Data Analyst Interview Questions & Answers 2026

Data analyst hiring varies widely between a bank, a retailer, a start-up and a big technology company, but most processes follow the same broad shape. Knowing the shape lets you prepare for each stage separately instead of revising everything at once.

Some employers, particularly in finance, insurance and the public sector, add an online aptitude stage before any human contact. If you are applying to a large organisation, expect a numerical reasoning test and sometimes a verbal reasoning test early on. Our guide to online assessment tests explains how those filters work.

SQL is the single most commonly tested technical skill in analyst interviews. You are usually given a small schema on a shared screen or in a browser-based tool and asked to write queries while explaining your thinking. Speed matters less than correctness, clear structure and spotting edge cases.

Questions you should be able to answer cold

• What is the difference between INNER, LEFT, RIGHT and FULL joins? An inner join keeps only matching rows; a left join keeps every row from the left table and fills unmatched right-side columns with NULL; a full join keeps unmatched rows from both sides. Be ready to say which you would use to find customers who have never placed an order (a left join filtered on a NULL key).
• What is the difference between WHERE and HAVING? WHERE filters rows before aggregation; HAVING filters groups after aggregation. You cannot use an aggregate function such as COUNT in a WHERE clause.
• How would you find duplicate records? Group by the columns that should be unique and keep groups where COUNT(*) is greater than one. A window function with ROW_NUMBER can also flag the duplicates so you can delete all but one.
• What do window functions do? They calculate across a set of related rows without collapsing them into one row. Know RANK versus DENSE_RANK versus ROW_NUMBER, LAG and LEAD for period-on-period comparisons, and running totals using SUM with an OVER clause.
• How do you get the second-highest salary or the top three products per category? Use DENSE_RANK or ROW_NUMBER partitioned by category and filter on the rank in an outer query or CTE. This pattern appears in a large share of interviews.
• What is the difference between UNION and UNION ALL? UNION removes duplicates and so requires extra work; UNION ALL keeps everything and is faster. Use UNION ALL unless you specifically need de-duplication.

You will not be asked to derive proofs, but you will be expected to use statistical terms correctly and to explain them to a non-technical colleague. The most common topics are listed below.

• Mean, median and mode — when does each mislead? The mean is pulled by outliers, so income, order values and session durations are usually better summarised with the median. State which one you would report and why.
• What is a p-value? The probability of seeing a result at least as extreme as the one observed if the null hypothesis were true. It is not the probability that the null hypothesis is true, and a small p-value says nothing about whether the effect is large enough to matter.
• Explain correlation versus causation. Two metrics moving together does not show one causes the other; a third variable, reverse causality or coincidence can explain it. Randomised experiments are the cleanest way to establish causation.
• How would you design and read an A/B test? Define one primary metric and a hypothesis, calculate the required sample size from the baseline rate and the smallest effect worth detecting, randomise at the right unit (user rather than session), run for whole weeks, and avoid peeking and stopping early. Report the effect size and confidence interval, not just significance.
• What are Type I and Type II errors? A Type I error is a false positive (rejecting a true null); a Type II error is a false negative (failing to detect a real effect). Lower significance thresholds reduce the first but raise the second.

Case questions test whether you can turn a vague business problem into a structured analysis. Typical prompts are: “Orders fell 10% last week — how would you investigate?”, “How would you measure the success of a new feature?” and “Which customers should we target with a retention offer?”

A reliable four-step structure

• Clarify: Ask what the metric means, over what period, compared with what, and who is asking. Many candidates skip this and solve the wrong problem.
• Break it down: Decompose the metric into drivers — for orders, that is traffic × conversion × basket size, split further by channel, device, region, product and customer type.
• Check the boring explanations first: Tracking changes, outages, seasonality, a public holiday, a pricing change or a data pipeline delay explain a surprising share of sudden movements.
• Recommend and quantify: Finish with what you would do next, how confident you are and what would change your mind.

Expect questions on the tools listed in the job description and on how you keep analysis trustworthy. Be honest about your level: claiming expertise in a tool you cannot demonstrate is quickly exposed in a live exercise.

If your role includes spreadsheet work, review the typical tasks in our Excel skills test guide, and if it includes programming, see the coding assessment guide.

Analysts spend much of their time persuading people who do not share their technical background, so behavioural questions carry real weight. Answer them using the STAR technique and quantify the outcome wherever you can.

• “Tell me about an analysis that changed a decision.” Choose a project with a clear business question, your specific method and a measurable result. Say what the decision would have been without your work.
• “Describe a time you explained complex findings to a non-technical audience.” Show that you started from their question, removed jargon, used one or two visuals and checked they understood the implication.
• “Tell me about a time your data was wrong or your analysis contained an error.” Own the mistake, explain how you found it, what you told stakeholders and what check you added afterwards. This is the same shape as our guide to “tell me about a time you failed”.
• “How do you prioritise when several teams want analysis at once?” Describe asking about the decision each request supports, its deadline and its value, then agreeing priorities openly rather than quietly working through a queue.
• “Tell me about a time you disagreed with a stakeholder’s interpretation of the data.” Show that you listened, tested their hypothesis with data and stayed respectful. See conflict interview questions for more structure.

Many of these overlap with the most common interview questions, so prepare your standard answers to “tell me about yourself” and “why this company” alongside the technical material.

Three focused weeks is enough for most candidates who already have working knowledge of SQL and a spreadsheet tool. If you are starting from scratch, double the timeline and spend the extra time on SQL.

If the employer also runs psychometric tests, use our free practice tests to keep your numerical speed sharp while you work on the technical material.

Good questions show that you think like an analyst: curious about data quality, impact and how decisions get made. Our guide to questions to ask at an interview has a longer list; these suit analytics roles particularly well.

• “What does the data stack look like, and who owns data quality?” Reveals whether you will spend your time analysing or fixing pipelines.
• “Can you give an example of an analysis that changed a decision recently?” Shows whether insights are used or filed away.
• “How do analysts work with product or commercial teams day to day?” Embedded and centralised teams offer very different working lives.
• “How will success in this role be measured after six months?” Gives you the criteria you will actually be judged on.

Send a short thank-you email after the interview that references one specific problem you discussed; it is a simple way to stay memorable.