Use these questions to interview Data Analyst candidates at {{Company name}}. They are grouped by competency so different interviewers can own different groups without overlap. For the SQL group, pair the questions with a live exercise on a realistic dataset — talking about SQL is not the same as writing it. Ask for specific projects and numbers throughout.

SQL and data manipulation

• Walk me through the most complex query you wrote recently. What made it complex, and could it have been simpler?
• Live exercise: join three tables to compute monthly active users and month-over-month change.
• Explain window functions to me with a real problem you solved using one.
• How do you approach a query that runs for twenty minutes? Walk me through your optimization steps.
• Live exercise: this query returns duplicated revenue. Find the fan-out and fix it.
• When do you move work out of SQL into a spreadsheet, a notebook, or a dashboard tool?

What good looks like: strong candidates narrate as they write — stating assumptions about grain and keys before joining, sanity-checking row counts after each step, and catching the planted fan-out by reasoning about cardinality rather than staring. Weak candidates produce plausible-looking SQL and declare it done without verifying anything.

Analytical reasoning and statistics

• A metric moved eight percent overnight. Walk me through your first hour of investigation.
• How do you decide whether a difference between two groups is real or noise?
• Explain a time you found the true cause of a metric change was a data or tracking issue, not a business change.
• What is survivorship bias, and where have you actually encountered it in your work?
• Describe an A/B test you analyzed. What would have invalidated the result?
• When has an average misled you, and what did you use instead?

What good looks like: the best answers show a checklist instinct — segment the change, check the pipeline, compare cohorts, then hypothesize — plus honest epistemics: candidates who say what they could not conclude are stronger than those with an answer for everything. Distrust candidates who have never found a tracking bug; it means they never looked.

Business framing and metrics

• Tell me about an analysis that changed a real decision. What was decided differently?
• A stakeholder asks for a dashboard. What questions do you ask before building anything?
• Describe a metric you defined from scratch, including the edge cases you had to settle.
• Tell me about a request you pushed back on because the question was not answerable with data.
• How do you decide an analysis is good enough to ship versus needing another week?
• What is a vanity metric you have seen teams chase, and what did you propose instead?

Communication and visualization

• Walk me through presenting a complex finding to a non-technical executive. How did you structure it?
• What makes a chart bad? Show me the worst chart you have rescued.
• How do you communicate uncertainty without losing the audience?
• Tell me about a time your analysis was misinterpreted. What did you change in how you present?
• How do you write the summary at the top of an analysis? What goes in the first three lines?

Data quality and tooling

• Describe a data-quality issue you caught before it reached a report. What was the signal?
• How do you document your work so another analyst can reproduce it in six months?
• What does your ideal analytics stack look like — warehouse, transformation, BI — and why?
• Two dashboards show different numbers for the same metric. Walk me through resolving it.
• What checks do you run before publishing a number executives will repeat in a board meeting?