Interview guide · Technology roles
Data analyst interview questions
In short
Good data analyst interview questions test whether a candidate can pull the right data with SQL, clean it, find what actually matters, explain it to non-technical people and spot when a number is wrong. Below are 31 questions for junior and experienced data analysts, covering SQL, Excel, statistics, visualization and business sense, with strong answers, red flags and a scorecard.
A data analyst's job is to turn messy operational data into decisions someone can act on. In practice that means writing SQL against tables nobody documented, reconciling two reports that disagree, building a dashboard that the operations manager will actually open, and saying clearly 'this is what changed, this is probably why, and here is what I would check next.' Technical skill is the entry ticket. What separates good analysts is skepticism about their own numbers and the ability to explain a finding in two sentences to a busy director.
Use the groups below to build a 45 to 60 minute interview plus a practical exercise: openers, SQL and data handling, statistics and analysis, visualization and communication, business and stakeholder questions, and questions for senior analysts. For junior data analyst roles, put more weight on SQL fundamentals, Excel, curiosity and clear writing; for senior roles, add experiment design, metric definition and influence. Always include a hands-on task. A 30 to 45 minute SQL exercise on a small realistic dataset, or a short take-home where the candidate explains a finding in a few slides, predicts performance far better than asking someone to define a left join.
What to look for in a data analyst
- SQL fluencyMost analyst time goes into querying data. Candidates should be comfortable with joins, aggregation, filtering, window functions and handling nulls without looking everything up.
- Skepticism about dataDuplicates, missing values, time zone shifts and changed definitions produce wrong answers that look right. Good analysts check row counts and sanity-test results before sharing them.
- Business framingThe best analysis starts from the decision someone needs to make, not from the data that happens to be available. Look for candidates who ask what the stakeholder will do with the answer.
- Clear communicationAn analysis that nobody understands changes nothing. Analysts need to lead with the conclusion, use simple charts and state uncertainty honestly.
- Statistical judgmentAnalysts do not need advanced math, but they must know the difference between correlation and causation, when a change is just noise and why averages can mislead.
- Organization and reproducibilityQueries saved and documented, logic that someone else can rerun, and version control where the team uses it save hours every month and prevent conflicting numbers.
Opening and background questions
Tell me about an analysis you did that changed a decision. What was the question, and what happened?
- Why ask it
- Shows impact, not just activity.
- A strong answer
- Describes the business question, the data used, the finding, how they presented it and what changed as a result, with a number if possible.
- Red flags
- Only describes building reports, with no example of anyone acting on them.
What tools do you use day to day, and which are you strongest in?
- Why ask it
- Checks fit with your stack: SQL dialect, spreadsheet, BI tool, Python or R.
- A strong answer
- Gives an honest picture of depth, such as 'SQL daily, Excel daily, Tableau for dashboards, some Python for cleaning'.
- Red flags
- Lists every tool on the market with no sense of depth.
What kinds of data have you worked with, and how big were the tables?
- Why ask it
- Experience with transactional, event, survey or financial data affects ramp-up time.
- A strong answer
- Names data types and approximate scale and mentions specific challenges such as event logs with duplicates or customer records with poor matching.
- Red flags
- Has only worked with small, clean sample files.
How do you decide whether a request needs a quick query or a full analysis?
- Why ask it
- Analysts drown when every question becomes a project.
- A strong answer
- Asks about the decision and its stakes, gives a quick answer when that is enough and proposes a deeper look only when the question warrants it.
- Red flags
- Treats every request the same way.
Why data analysis, and what part of the work do you enjoy most?
- Why ask it
- Motivation and fit with the actual mix of the job.
- A strong answer
- Specific enjoyment, such as finding the cause of an unexpected change or making a messy process measurable.
- Red flags
- Only wants to do machine learning and sees reporting as beneath them.
SQL and data handling questions
Explain the difference between an inner join and a left join, and give an example where choosing wrong changes the answer.
- Why ask it
- Join mistakes are one of the most common sources of wrong numbers.
- A strong answer
- Explains that an inner join keeps only matching rows and a left join keeps all rows from the left table, then gives an example such as customers with no orders disappearing from a count.
- Red flags
- Knows the definitions but cannot give a real consequence.
Write a query that finds the top three products by revenue in each region.
- Why ask it
- Tests window functions, a core intermediate SQL skill.
- A strong answer
- Aggregates revenue by region and product, applies ROW_NUMBER or RANK partitioned by region and ordered by revenue, filters to rank three or less, and mentions how ties are handled.
- Red flags
- Tries to solve it with LIMIT 3 on the whole result.
Your query joining orders to customers returns more rows than the orders table has. What is going on?
- Why ask it
- Fan-out from duplicate keys produces inflated totals in many real reports.
