Power BI Scenario-Based Interview Questions: 12 Real-World Problems With Model Answers

Gulab Chand Tejwani
PMP · PL-300 · Microsoft Certified Trainer
· 11 min read

Scenario-based Power BI interview questions describe a realistic problem, such as a total that doesn't match, a slow report or a manager who should see only their region, and ask how you'd solve it. Interviewers use them because they show whether you can diagnose a model rather than recite definitions. A strong answer follows a pattern: restate the problem, name the likely cause, describe the fix, and say how you'd check it worked.
I'm Gulab Chand Tejwani, Microsoft Certified Trainer. When I help companies interview Power BI candidates, these are the questions that separate people who've built real reports from people who've watched tutorials. Below are twelve scenarios I'd actually use, with the answer I'd hope to hear.
How should you answer a scenario-based Power BI question?
Don't jump to a fix. Interviewers want to hear how you think. A structure that works:
- Restate it in one sentence so you both agree on the problem.
- Name the likely causes, most likely first.
- Describe the fix, with the DAX or setting if it's short.
- Say how you'd verify it: a test visual, a check against the source, Performance Analyzer.

Model and relationship scenarios
1. The total in a table visual doesn't equal the sum of the rows
Likely cause: the measure isn't additive. A ratio, a distinct count or an IF-based measure is recalculated at the total level, not summed. For example, a margin % measure at the total row is total margin divided by total sales, not the sum of each row's percentage.
Fix: if the business really wants the sum of row values, iterate explicitly:
Sum of Row Values =
SUMX ( VALUES ( Product[Category] ), [Some Measure] )
Check: explain to the stakeholder which number is "right" for their question. Often the total is correct and the expectation is wrong.
2. Sales by region shows the same number on every row
Likely cause: no relationship (or an inactive one) between the Sales table and the Region dimension, or the filter can't flow in the needed direction. With no filter reaching Sales, every row shows the grand total.
Fix: create or activate the relationship (Region one-to-many Sales), and check the key columns have matching data types and values.
Check: a table visual with Region and a row count. If counts look right, the relationship works.
3. You need to compare sales against monthly targets, but targets are at month and category level
Likely cause of trouble: the fact tables have different grain. Sales is daily per product; Targets is monthly per category.
Fix: keep Targets as its own fact table and relate both facts to shared dimensions at the level they support: a Date table and a Category-level dimension. For the date, relate targets on a month key (or the first day of the month) and only show targets at month level or above.
Check: a matrix by month and category showing both measures, with no repeated target values.
4. Students and courses: a many-to-many relationship
Scenario: one student enrols in many courses, and each course has many students.
Fix: use a bridge table (Enrolment with StudentID and CourseID), with one-to-many relationships from Student and Course into it. Be careful with bidirectional filtering. Only enable it where you need it, or use CROSSFILTER inside a specific measure, because bidirectional filters across a model can create ambiguity and slow things down.
| Choice | When it's right | Risk |
|---|---|---|
| Single direction (default) | Star schema, dimension filters fact | None in a clean star |
| Both directions | Bridge tables, specific many-to-many cases | Ambiguous paths, slower queries |
| Inactive + USERELATIONSHIP | Role-playing dates (order, ship, due) | Forgetting to activate in a measure |
| Many-to-many cardinality | Dimension at a different grain than the fact | Totals that surprise users |
DAX scenarios
5. The manager wants "% of total" that still respects the slicer
Fix: use ALLSELECTED in the denominator so the percentage is of the visible total, not the all-time total:
% of Visible Total =
DIVIDE (
[Total Sales],
CALCULATE ( [Total Sales], ALLSELECTED ( Product[Category] ) )
)
Mention the alternative: REMOVEFILTERS(Product[Category]) would ignore the category selection, which is right if they want each category's share of all categories. It only removes filters on that column, so a Date slicer still applies. For a share of all-time sales you'd also remove the date filter, for example with REMOVEFILTERS('Date').
6. Year-to-date shows blank or wrong values
Likely causes: no proper date table, the date table isn't marked as a date table, it has gaps, or the relationship uses a datetime column with times in it.
Fix: a continuous date table covering the full range, marked as a date table, related on a date-only column. For an April to March year, pass the year-end date to TOTALYTD.
7. A measure is slow because it uses FILTER over a large table
Scenario: CALCULATE([Total Sales], FILTER(Sales, Sales[Region] = "West")).
Fix: filter the column, not the whole table: CALCULATE([Total Sales], KEEPFILTERS(Sales[Region] = "West")), or better, filter the dimension column Region[Region] the same way. Filtering a whole fact table forces Power BI to iterate every row. KEEPFILTERS keeps the result identical to the original: without it, the column filter would replace an existing Region selection (so a visual filtered to East would show West sales instead of blank). Then check the change in Performance Analyzer.
Performance and refresh scenarios
8. A report page takes 20 seconds to load
How I'd diagnose it:
- Open Performance Analyzer, refresh visuals and find the slowest ones.
- Copy the slow DAX query and look for heavy iterators or table-wide FILTERs.
- Check the model: high-cardinality columns (timestamps, GUIDs), unused columns, calculated columns that could be measures, bidirectional relationships.
- Reduce visuals on the page. Twenty visuals means twenty queries.
Fixes: remove unused columns, split datetime into date and time, replace calculated columns with measures, simplify DAX, and use aggregations for very large data.
9. The dataset has five years of data and refresh takes too long
Fix: set up incremental refresh. Define RangeStart and RangeEnd parameters, filter the table with them in Power Query, and configure a policy such as "store 5 years, refresh the last 10 days". Mention that the source should support query folding so the filter runs at the source.
10. Import, DirectQuery or Direct Lake?
Answer framework:
- Import: fastest queries and full DAX and Power Query features, but data is only as fresh as the last refresh, and the model has size limits.
- DirectQuery: near real-time and no data copy, but every visual sends queries to the source, so performance depends on the source and some features are limited.
- Direct Lake (Fabric): reads Delta tables in OneLake directly into memory with performance close to Import and without a full data refresh. It needs Fabric capacity and data prepared in the lake.
Then pick based on the scenario's freshness need, data size and licensing.

