Should a data analyst interview include a live SQL test?

Yes, in some form, because SQL is daily work. Keep it realistic: a small schema, a business question, and permission to look up syntax. Watch how the candidate checks the answer, not whether they remember function names.

Should we use a take-home case study for data analysts?

A take-home can show depth, but it costs the candidate unpaid time. If you use one, provide a clean dataset, cap it at two to three hours, and spend part of a later interview on the candidate's choices. A 45-minute live exercise is a fair alternative.

What if a candidate works in Python or R rather than SQL?

If the job needs SQL, test SQL. If it accepts either, let the candidate use the tool they know for the work sample and assess the reasoning and the checks the same way.

How much statistics should a data analyst know?

Enough for the decisions they will support. Most analyst roles need sampling, variability, comparisons between groups and the limits of correlation. Roles that run experiments need more. Test what the job uses, with a business scenario rather than a definition quiz.

Free interview kit

Data Analyst interview kit: questions and rubric

For hiring teams interviewing a mid-level data analyst who answers business questions with SQL, spreadsheets and a BI tool. It covers a short screen, five competencies with behavioral, situational and work-sample questions, anchored ratings, the questions never to ask, and a rule for deciding.

5 competencies · 11 questions · anchored 1-to-5 scale · Updated 10 October 2026 · Written by the Itya team

Share this kitLinkedIn X WhatsApp Email

How to use this kit

Agree the competencies with the hiring manager before the job is posted.
Give each interviewer one or two competencies, so nothing is asked twice.
Ask every candidate the same core questions; follow up freely.
Rate each competency against the anchors straight after the interview, alone.
Debrief on quotes, not adjectives, then decide with the rule at the end.

First-round screen

Fifteen to twenty minutes, by phone, video or a disclosed AI screen. The aim is to confirm the basics before the panel spends its time.

What draws you to this role, and what kinds of business questions do you most enjoy answering?
Tell me about an analysis you did that changed a decision. What was the question, and what did you find?
Which of the tools in the job description (SQL, spreadsheets, a BI tool, Python or R) have you used at work, and for what?
The job description sets out the location, working pattern and start date. Do those work for you, what is your notice period, and what compensation range are you expecting?

Competency 1

SQL and data manipulation

Most analysis lives or dies in the query. An analyst must get correct data out of messy tables without help.

Work sample: here are two tables, orders and customers. Write a query for monthly revenue by customer sign-up cohort over the last six months, and talk me through it. Looking up syntax is fine.
Follow up with
- What happens to customers with no orders?
- How would you check the result is right?
- How would the query change if orders could be refunded?
Strong answer
Writes a correct join and grouping, chooses the join type deliberately, handles nulls, checks the result against a known total, and asks about edge cases such as refunds and time zones.
Weak answer
Writes a query that double counts or drops rows without noticing, cannot explain the join choice, and does not sanity-check the output.
Tell me about the messiest dataset you have worked with. What did you do to make it usable?
Follow up with
- How did you record what you changed?
- What did you raise with the owner of the data?
Strong answer
Names specific problems (duplicates, inconsistent IDs, missing values), the checks and fixes applied, a record of each step, and raising the root cause with the data owner.
Weak answer
Cleaned it by hand in a spreadsheet with no record of the changes and no check of their impact on the result.

Rating	What it looks like
5 · Strong	Writes correct, readable queries, handles edge cases deliberately, validates results against known totals, and documents every cleaning step.
3 · Mixed	Writes correct queries for standard questions, but misses edge cases such as nulls, duplicates or time zones until prompted.
1 · Weak	Cannot write a correct join or aggregation unaided, and does not check results.

Competency 2

Analytical reasoning and statistics

Numbers mislead without judgment. An analyst must tell signal from noise and correlation from cause.

Sales rose 12% in the week after a marketing campaign launched, and the marketing lead wants to call it a win. What would you check before agreeing?
Follow up with
- What else happened that week?
- How would you estimate what would have happened without the campaign?
Strong answer
Checks seasonality, other launches, the prior trend and normal week-to-week noise, proposes a comparison group or holdout, and explains the uncertainty in plain language.
Weak answer
Agrees because the number went up, or dismisses it without proposing a way to test it.
What would you do if a regional head asked whether a two-point difference in conversion between two regions is real?
Follow up with
- What would you need to know about the sample sizes?
- How would you phrase the answer for them?
Strong answer
Asks about sample size and variability, uses a suitable test or interval, and gives the answer as a range with a practical meaning, not just 'significant' or 'not significant'.
Weak answer
Declares it real or not real without considering sample size, or answers in jargon the regional head cannot act on.
Tell me about a time your analysis turned out to be wrong. What happened?
Follow up with
- How was the mistake found?
- What do you check now that you did not check before?
Strong answer
Owns the error, explains the cause (a bad join, a skewed sample), how it was corrected and communicated, and a habit that changed as a result.
Weak answer
Cannot recall an error, or blames the data with no change to their own checks.

Rating	What it looks like
5 · Strong	Tests alternative explanations unprompted, chooses methods that fit the question, and states uncertainty in terms a decision-maker can use.
3 · Mixed	Spots the obvious alternative explanations and knows the basic statistics, but needs prompting to quantify uncertainty.
1 · Weak	Takes numbers at face value, confuses correlation with cause, and ignores sample size.

Competency 3

Framing the business question

A precise answer to the wrong question is wasted work. Strong analysts find out which decision the analysis will inform before they start.

