare four STAR stories about analyses people acted on. Then pick one project and rehearse explaining it in two minutes to someone with no data background.
The free interview question generator lets you paste the job and pick the round, and it returns likely questions, why they come up, and what the interviewer listens for. Analysts often work beside the roles covered on the business analyst, product manager and software engineer pages, so those questions are worth a look.
If your resume lists tools but few results, the guide to a resume with no numbers helps you find them. Prept's Mock room lets you answer out loud on a clock, scores each answer out of 10 and checks it against your resume, and it is included in the free plan.