Security and service scenarios
11. Regional managers should see only their region
Fix: row-level security. Create a role with a DAX filter on the Region dimension, because a filter on Region flows down to Sales through the normal one-to-many relationship. For dynamic RLS, keep a Security table that maps each manager's email to a region, and use it inside the Region filter:
-- Role filter on the Region table
Region[Region]
IN CALCULATETABLE (
VALUES ( Security[Region] ),
Security[ManagerEmail] = USERPRINCIPALNAME ()
)
If you filter only the Security table instead, the filter won't reach Sales unless a relationship carries it there, and managers may still see every region.
Then assign users or groups to the role in the service and test with View as, ideally signed in as a real Viewer too. Mention that RLS applies to viewers, and that workspace admins, members and contributors can see all data. Share through an app or with the Viewer role.
12. The CEO wants to ask questions in plain English
Answer: Copilot in Power BI can do this, if the workspace is on a paid Fabric capacity (F2+) or Premium (P1+), or the CEO is assigned to a Fabric Copilot capacity, and the admin has enabled it. The real work is preparing the model: clear names, measures with descriptions, hidden technical columns, and Prep data for AI (AI data schema, verified answers for common questions, AI instructions for business rules). I covered the details in Copilot in Power BI explained.
Practice: two quick scenario checks
You set up RLS so each manager sees only their region. A manager who is a Member of the workspace says she can see all regions. Why?
- ARLS only works in Power BI Desktop
- BUSERPRINCIPALNAME returns the wrong value
- CRLS doesn't restrict workspace Admins, Members or Contributors
- DThe report needs to be republished
Show answer and explanation
Correct answer: C
RLS filters data for viewers. People with edit roles in the workspace see all data. Give her the Viewer role or share through an app.
A model has a DateTime column with second-level precision, used only for daily reporting. The model is large and slow. What change helps most?
- AChange the column's format to show only the date
- BSplit it into a Date column (and a Time column if needed) and relate the Date column to the date table
- CCreate a calculated column that extracts the date
- DTurn on bidirectional filtering
Show answer and explanation
Correct answer: B
Formatting doesn't change what's stored, and a calculated column adds more data. Splitting at source (Power Query) cuts the column's cardinality sharply, which shrinks the model and speeds up queries.
How do you prepare for Power BI scenario interviews?
Build two or three complete projects and break them on purpose: remove a relationship, add a datetime column, write a slow FILTER, then fix each one. You'll recognise the scenarios instantly because you've caused them yourself. The star schema guide and Power Query data cleaning steps cover the modelling and preparation side, and DAX functions in Power BI covers the measures.
If you want guided labs, the Power BI Mastery course works through modelling, DAX, performance and security with datasets and step-by-step solutions, plus a scenario quiz at the end of every module, for ₹1,999 personal access. Many Power BI interviews also include SQL, and the SQL interview practice store has a free company set (ShopX) to start with.
FAQ
What are scenario-based questions in a Power BI interview?
Questions that describe a realistic problem, like a wrong total, a slow report or a security requirement, and ask how you'd solve it. They test diagnosis and judgement rather than definitions.
How do I answer Power BI scenario questions?
Restate the problem, list the likely causes, describe the fix and explain how you'd verify it. Mention trade-offs where they exist.
What Power BI topics come up most in scenario interviews?
Relationships and star schema, filter context in DAX (CALCULATE, ALL, ALLSELECTED), time intelligence, performance tuning, incremental refresh, row-level security and storage modes.
Do Power BI interviews ask about Copilot and Fabric?
Increasingly, yes. Expect questions about Copilot's requirements and model preparation, and about Direct Lake versus Import and DirectQuery.
How many projects should I build before a Power BI interview?
Two or three complete projects you can explain end to end, from raw data to a published, secured report, are usually more convincing than many small dashboards.