- A strong answer
- Recognizes duplicate keys on the joined table, checks uniqueness with a GROUP BY and HAVING count above one, and fixes the join or deduplicates first.
- Red flags
- Adds DISTINCT without understanding why the rows multiplied.
How do NULL values affect counts, averages and filters in SQL?
- Why ask it
- Null handling causes silent errors.
- A strong answer
- Knows COUNT(column) skips nulls while COUNT(*) does not, AVG ignores nulls, comparisons with NULL need IS NULL, and uses COALESCE deliberately.
- Red flags
- Assumes nulls behave like zero.
You receive a spreadsheet of 50,000 customer records from another team. What do you check before using it?
- Why ask it
- Tests data cleaning habits.
- A strong answer
- Checks row counts against the source, duplicates, missing values, data types, date formats, outliers, inconsistent categories and how the file was produced.
- Red flags
- Starts building charts immediately.
How would you calculate a 7-day rolling average of daily signups?
- Why ask it
- Rolling metrics are common in operational reporting.
- A strong answer
- Uses a window function with a frame of the current row and six preceding rows, ordered by date, and handles missing days by building a date spine first.
- Red flags
- Calculates weekly totals instead and calls them rolling.
In Excel or Google Sheets, how would you combine data from two tables and summarize it?
- Why ask it
- Many stakeholders live in spreadsheets, and junior analysts use them heavily.
- A strong answer
- Uses XLOOKUP or INDEX-MATCH, a pivot table, and checks for unmatched rows; mentions when it would be better to do the work in SQL.
- Red flags
- Copies and pastes values by hand.
Statistics and analysis questions
Weekly sales dropped 8 percent. Walk me through how you would investigate.
- Why ask it
- The most realistic analyst question there is.
- A strong answer
- Confirms the data is correct first, checks whether the drop is unusual compared with normal variation and seasonality, segments by region, channel, product and customer type, looks for known events such as outages or price changes, and reports what is explained and what is not.
- Red flags
- Jumps to a single cause without checking the data or segmenting.
When would you use the median instead of the mean?
- Why ask it
- Skewed data such as revenue or handle time makes the mean misleading.
- A strong answer
- Uses the median when data is skewed or has outliers, such as order values or time to resolve tickets, and often reports both.
- Red flags
- Always uses the average.
Customers who use our mobile app spend 30 percent more. Does the app cause higher spending?
- Why ask it
- Tests understanding of correlation and causation.
- A strong answer
- Explains that high-spending customers may be more likely to download the app, suggests comparing similar customers before and after adoption or running an experiment, and avoids claiming causation.
- Red flags
- Says yes and recommends pushing everyone to the app.
How would you analyze the results of an A/B test on a checkout page?
- Why ask it
- Experiment analysis is common for analysts in product and marketing teams.
- A strong answer
- Checks randomization and sample sizes, compares the primary metric with a confidence interval or significance test, watches guardrail metrics, avoids stopping early on a good day and checks segments with caution.
- Red flags
- Declares a winner after one day because the conversion rate is higher.
What is a p-value, in plain English?
- Why ask it
- Checks whether the candidate can explain a common statistical idea without jargon.
- A strong answer
- Roughly: how surprising the observed result would be if there were really no difference. A small p-value suggests the difference is unlikely to be pure chance, but it says nothing about how big or important the difference is.
- Red flags
- Says it is the probability the result is true.
How do you spot outliers, and what do you do with them?
- Why ask it
- Outliers can be data errors or the most interesting rows in the table.
- A strong answer
- Uses plots, percentiles or interquartile range, investigates the cause, removes only clear errors, and reports results with and without extreme values when they matter.
- Red flags
- Deletes outliers by default.
Visualization and communication questions
How would you design a dashboard for a contact center operations manager?
- Why ask it
- Tests audience awareness and metric selection.
- A strong answer
- Asks what decisions the manager makes daily, then shows a small number of key metrics such as service level, handle time, abandonment and staffing against forecast, with trends, targets and filters by team or site.
- Red flags
- Puts every available metric on one screen.
Explain a recent analysis to me as if I were a VP with five minutes.
- Why ask it
- A direct test of executive communication.
- A strong answer
- Starts with the conclusion and recommendation, gives two or three supporting points, states uncertainty and offers detail on request.
- Red flags
- Starts with the data sources and methodology.
When would you use a bar chart versus a line chart, and what charts do you avoid?
- Why ask it
- Chart choice affects whether people read the data correctly.
- A strong answer
- Line charts for trends over time, bar charts for comparisons across categories; avoids 3D charts, dual axes that mislead and pie charts with many slices.
- Red flags
- Chooses charts for how they look.
A stakeholder says your number is wrong because their report shows something different. What do you do?
