Data Analyst Interview Questions & Answers (2026)

Statistics & analysis

How do you handle missing data? Name three strategies and their trade-offs.
What's the difference between correlation and causation? Give a real example.
When is the median a better measure than the mean?
How would you identify outliers, and would you always remove them?

Tools & visualization

How do you decide which chart type to use?
VLOOKUP vs INDEX/MATCH — why might you prefer one?
Walk me through a dashboard you built. What decision did it drive?

Case & behavioral

Revenue is down this quarter. What data do you pull first?
Tell me about an analysis that changed a business decision.
How do you present a finding a stakeholder doesn't want to hear?

How to approach: for case questions, state the metric, the segments you'd cut by, and the hypothesis you'd test. Always close with the "so what" — the business action.

What hiring managers screen for

Analyst loops weight three things heavily: SQL fluency (can you get the right data without hand-holding?), analytical judgment (do you ask the right question before running numbers?), and communication (can you make a stakeholder act on your finding?). The last one is where many technically strong candidates lose — a correct analysis nobody understands changes nothing.

How to answer a "metric dropped" case

Clarify the metric definition and the time window.
Segment: by channel, device, geography, and new vs returning users.
Check for data issues first — a broken pipeline looks like a real drop.
Form a hypothesis, then pull the cut of data that would confirm or kill it.
Recommend an action and the follow-up you'd monitor.

Mistakes that quietly fail analysts

Writing a query that returns the right shape but the wrong grain (double-counting via joins).
Confusing correlation with causation in the readout.
Removing outliers without asking whether they're errors or signal.
Presenting a chart without the one-sentence "so what" a decision-maker needs.

Practice explaining a finding to a non-technical audience in three sentences. That skill is what turns an analyst into a trusted partner.

Worked answers to the questions that decide analyst loops

"WHERE vs HAVING." Keep it concrete: "WHERE filters rows before aggregation; HAVING filters groups after aggregation. So WHERE amount > 100 filters individual transactions, while HAVING SUM(amount) > 1000 filters customers whose total exceeds a thousand. You can't put an aggregate in WHERE — that's the classic mistake." A one-sentence example beats a textbook definition every time.

"Revenue is down this quarter. What data do you pull first?" Show structure: "I'd confirm the metric definition and the comparison window first, then decompose revenue into its drivers — number of customers times average order value times purchase frequency. Decomposing tells me whether we lost customers, they're spending less, or buying less often. Then I segment by product line, region, and channel to localize it before recommending anything." Decomposition signals analytical maturity.

"How do you present a finding a stakeholder doesn't want to hear?" Lead with the business impact, not the chart. "I'd open with what it means for their goal, show the evidence simply, acknowledge the uncertainty honestly, and come with a recommended next step rather than just bad news. People accept hard findings when you respect their goal and give them a path forward."

Round-by-round: the analyst loop

A typical loop is a recruiter screen, a SQL screen (often live), and an onsite that mixes a deeper SQL round, a case or business-scenario round, sometimes an Excel or spreadsheet exercise, a dashboard or visualization discussion, and a behavioral round. Many companies use a take-home dataset. The take-home rewards clarity — a clean analysis with a clear recommendation beats a kitchen-sink of charts.

What a strong analyst answer looks like

Strong candidates get the grain right — they never double-count by joining carelessly. They distinguish correlation from causation in the readout, they question whether outliers are errors or signal before removing them, and they always end with the "so what." They treat the stakeholder's decision as the goal, not the query.

Weak candidates write technically correct SQL that answers the wrong question, or produce a beautiful dashboard that drives no decision. The job is judgment plus communication, with SQL as the table-stakes tool.

How expectations differ by company and team

Marketing analytics roles probe funnel metrics, attribution, and campaign measurement. Product analytics roles probe retention, cohorts, and experimentation. Finance and ops analytics roles probe forecasting and variance analysis. Some teams live in SQL and a BI tool; others expect Python or R for deeper work. Read the stack in the job description and rehearse the domain it implies.

Frequently asked questions

How good does my SQL need to be? Comfortable with joins, aggregations, window functions, CTEs, and subqueries — fast and correct under light pressure. This is the most common filter.

Do I need Python? For many analyst roles, strong SQL plus a BI tool is enough. Python helps for heavier analysis and is increasingly expected at product-led companies.

How do I prep for the case round? Practice decomposing metrics out loud and stating the segments you'd cut by. Always finish with a recommendation and the follow-up you'd monitor.

What's the fastest way to stand out? Communication. The analyst who can turn a query result into a clear, decision-ready sentence is the one who gets hired and promoted.

An expanded question bank by theme

Broaden your reps with these commonly asked prompts. Practice the approach out loud, not just the syntax.

