Business Intelligence Analyst interview questions
A BI interview is really checking two things: can you write the query, and can you work out what the person asking for it actually needs. You'll get tested on both, often in the same hour. Here's what each round is looking for, the questions that come up, and how to answer them without sounding rehearsed.
What happens in each round
- 1
Recruiter screen
What happens
Whether your tools match the stack. Expect to be asked which of Power BI, Tableau or Looker you've used day to day, how comfortable you are in SQL, and who your reports went to.
- 2
Hiring manager conversation
What happens
How you handle stakeholders and messy requests. They want a story about a dashboard that changed a decision, and a sense of whether you push back or just build what you're told.
- 3
SQL assessment
What happens
Joins, aggregation, window functions and whether you notice duplicates. It's either a live screen share or a take-home against a small sample schema.
- 4
Dashboard or case presentation
What happens
Whether you can take a dataset and a vague business question, pick the right measures, and explain the result to people who won't read the SQL.
- 5
Team or stakeholder panel
What happens
Fit with the people who'll send you requests. Finance or sales leads often sit in, and they're judging whether they'd trust your numbers in their own meetings.
Questions you're likely to get
1.Walk me through a dashboard you built that people actually used.
Why they ask
Anyone can build a pretty report. They want proof you started from a real question and that someone changed what they did because of your work.
How to answer
- Name the audience and the decision they were stuck on
- Explain how you found out what they needed before you built anything
- Describe the data sources and the model underneath, briefly
- Say how you know it got used: usage logs, a meeting it replaced, a process that changed
2.Write a query that returns each customer's most recent order.
Why they ask
It's the classic window function test. It also shows whether you think about ties and customers with no orders.
How to answer
- Use ROW_NUMBER partitioned by customer and ordered by order date descending, then filter to the top row
- Mention the alternative of joining to a MAX date subquery and why ties can duplicate rows there
- Ask whether customers with no orders should appear, and switch to a LEFT JOIN if so
- Talk through your reasoning out loud while you type
3.How would you design a data model for sales reporting?
Why they ask
Modeling is what separates BI from report pulling. A bad model gives wrong totals no matter how nice the dashboard looks.
How to answer
- Start with the grain of the fact table, such as one row per order line
- Lay out the dimensions: customer, product, date, region, sales rep
- Explain why a star schema keeps filters predictable in Power BI or Tableau
- Mention how you'd handle a rep changing territory over time
4.Two dashboards show different revenue for the same month. How do you figure out why?
Why they ask
This happens constantly in the job. They want a calm, methodical process, not a guess.
How to answer
- Check the definitions first: booked versus invoiced, gross versus net of refunds
- Compare filters, date logic and time zones between the two reports
- Trace both back to the source tables and reconcile at the lowest level you can
- Agree one definition with the owners and document it so it doesn't happen again
5.What's the difference between a measure and a calculated column in Power BI?
Why they ask
It's a quick check that you've built real models in DAX, not just dragged fields onto a canvas. Tableau shops ask the same thing about calculated fields and level of detail expressions.
How to answer
- A calculated column is computed row by row at refresh and stored in the model
- A measure is computed at query time and responds to filter context
- Use measures for anything aggregated, like totals and ratios, to keep the model small
- Give a real example where you switched one for the other and why
6.A dashboard takes forever to load. What do you do?
Why they ask
Slow reports stop getting opened. They want to see that you know where performance problems actually live.
How to answer
- Find out whether it's the query, the model or the visuals, using the tool's performance analyzer
- Push heavy transformations upstream into the warehouse or a dbt model
- Cut unused columns, reduce high-cardinality fields and pre-aggregate where it makes sense
- Trim the number of visuals on the page and check import versus live connection
7.How do you decide what goes on an executive dashboard?
Why they ask
Leadership dashboards fail when they try to show everything. They're testing judgment and restraint.
How to answer
- Start from the decisions the executive makes each week or month
- Pick a small set of KPIs tied to those decisions, each with a target or comparison
- Put detail behind drill-throughs instead of on the front page
- Check back after launch to see what they actually look at
8.Tell me about a time a number you reported was wrong.
Why they ask
Everyone in BI ships a wrong number eventually. They want to know you own it quickly and fix the cause, not only the symptom.
How to answer
- Say plainly what was wrong and who had already seen it
- Explain how you told them and how fast
- Describe the root cause, such as a join that duplicated rows
- Name the check you added so it couldn't happen the same way again
9.Marketing and finance disagree on what an active customer is. What do you do?
Why they ask
Metric definitions are the political heart of the job. They want someone who can broker an answer instead of quietly picking a side.
How to answer
- Get both definitions written down and show each team the number theirs produces
- Find out what decision each team uses the metric for
- Propose either one shared definition or two clearly labelled metrics
- Get a named owner to sign off and record it in the data dictionary
10.A stakeholder asks for a giant export with every column. How do you respond?
Why they ask
Raw export requests are a sign the real need hasn't been understood. They're checking whether you can dig for it without being difficult.
How to answer
- Ask what they plan to do with the file once they have it
- Look for the recurring question behind the request
- Offer a filtered view or a self-service report if it'll come up again
- Still deliver something quickly if the need is genuinely one-off
11.How do you check a report is right before you share it?
Why they ask
Trust is the whole product. One wrong number can make a team stop using your dashboards for good.
How to answer
- Reconcile totals against the source system or a finance-approved figure
- Spot-check a few individual records end to end
- Test edge cases like returns, cancelled orders and missing dates
- Have the stakeholder confirm a number they already know before launch
Mistakes that sink good candidates
Talking about tools for the whole interview and never about the business question the work answered
Writing SQL in silence during the live assessment, so they can't follow your thinking
Blaming data engineers or stakeholders for every bad number in your stories
Presenting a case study dashboard crammed with charts and no clear headline
Need more Business Intelligence Analyst interviews to prep for?
HeroApply applies to Business Intelligence Analyst jobs that match you, every day. 169 jobs are open today.