Tell me about a request where the question you were asked was not the question that needed answering.
Follow up with
- How did you find out?
- What did you deliver in the end?
Strong answer
Asked what decision the analysis would inform, reframed the question with the stakeholder, and delivered something that changed the decision.
Weak answer
Delivered exactly what was asked even though it would not help, or reframed the question without checking with the stakeholder.
What would you do if a manager asked you for 'a dashboard of everything' about customer churn?
Follow up with
- What three questions would you ask first?
- How would you know, six weeks later, whether the dashboard is useful?
Strong answer
Asks which decisions the dashboard should support and who will use it, proposes a few core measures with written definitions, and plans to retire views nobody uses.
Weak answer
Builds dozens of charts, or refuses without offering a smaller alternative.

Rating	What it looks like
5 · Strong	Starts every request by finding the decision behind it, scopes the work to that decision, and agrees definitions before building.
3 · Mixed	Clarifies the question on larger requests, but sometimes delivers analysis that does not connect to a decision.
1 · Weak	Starts building before understanding the question. Treats every request literally.

Competency 4

Communicating insights

An analysis matters only if someone acts on it. Analysts must make the finding clear to people who will never read the query.

Work sample: here is a chart and a short table from a real, anonymized analysis. In two minutes, explain to me, as a non-technical sales director, what it shows and what I should do.
Follow up with
- What is the main caveat I should know?
- What would you show me next?
Strong answer
Leads with the finding and a recommended action, uses the one number that matters, states the main caveat simply, and avoids jargon.
Weak answer
Describes the chart axis by axis, buries the finding, or overstates how certain it is.
Tell me about a time your finding was unwelcome to the person who asked for it.
Follow up with
- How did you present it?
- What happened to the decision?
Strong answer
Presented the evidence clearly and fairly, separated fact from interpretation, and held firm on the data while staying open about what to do next.
Weak answer
Softened the finding until it said nothing, or presented it combatively.

Rating	What it looks like
5 · Strong	Leads with the finding and the action, chooses the one number that matters, states caveats plainly, and holds firm on unwelcome results.
3 · Mixed	Explains findings clearly to technical peers, but needs prompting to lead with the action for non-technical audiences.
1 · Weak	Explains the method instead of the finding. Stakeholders could not say what to do after hearing it.

Competency 5

Rigor and data quality

Stakeholders trust the numbers until one is wrong. Analysts who check their work and maintain definitions protect that trust.

How do you check an analysis before you send it? Take me through the last one you did.
Follow up with
- What would make you delay sending it?
- Who reviews your work?
Strong answer
Describes concrete checks: row counts, reconciliation to a trusted source, spot-checking individual records, and a peer review for high-stakes work.
Weak answer
Says 'I double-check it' with no specific checks, or relies on stakeholders to spot errors.
What would you do if you found that a measure on a widely used dashboard had been calculated wrongly for months?
Follow up with
- Who do you tell first?
- How do you handle the historical numbers?
Strong answer
Confirms the error, tells the owner and the main users promptly with the size of the impact, fixes and documents the definition, and adds a check so it cannot recur.
Weak answer
Fixes it quietly, or waits until someone else notices.

Rating	What it looks like
5 · Strong	Checks every analysis in a set routine, keeps definitions documented, and surfaces errors openly with their impact.
3 · Mixed	Checks important work against a source, but checks are inconsistent and definitions are not written down.
1 · Weak	Has no routine for checking work, and has let errors stand or fixed them silently.

Never ask

Don't ask a candidate's age, or their graduation year to work it out. Why: age says nothing about analytical skill, and age-based decisions are unlawful in many places.
Don't ask about marital status, children, pregnancy or plans to start a family. Why: none of it is job-related, and the question is still put mostly to women, which makes it discriminatory as well as irrelevant.
Don't ask about religion, caste, community or the origin of a surname, directly or by proxy ('What does your father do?'). Why: in India these questions signal caste and religion, which have no bearing on the job.
Don't ask 'Where are you from originally?' or 'What is your native place?'. Why: it invites judgments about region and community. If location matters, ask whether the candidate can work from the job's location.
Don't ask about health conditions, disabilities or past medical leave. Why: the only relevant question is whether the candidate can perform the essential functions of the job, with or without reasonable accommodation.
Don't ask for salary history where the law restricts it, as many US states and cities do. Why: it carries past pay gaps into the new offer. Ask for the expected compensation range instead.
Don't screen on college tier, board percentages or a statistics degree as a proxy for skill. Why: plenty of strong analysts learned on the job. The SQL and reasoning work samples show the skill directly.

Making the decision

The must-haves are SQL and data manipulation, analytical reasoning, and rigor and data quality. Any 1 on a must-have is a no, and so is any competency rated 1 by two interviewers. A hire needs an average of 3.5 or higher. The hiring manager decides after the debrief and records the work-sample results and quotes behind each rating.

Questions people ask

Should a data analyst interview include a live SQL test?: Yes, in some form, because SQL is daily work. Keep it realistic: a small schema, a business question, and permission to look up syntax. Watch how the candidate checks the answer, not whether they remember function names.
Should we use a take-home case study for data analysts?: A take-home can show depth, but it costs the candidate unpaid time. If you use one, provide a clean dataset, cap it at two to three hours, and spend part of a later interview on the candidate's choices. A 45-minute live exercise is a fair alternative.
What if a candidate works in Python or R rather than SQL?: If the job needs SQL, test SQL. If it accepts either, let the candidate use the tool they know for the work sample and assess the reasoning and the checks the same way.
How much statistics should a data analyst know?: Enough for the decisions they will support. Most analyst roles need sampling, variability, comparisons between groups and the limits of correlation. Roles that run experiments need more. Test what the job uses, with a business scenario rather than a definition quiz.

A kit is step one.

Itya holds every interviewer to the same rubric, records the interview, and drafts the scorecard with the quotes behind each rating. The AI drafts; your team decides. Free plan, unlimited teammates.

Run this kit in Itya All interview kits