Data Analyst Interview Questions
20 questions
Question 1: Share an example of an unexpected pattern you found while analyzing data to answer a business question.
What interviewers listen forListen for the comparisons they used to investigate the pattern and what the evidence supported. Calling two measures causally linked just because they moved together is a red flag.
Candidate tipSet out the business question, then focus on the pattern you found and the comparisons you used to test possible explanations. End with the conclusion and its limits.
Question 2: What would you check in a SQL query that returns a higher sales total after you add a join?
What interviewers listen forExpect checks of join keys, matches per row, and totals before and after the join. Adding DISTINCT without finding the cause can hide valid sales and is a red flag.
Candidate tipWork from what each row represents to the number of matches on the join keys. Explain how you would find repeated sales rows and verify totals after fixing the query.
Question 3: Describe a time you found a data quality issue that could have changed the conclusions of your analysis.
What interviewers listen forLook for evidence of the data issue, its effect on the analysis, and a checked correction or stated limitation. Quietly deleting inconvenient records is a red flag.
Candidate tipPick an issue such as missing records or inconsistent values. Explain the check that exposed it, how you assessed its effect, and what changed in the final analysis.
Question 4: Describe a time you changed a dashboard because it was not helping people track business performance.
What interviewers listen forA strong example links a dashboard change to a user's tracking need and shows whether it helped. Treating a new chart as proof of improvement without checking with users is a red flag.
Candidate tipChoose one dashboard that missed its purpose. Explain what users needed to track, the changes you made, and how you checked whether the revised dashboard helped.
Question 5: Give an example of how you turned a vague business request into a clearly defined analysis.
What interviewers listen forThe example should show how they clarified the decision, scope, and expected output with the requester. Starting analysis without finding out what the request meant is a red flag.
Candidate tipBriefly state the original request. Focus on the questions you asked the requester, the scope you agreed on, and how that agreement guided the analysis.
Question 6: Tell me how you explained an analysis finding to people who were unfamiliar with the data.
What interviewers listen forListen for how they made the finding understandable without losing its limits, and how they checked understanding. Repeating technical terms without explanation is a red flag.
Candidate tipUse one presentation or written explanation. Describe what the audience needed to understand, the words or chart you chose, and how you knew the message was clear.
Question 7: What would you do if two teams wanted different definitions of an active customer in a shared dashboard?
What interviewers listen forStart with the decisions each team needs the metric for, then clarify calculation rules and labels. Showing incompatible counts under the same unexplained label is a red flag.
Candidate tipExplain how you would compare the teams' needs, agree on calculation rules, and label any justified differences so dashboard users can interpret the counts.
Question 8: How would you prepare an analysis that combines a weekly outside data feed with daily records from an internal system?
What interviewers listen forExpect checks of date coverage, matching identifiers, and a common reporting period. Inventing daily values from weekly totals without a stated basis is a red flag.
Candidate tipDescribe the sources' date ranges and level of detail. Explain how you would match records and align periods while preserving what the weekly feed can actually support.
Question 9: Walk me through a time you worked with a data engineer or IT colleague to resolve a recurring data feed problem.
What interviewers listen forLook for a concrete data feed symptom, evidence they gave the engineer, and checks after the fix. Assuming the feed is reliable just because it runs again is a red flag.
Candidate tipChoose one recurring feed problem. Separate your investigation from the engineer's work, explain the evidence you shared, and end with how you checked that the problem was resolved.
Question 10: Tell me about a time you delivered a useful answer to an urgent data request with little time for analysis.
What interviewers listen forA strong example shows a focused answer, checks on the numbers, and clear limits within the deadline. Presenting unchecked figures as final because the request was urgent is a red flag.
Candidate tipChoose one urgent request. Explain the deadline, the data and checks you used, and what you delivered. Include any limits you told the requester and whether the answer met their need.
Question 11: How do you organize transformation code when several reports need the same calculation from raw transaction data?
What interviewers listen forExpect one shared calculation with clear inputs and checks that reports use it consistently. Keeping separate, conflicting copies of the same business rule is a red flag.
Candidate tipExplain where you would clean the transaction data and define the shared calculation. Describe how reports would reuse that logic and how you would check their results.
Question 12: Describe a time your documentation helped another analyst reuse a report without having to ask you how it worked.
What interviewers listen forListen for the sources, definitions, and reporting logic they documented, plus evidence another analyst could use them. Leaving essential filters undocumented undermines reuse.
Candidate tipPick a report you handed over. Describe the notes you wrote about its data and calculations, then give evidence that the other analyst could use them. Include anything you later clarified.
Question 13: What would you do if an AI-written query produced plausible results but you could not explain part of its logic?
What interviewers listen forThey should resolve the unclear logic and test the query against known cases before relying on it. Accepting plausible totals as proof that AI-generated code is correct is a red flag.
Candidate tipStart with how you would understand the unclear clause and compare it with the business request. Then describe checks using known records before trusting the report's results.
Question 14: Share an example of how you used operational data to pinpoint a bottleneck that was slowing work down.
What interviewers listen forSeek a specific delay, evidence that located the constraint, and the candidate's own analysis. Blaming the busiest team without examining workload or waiting time is a red flag.
Candidate tipSet out the workflow and delay. Focus on the records or timing comparisons you used to locate the bottleneck, and finish with what the analysis established and any limits.
Question 15: How would you set up an A/B test to measure whether a product change increases sign-ups?
What interviewers listen forExpect random assignment, a defined sign-up rate, and a sample size and run length chosen before testing. Stopping as soon as one version looks better risks a false win.
Candidate tipDefine who enters the test and what counts as a sign-up. Describe assignment to versions, the data you would collect, and how you would plan enough observations to compare results.
Question 16: Walk me through a time you automated a manual step in a recurring report.
What interviewers listen forThe example should identify the manual step, their automation, and checks against expected report output. Calling a successful script run proof of correct figures is a red flag.
Candidate tipExplain the old reporting step and the change you built. Spend most of the answer on your work and validation, then describe the effect on manual effort and report reliability.
Question 17: Give an example of how you helped a stakeholder learn to answer their own questions using a self-service data tool.
What interviewers listen forListen for hands-on help with the tool and evidence the stakeholder could choose filters and interpret results independently. A demonstration alone does not establish that they learned to use it.
Candidate tipUse one stakeholder's data question. Describe how you helped them use the tool, including filters or metric definitions, and how you checked that they could repeat the work themselves.
Question 18: What would you do if a manager requested a customer-level report but only had permission to view summary data?
What interviewers listen forFirst protect the customer records, then consider a permitted summary or a request to the data owner for access. A manager's title does not override the stated access restriction.
Candidate tipExplain why you would hold the customer-level report. Describe how you would offer a summary within their permissions or help them seek authorization from the data owner.
Question 19: Talk about a time you used customer usage data to help a product team prepare for a launch.
What interviewers listen forA strong example connects the candidate's usage analysis to a launch decision and states the data's limits. Treating a small group of active users as proof of demand from all customers is a red flag.
Candidate tipName the launch decision the team faced. Explain which customer activity you analyzed, the insight you shared, and how it informed preparations. Include gaps in the users or activity covered.
Question 20: Describe a time you worked with a system owner or data vendor to start collecting a field your analysis needed.
What interviewers listen forLook for a clear field definition, the candidate's work with its owner, and checks on newly collected values. Assuming a field with the right name contains the needed information is a red flag.
Candidate tipExplain the missing field's purpose and how you agreed on its meaning and collection with the owner or vendor. End with how you tested the new values and whether they filled the analysis gap.