SQL: Calculate month-over-month and year-over-year growth. Find the top three products per category. Compute a running total and a moving average. Identify customers who churned and came back. Write a funnel conversion across four steps. Pivot rows to columns. Find gaps in a sequence of dates.

Statistics and analysis: When is the mean misleading? How do you test whether two groups differ? What's a confidence interval in plain terms? How do you handle seasonality in a trend? What's Simpson's paradox, and why should an analyst care?

Business cases: Conversion dropped on mobile but not desktop — what do you check? A marketing campaign looks successful — how do you confirm it caused the lift? Which customers should we target for a retention offer? How would you forecast next quarter's revenue?

Tools and communication: How do you decide between a bar, line, and scatter chart? What makes a dashboard actually used versus ignored? How do you summarize a complex analysis in one sentence for an executive?

Follow-up questions interviewers love

After your first answer, expect: "Is that correlation or causation?" "What's the grain of that table — could you be double-counting?" "What would you do if the data looked wrong?" "How confident are you, and why?" "What's the action you'd recommend?" The follow-ups separate analysts who run queries from analysts who drive decisions.

A realistic two-week study plan

Days 1–4: SQL fluency. Drill joins, aggregations, window functions, CTEs, and the themed prompts above until they're fast and correct.
Days 5–7: Statistics and analytical judgment. Practice explaining correlation versus causation, outlier handling, and seasonality with plain examples.
Days 8–10: Business cases. Practice decomposing metrics and stating the segments you'd cut by, always ending with a recommendation.
Days 11–12: Visualization and communication. Rehearse the one-sentence "so what" for several analyses.
Days 13–14: Mock interviews and, if there's a take-home, a timed practice run that prioritizes a clear recommendation over chart quantity.

The day before and the day of

The night before, review your SQL patterns and a few business-case structures rather than cramming new syntax. On the day, for every query confirm the grain before you write, and for every case state the metric, the segments, and the hypothesis before diving in. Close each answer with the business action it implies. The analyst who turns numbers into a clear decision is the one who gets hired.

How to turn this question list into real readiness

A list of questions is raw material, not preparation. The candidates who convert practice deliberately, and the method is the same regardless of role: focus on SQL accuracy, analytical judgment, and decision-ready communication.

Start by answering out loud, never silently. Comprehension and recall under pressure are different skills, and only spoken practice builds the second. Record yourself so you can hear the filler words, the hedging, and the moments where your structure falls apart — things you never notice while speaking.

Then score yourself against a simple rubric: was the answer structured, specific, and relevant to what was asked? Did it land on a concrete result or trade-off? Rebuild the weakest answers and run them again. A useful daily rep is to write a query, then say the one-sentence business takeaway it produces.

Use spaced repetition rather than a single cram. Three short sessions across a week beat one long session the night before, because the goal is durable recall under stress, not short-term familiarity. Finally, simulate pressure with at least two timed mock interviews before the real thing — pressure changes how you think, and you want to have felt it before it counts.

A final pre-interview checklist

Run through this the day before:

Do you confirm the grain of a table before writing a query, to avoid double-counting?
Can you decompose a metric into its drivers on the spot?
Do you distinguish correlation from causation in your readouts?
Does every analysis you present end with a clear recommended action?
Have you researched the company, the team, and the specific role enough to tailor your answers and ask sharp questions of your own?
Have you prepared two or three genuine questions to ask the interviewer that show you understand the role?

If you can answer yes to each, you're ready. Get a good night's sleep — being rested will do more for your performance than one more hour of practice.

The mindset that wins analyst loops

Strong analysts are skeptics with a bias toward action. They question the data before they trust it — checking the grain, the definitions, and whether a surprising number is a real signal or a broken pipeline. But they don't get paralyzed; they form a hypothesis, pull the cut that tests it, and come back with a recommendation. In the interview, narrate that loop out loud: here's what I'd verify, here's the question I'd ask the data, here's the action I'd recommend. That blend of rigor and decisiveness is exactly what hiring managers are listening for, and it's far rarer than raw SQL skill.

One last thing: practice explaining, not just querying

The single highest-leverage habit for an analyst interview is rehearsing how you explain a result, not just how you compute it. Take any query you write during prep and force yourself to finish with a single, plain-English sentence a busy executive could act on. Practice saying it without the words "correlation," "p-value," or "statistically significant" unless you immediately translate them. When you can move fluidly from a SQL result to a confident recommendation — and back up that recommendation with the segments and checks you ran — you stop sounding like someone who runs reports and start sounding like someone who drives decisions. That shift is what turns an interview into an offer.

Practice these questions with AI

Reading questions is step one. The candidates who convert are the ones who rehearse out loud and iterate on feedback. Paste your target job description into ClavePrep to generate role-specific questions, run a free AI mock interview (text or voice), and get structured feedback on each answer. Build your behavioral stories first with the free STAR Answer Builder.

SQL (the core)