All posts
    Interviews

    Data Analyst Interview Questions: SQL, Metrics and Case Practice

    Practise explaining joins, data quality, metric changes and business recommendations, including what would make your answer wrong.

    RR
    Rachit Rastogi
    Founder, JobMate
    Nov 21, 2026 5 min read

    A useful data analyst answer connects a business question to a reliable dataset, an analysis and a decision. Practise SQL and statistics, but also practise explaining assumptions and checking whether the result answers the original question.

    The questions below are original practice exercises, not a frequency-ranked list or a particular employer's actual questions. O*NET's Business Intelligence Analysts profile provides occupational context for querying data, reporting and identifying patterns; actual data analyst roles vary.

    Reference: O*NET Business Intelligence Analysts

    1. Why did a join increase the number of rows?

    Practice prompt: joining orders to order_items doubles the row count. Explain that one order can have several items, so the result has a different grain. Check uniqueness of the join keys and the expected relationship. For order-level revenue, aggregate items to one row per order before combining with another order-level table.

    Follow-up: what happens if a customer has no orders? Describe the effect of an inner versus a left join and how a WHERE condition on the joined table can accidentally exclude unmatched customers.

    2. How would you calculate a conversion rate?

    Define the eligible population, conversion event and time window before dividing. If the question is 'Which acquisition cohort signed up?', use eligible visitors and signups from that same cohort. Do not divide signups this month by unrelated search clicks this month and call it a conversion rate.

    3. Revenue fell. Where would you start?

    First check whether the data is complete and the definition changed. Then break the result into useful components, such as customers, orders per customer and revenue per order. Segment by a relevant dimension, compare like periods and name the next check that would distinguish your explanations.

    4. How do you handle missing or duplicate data?

    Ask what one row should represent and why data is missing. Separate an absent event from a zero value. Identify a stable key for deduplication and preserve a way to inspect discarded records. Explain when filling values would bias the result rather than treating imputation as an automatic cleaning step.

    5. Can this experiment prove that a change worked?

    Practice prompt: signups rose after a new homepage went live. A before-and-after comparison can reveal a change, but traffic sources, seasonality or another release may explain it. Discuss the assignment process, consistent outcome definition, observation period and uncertainty before making a causal claim.

    6. How would you explain a result to a non-technical manager?

    Start with the decision and one relevant number. Give its population and period, show the comparison and name the main limitation. End with a specific action or test. Keep SQL mechanics available for questions rather than leading with the query.

    7. Walk me through a project

    Use this order: question, dataset, grain, quality checks, method, result and limitation. Show one example where a check changed your answer. For a portfolio dataset, explain how you acquired it and what cannot be concluded from it; do not claim a business adopted your recommendation if it did not.

    Practise with one small dataset

    • ●Write the grain and key before querying.
    • ●Check row counts before and after joins.
    • ●Build one metric with an explicit denominator and window.
    • ●Write a three-sentence recommendation with one limitation.
    • ●Explain which additional data would change your answer.

    Practise business analyst scenarios

    Prepare for a case-study interview

    Frequently asked questions

    What should I practise for a data analyst SQL interview?

    Start with joins, grouping, filtering, NULL handling and window functions where relevant to the role. Explain the row grain and test edge cases, rather than only memorising syntax.

    Do I need Python for every data analyst interview?

    No single tool is universal. Use the vacancy and recruiter instructions to prioritise SQL, spreadsheets, BI tools or Python, and be honest about your level.

    Ready to apply what you read?

    Find matching roles, tailor your resume and let Otto handle supported applications. Start with 10 free automated applications.