- Why ask it
- Conflicting numbers are a weekly event in most companies.
- A strong answer
- Stays curious, compares definitions, filters, date ranges and sources line by line, documents the difference and agrees on a single definition going forward.
- Red flags
- Becomes defensive or simply changes their number to match.
How do you document your work so someone else can rerun it?
- Why ask it
- Reproducibility saves time and avoids conflicting numbers.
- A strong answer
- Saves queries in a shared repository or folder, comments logic, notes data sources and definitions, and uses version control where available.
- Red flags
- Work lives only on their laptop.
Business and stakeholder questions
How would you define and measure customer churn for a subscription business?
- Why ask it
- Metric definition is a core analyst skill with real business consequences.
- A strong answer
- Asks about billing cycles and what counts as cancellation, defines churn rate for a period, discusses voluntary versus involuntary churn and cohort views.
- Red flags
- Gives a formula without asking how the business works.
Tell me about a time a stakeholder asked for the wrong analysis. How did you handle it?
- Why ask it
- Good analysts reshape requests to answer the real question.
- A strong answer
- Asked what decision the stakeholder needed to make, proposed a different approach, agreed on it and delivered something more useful.
- Red flags
- Delivers exactly what was asked even when they knew it would not help.
You have five requests from three teams and time for two. How do you prioritize?
- Why ask it
- Analysts are almost always oversubscribed.
- A strong answer
- Considers business impact, deadlines and effort, talks openly with requesters and their manager, and offers quick partial answers where possible.
- Red flags
- Works first in, first out, or on whatever is most interesting.
Tell me about a mistake you made in an analysis that someone else relied on.
- Why ask it
- Honesty about errors is a must in an analyst.
- A strong answer
- Describes the error, how it was found, how quickly they told people and corrected it, and the check they added afterward.
- Red flags
- Has never made one, or hid it.
What would you need to know about our business in your first month to be useful?
- Why ask it
- Shows curiosity and a business-first mindset.
- A strong answer
- Core metrics and how they are defined, the main data sources and their quirks, who the key stakeholders are and which decisions are coming up.
- Red flags
- Only asks which tools the team uses.
Questions for senior data analysts
How have you improved the way a team uses data, beyond your own analyses?
- Why ask it
- Senior analysts change how a team works.
- A strong answer
- Built shared definitions, a self-service dashboard, a data quality check or training that reduced ad hoc requests, with an example of impact.
- Red flags
- Only talks about their own outputs.
How do you handle sensitive data such as employee or customer personal information?
- Why ask it
- Analysts often access personal data covered by privacy rules.
- A strong answer
- Uses the minimum data needed, aggregates or masks personal identifiers, follows access controls and is aware that rules such as GDPR or state privacy laws can apply.
- Red flags
- Downloads full customer tables to personal devices.
How do you decide when a question needs a data scientist or engineer rather than an analyst?
- Why ask it
- Shows awareness of role boundaries and when to escalate.
- A strong answer
- Predictive modeling at scale or production pipelines need other skills; analysts handle descriptive and diagnostic questions and simple forecasts, and collaborate on the rest.
- Red flags
- Attempts everything alone or hands off anything difficult.
Sample answers
First I'd make sure the drop is real. I'd check whether the data loaded completely, whether a definition or filter changed and whether the week had fewer trading days. Then I'd compare it with the same week last year and with normal week-to-week variation, because 8 percent might be within the usual range. If it's real, I'd split it by region, channel, product category and new versus returning customers, and see where most of the drop sits. Usually it's concentrated. Then I'd look for causes in that segment, like a stockout, a price change, a site outage or a marketing campaign that ended. I'd send a short note to the stakeholder the same day saying what we know and what we're still checking, rather than waiting for a perfect answer.
I'd start by assuming either of us could be right. I'd ask for their report and compare the two side by side: what date range, what filters, which source table, how they define an active customer, whether refunds are included. In my last role, finance and marketing had revenue figures 6 percent apart, and it turned out marketing counted orders at checkout while finance counted them at shipment. Neither was wrong, they were measuring different things. I wrote a one-page note explaining both, we agreed which number to use in the weekly business review, and I labeled both dashboards clearly so the question stopped coming up.
Not necessarily. The customers who download the app are probably already our most engaged buyers, so they'd likely spend more anyway. To get closer to the real effect, I'd compare the same customers' spending before and after they installed the app, against similar customers who didn't install it in the same period. Even better would be an experiment, like randomly offering an app incentive to some customers and not others. I'd present the 30 percent as a correlation with a clear caveat and give a range for the likely true effect once we've done the comparison.
