Skip to main content
VirtualInterview.ai
Sign InSign Up

Mock interviews by role · Data, AI & Cloud

Data Analyst mock interview

Data analyst interviews ask you to define a metric before you calculate it, and to say what the number does not mean. A correct query that answers the wrong question still fails the round.

Practice saying the grain of the table, the filter, and the caveat in plain language a hiring manager can repeat.

Practice this role with an AI interviewer

What the loop usually looks like

  1. Screen

    25 min

    A analysis you delivered and who used it.

  2. SQL or case

    45 min

    A metric to define, or a query to write and explain.

  3. Communication

    30 min

    Explain a messy result to a non-analyst.

What interviewers score

  • Metric definition
  • Query correctness
  • Caveats
  • Stakeholder language

10 data analyst interview questions and model answers

  1. Technical

    1. How do you define monthly active users so two teams do not argue?

    What they are testing: Metric contracts.

    I write the event that counts, the identity, and the window, and I exclude internal accounts. I publish that sentence next to the chart. If marketing wants a different event, that is a different metric with a different name.

  2. Technical

    2. A conversion rate jumped. What do you check before celebrating?

    What they are testing: Sanity checks.

    I check volume, a tracking change, a bot filter, and a mix shift toward an easier segment. A rate can rise because the denominator shrank. I look at counts beside the rate.

  3. Domain

    3. How do you explain a statistically small change to an executive who wants a story?

    What they are testing: Restraint.

    I say the change is inside normal week-to-week movement and I would not act on it yet. I offer the next read date. I do not dress noise up as a win.

  4. Technical

    4. When do you aggregate in SQL instead of in a spreadsheet?

    What they are testing: Reproducibility.

    If the number will be reused or the data is large, it lives in a query someone else can run. A spreadsheet is fine for a one-off check I label as such. I do not paste a number into a deck with no query behind it.

  5. Situational

    5. A stakeholder asks for data you suspect will be used to punish a team. What do you do?

    What they are testing: Ethics.

    I ask what decision the data will inform. I provide the metric with context and the limits. I do not hand over a ranking of people without a definition that those people have seen.

  6. Behavioral

    6. Tell me about an analysis that changed a decision.

    What they are testing: Impact.

    Support thought refunds were random. The query showed they clustered on one SKU after a packaging change. The team reverted the packaging. I kept the query so the next spike was visible in a day, not a month.

  7. Technical

    7. How do you handle nulls in a customer id?

    What they are testing: Data quality.

    I count them and find out whether they are logged-out users or a broken tag. I do not silently drop them if they are a large share of traffic. The chart notes the share that could not be identified.

  8. Domain

    8. What is the difference between a dashboard and an answer?

    What they are testing: Product of the work.

    A dashboard shows the current numbers. An answer says what changed, why you believe it, and what you would do. I start with the answer and link the dashboard as the source.

  9. Behavioral

    9. Describe a time your first query was wrong.

    What they are testing: Humility and checks.

    I joined on a key that duplicated rows and doubled revenue. A colleague’s finance total did not match. I fixed the grain and sent a correction the same day. I now check totals against a known source before I share.

  10. Situational

    10. You are asked for a number in an hour and the definition is ambiguous. What do you deliver?

    What they are testing: Time-boxed honesty.

    I send one number with the definition I used and the two decisions that definition cannot support. I do not disappear for a day to build the perfect model.

Frequently asked questions

Will I have to write SQL live?

Often yes. Practice talking through the join grain while you write. Interviewers coach candidates who narrate. They cannot help candidates who go silent.

How much statistics is expected?

Enough to know when a change is too small to act on, and the difference between correlation and a cause you have not shown. Deep modeling belongs in a data scientist loop.

What should I bring as a portfolio piece?

One analysis with the decision it changed and the query or notebook you can walk through. Hide anything confidential.

How do I practice the explanation, not just the query?

After you write the query, say the result in two sentences to a mock interviewer and invite a follow-up on the caveat.

Keep practicing