Data Analyst interview questions and sample answers

A data analyst interview usually mixes two kinds of question: ones that test whether you know the job, and behavioral ones that ask for proof from your past. Below are 10 of the first with a short model answer, and 5 of the second with what the interviewer is really checking. Rewrite every answer in your own words and with your own numbers.

Job-specific questions

1. How do you validate a dataset before starting analysis?

I check row counts against the source system, look for nulls and duplicates with pandas' isnull and duplicated methods, and spot-check a sample of records against the raw source. I also confirm date ranges and units match what the business expects.

2. Walk me through how you would build a dashboard for weekly sales tracking.

I would start by asking stakeholders what decision the dashboard drives, then pick 3 to 5 key metrics like revenue, conversion rate, and average order value. I'd build it in Tableau or Looker with a weekly refresh and region filters.

3. What is your process when a stakeholder asks for a number that contradicts another report?

I trace both numbers back to their source queries and check the date ranges, filters, and definitions used, like whether active user means 30-day or 90-day. Usually the discrepancy is a definition mismatch, which I document going forward.

4. How do you decide which statistical test to use for an A/B test result?

For comparing conversion rates between two groups I use a two-proportion z-test or chi-square test. For continuous metrics like revenue per user I check for normality first, then use a t-test or Mann-Whitney U if the data is skewed.

5. Explain a time you found an error in your own analysis before presenting it.

I was calculating churn rate and caught that I'd excluded customers who canceled mid-month due to a date filter bug. I re-ran the query, the rate jumped from 4 percent to 6.5 percent, and flagged the correction before the meeting.

6. How do you handle a dataset with significant missing values?

I first check if the missingness is random or systematic, for example if a survey field is only missing for one region. Depending on the pattern I either impute with median or exclude the field if over 40 percent is missing.

7. What SQL techniques do you use to find duplicate records?

I use GROUP BY on the key columns with a HAVING COUNT(*) > 1 clause to spot duplicates, or a window function like ROW_NUMBER() OVER PARTITION BY to identify and remove extras while keeping the most recent record.

8. How would you explain a complex regression result to a non-technical executive?

I'd skip the coefficients and R-squared and translate it into a business sentence, like for every dollar we spend on ads we see about 3 dollars back in the first month. I'd back it with one simple chart, not the full model output.

9. Describe your approach to detecting outliers in a dataset.

I use the IQR method, flagging points beyond 1.5 times the interquartile range, or a z-score above 3 for normally distributed data. Then I investigate manually since some outliers are data errors and others are legitimate, like a bulk order.

10. What tools do you use for version-controlling your analysis?

I keep SQL queries and Python notebooks in Git, commit with descriptive messages tied to ticket numbers, and use nbstripout to keep notebook diffs clean. For recurring reports I parameterize the script so anyone can rerun it against a new date range.

Behavioral questions

Answer these with STAR: the Situation in one sentence, the Task, the Action you took (most of the answer), the Result with a number.

11. Tell me about a result you are proud of.

What they are checking: A specific outcome with a number, and what you personally did to get it.

A result to build the answer around: Found a discount-stacking leak in checkout data worth $240k/year; fix shipped within 2 weeks.

12. Describe a time you improved how something was done.

What they are checking: That you notice waste and fix it without being told, then measure the difference.

A result to build the answer around: Built 14 Tableau dashboards replacing 30 manual weekly spreadsheets, saving the team ~10 hours a week.

13. Tell me about a time you had to deliver under pressure.

What they are checking: How you prioritise, communicate early and still finish to standard.

A result to build the answer around: Designed and analysed 20+ A/B tests; the winning onboarding flow lifted activation by 7%.

14. Give an example of working with a difficult colleague or customer.

What they are checking: Calm, the other person's view stated fairly, and a result that helped both sides.

A result to build the answer around: Modelled churn drivers with logistic regression; the retention campaign it informed cut churn by 1.5 points.

15. What is something you learned from a mistake?

What they are checking: Ownership without excuses, and the habit you changed so it did not happen again.

A result to build the answer around: Cleaned and documented 60 dbt models so other teams could self-serve.

Before the interview

Interviewers read your resume just before they walk in, and most questions come from it. Every bullet on it should be one you can expand into a two-minute STAR story. Check that it matches the posting with the free ATS Match Score, and look at the data analyst resume keywords the ad is likely to use.

Turn your notes into a data analyst resume, free

No signup, no card. See the full data analyst resume example and the cover letter example.

Interview questions for related jobs

All interview question lists · CV Forge — free resume generator