Interview scorecard
| Competency | Weight | Strong (4–5) | Weak (1–2) |
|---|---|---|---|
| SQL and data handling | 30% | Writes correct joins, aggregations and window functions, handles nulls and duplicates, and checks results. | Struggles with joins or aggregations, or misses duplicates and nulls. |
| Analytical thinking | 20% | Validates data first, segments systematically and separates correlation from causation. | Jumps to conclusions or treats every difference as meaningful. |
| Communication and visualization | 20% | Leads with the conclusion, chooses clear charts and explains uncertainty plainly. | Starts with methodology and produces cluttered dashboards. |
| Business sense and stakeholder management | 15% | Starts from the decision, defines metrics carefully and prioritizes requests openly. | Delivers exactly what was asked without understanding why. |
| Data quality and rigor | 10% | Documents work, makes it reproducible and owns mistakes quickly. | Work cannot be rerun and errors go unreported. |
| Learning and curiosity | 5% | Asks probing questions about the business and data during the interview. | Shows little interest in the business behind the data. |
Score each competency from 1 to 5 straight after the interview, before discussing the candidate with anyone else, then multiply by the weight.
How to structure a data analyst interview
A three-stage process works for most teams. Start with a 30-minute screen covering background, tools and a couple of simple SQL questions asked out loud. Then run a practical exercise: either a 45-minute live SQL session on a small realistic dataset, or a take-home of no more than three hours where the candidate answers a business question and presents findings in a few slides. Finish with a 60-minute panel with the hiring manager and a stakeholder from the business, covering the take-home walkthrough, statistics, communication and the stakeholder questions above.
Keep the exercise close to the job. If the analyst will support operations, give them call volume or order data with a few planted problems such as duplicates and a missing day, and see whether they catch them. Scoring is far more consistent when every candidate gets the same dataset and the same rubric. A structured skills assessment can handle the SQL screen for high-volume or graduate hiring.
Check the data analytics specialist job description when deciding which tools and skills are truly required. Many teams list Python as essential when the role is 90 percent SQL and spreadsheets.
Data analyst versus related analytics roles
Data analyst interviews overlap with several other roles, and it helps to be clear about the difference before you write the job post. A data analyst works mostly with existing data to describe and explain what happened. A data scientist builds predictive models and experiments at greater statistical depth; our data scientist interview questions cover modeling, machine learning and production concerns. A business analyst focuses more on processes and requirements than on querying data; see the business analyst interview questions for that profile. A financial analyst concentrates on budgets, forecasts and financial models, covered in our financial analyst interview questions.
In smaller companies one person may do parts of all four. If so, decide which skills are essential on day one and which can be learned, and weight the scorecard to match. Interviewing a data analyst on machine learning theory when the job is building weekly dashboards wastes everyone's time and screens out strong practical candidates.
Mistakes interviewers make with analyst candidates
Asking for definitions instead of application is the most common one. Almost everyone can define a left join. Far fewer notice that a join tripled their revenue total. Use questions and exercises with real traps in them.
Second, ignoring communication until the final round. A candidate with excellent SQL who cannot explain a finding to a manager will struggle in most analyst roles. Score communication at every stage, including the take-home write-up.
Third, oversized take-homes. Exercises that take a full weekend put off strong candidates with jobs and families and favor those with free time. Cap the time, tell candidates the cap, and score what can reasonably be done in it.
Fourth, unclear expectations about tools. If candidates can use AI assistants, documentation or a specific SQL dialect, say so up front. Tell them how the exercise will be scored, and provide reasonable accommodations under the ADA when asked.
Frequently asked questions
What are the most common data analyst interview questions?
Expect SQL questions on joins, aggregation, window functions and nulls, a data cleaning scenario, a question on investigating a drop in a metric, correlation versus causation, how you would build a dashboard, and behavioral questions about stakeholders, conflicting numbers and mistakes.
What SQL questions are asked in a data analyst interview?
Typical SQL data analyst interview questions include the difference between join types, finding the top N per group with window functions, calculating rolling averages, removing duplicates, handling NULL values and explaining why a join returns more rows than expected.
What should a junior data analyst know for an interview?
Solid basics in SQL (selects, joins, group by and simple window functions), spreadsheets including lookups and pivot tables, basic statistics such as mean versus median, clear chart choices, and an example of an analysis from coursework, an internship or a personal project.
Do data analytics interviews include a technical test?
Most do. It is usually a live SQL exercise or a short take-home where you analyze a dataset and present findings. Good employers keep take-homes to a few hours and score them against a clear rubric.
How is a data analyst interview different from a data scientist interview?
Data analyst interviews focus on SQL, data cleaning, descriptive analysis, dashboards and communication. Data scientist interviews add machine learning, deeper statistics, experiment design and often programming in Python or R.
How long should a data analyst interview process take?
Usually two to three weeks from first screen to offer, with about three to four hours of total candidate time across a screen, a technical exercise and a final panel.