Data Analyst Interview Questions: SQL, Metrics and Business Cases
Common data analyst interview questions with what a strong answer includes: SQL joins and grain, metric definitions, A/B tests and explaining results.
Data analyst interviews usually mix three things: technical checks (mostly SQL), metric and experiment reasoning, and a project or case discussion. The best answers in all three show the same habit: being precise about what each number means.
SQL questions
"Revenue doubled after you joined the orders table to the events table. What happened?"
A strong answer starts with grain. The orders table has one row per order; the events table has many rows per order. Joining them repeats each order's revenue once per event. Fix it by aggregating events to the order level before the join, or by joining on a unique key. Mention that you would check row counts before and after every join.
"What is the difference between WHERE and HAVING?"
WHERE filters rows before aggregation; HAVING filters groups after aggregation. A good follow-up to prepare: when filtering in WHERE versus HAVING changes the answer.
"How would you find each customer's first purchase?"
Use a window function such as ROW_NUMBER() OVER (PARTITION BY customer_id ORDER BY purchased_at) and keep row 1. Mention how you would handle ties and refunds.
Metric questions
"Daily active users went up 15% but revenue is flat. What would you check?"
- Is the definition of "active" unchanged? A tracking change is the most common cause.
- Which segment grew? New free users, a country, a platform?
- Did conversion or revenue per user fall for the same segment?
- Is there a time lag between activity and revenue?
Interviewers want to see you question the data before explaining the business.
Experiment questions
"An A/B test shows a 3% lift in conversion. Would you ship it?"
Ask about sample size and whether the test ran for full weekly cycles, check the confidence interval rather than only the p-value, look for sample ratio mismatch, and confirm that guardrail metrics such as refunds or retention did not get worse. Then make a decision and state what would change your mind.
Communication questions
"Explain a result to a non-technical stakeholder."
Lead with the decision it supports, give one number with its uncertainty, and say what you did not measure. Avoid method details unless asked.
The project deep dive
Expect at least one question like "Walk me through an analysis that changed a decision." Prepare the question you answered, the data and its limits, the method, the result and what the business did with it. Interviewers often follow up with "How did you validate the data?" and "What would you do differently?"
Practise these questions
The data analyst mock interview page has more scenarios and answer checkpoints, and a free three-question practice. For a full AI interview based on your own resume and target job, start a MockInterview session. You can also read the data analyst practice case.