Data Analyst interview questions
Analyst interviews test three things in sequence: can you get the data, do you know what the number means, and can you make someone act on it. The last one is where most candidates are weakest — technically correct answers that never reach a recommendation.
12 questions across 4 areas, with what separates a strong answer in each.
Practise these out loud and get your answers graded — calibrated to the actual posting you're applying for.
SQL and data manipulation
Usually a live exercise. Correctness on edge cases is the differentiator.
Write a query returning each user's first purchase and the days until their second.
Strong answers cover: Window functions over a self-join where appropriate, and explicitly handling users with only one purchase rather than silently dropping them.
This query is slow. What would you check?
Strong answers cover: Reading the plan rather than guessing, and knowing which of indexing, join order, or scanned volume is actually the constraint.
How would you find duplicate records when there's no reliable unique key?
Strong answers cover: Defining what 'duplicate' means for this data before writing anything, and proposing a check for how many you'd expect.
Metrics and definition
Whether you can turn a vague business question into something measurable.
How would you define an 'active user' for our product?
Strong answers cover: Tying the window and the action to how the product is actually used, and naming what the definition would wrongly include or exclude.
Revenue is up but the team is worried. What might the number be hiding?
Strong answers cover: Mix effects, one-off contracts, a shrinking base with rising prices — and which cut of the data you'd pull first.
A stakeholder asks for a dashboard. How do you respond?
Strong answers cover: Finding the decision behind the request; strong analysts push back on a metric with no decision attached to it.
Experiments and inference
Enough statistics to avoid confidently wrong conclusions.
An A/B test shows a 2% lift with p = 0.04. Do you ship?
Strong answers cover: Asking about the pre-registered hypothesis, sample size, test duration, and how many variants were tried before this one looked good.
How would you measure something you can't randomise?
Strong answers cover: Naming a quasi-experimental approach — difference-in-differences, a matched cohort, a geographic holdout — and its specific weakness.
Two teams report different numbers for the same metric. What now?
Strong answers cover: Reconciling definitions and time zones and filters before questioning anyone's competence, then proposing a single source of truth.
Communication
The round that decides seniority.
Explain a technical finding to someone who won't read the methodology.
Strong answers cover: Leading with the recommendation and the confidence level, keeping the method available but not in the way.
Tell me about an analysis that changed a decision.
Strong answers cover: What the decision would have been without you, and how you know — this is the most direct evidence of impact you can give.
You find a result the requester clearly doesn't want. How do you deliver it?
Strong answers cover: Presenting the limitations honestly without burying the finding, and separating what the data says from what you recommend.
Data Analyst interview FAQ
- How much statistics do data analyst interviews expect?
- Enough to reason about sample size, significance and confounding out loud. Deriving tests from scratch is rarely asked; misreading a p-value is frequently disqualifying.
- Is SQL still the main technical screen?
- For most analyst roles, yes. Python or R usually supplements it rather than replacing it, and window functions are the most common place candidates stumble.
Other roles
Ready to practise for real?
Calibr builds a session from your actual job posting, grades every answer, and tracks the areas you keep losing points on.