Modeling & Scenario Questions
Power BI · Interview & Certification
Theory
Modeling and scenario questions check whether you can design and troubleshoot real solutions. Typical modeling questions: what is a star schema and why use it; fact versus dimension; what cardinality and cross-filter direction mean; why to avoid bidirectional relationships; how to handle two date columns (role-playing with USERELATIONSHIP); why a dedicated date table is needed; Import versus DirectQuery versus composite.
Scenario questions describe a problem and ask what you would do. 'The report is slow': use Performance Analyzer, check for heavy measures, too many visuals, bidirectional or many-to-many relationships. 'Totals do not match the source': check relationships, duplicates in dimension keys, filters and blanks. 'Each manager must see only their region': dynamic RLS with USERPRINCIPALNAME. 'Refresh takes hours': query folding, removing unused columns and incremental refresh. Answer as: diagnose, fix, verify.
Why we use it
Real jobs are mostly troubleshooting and design decisions. Good scenario answers show you can think like a working analyst, not just recite product features.
When to use it
Prepare these alongside the DAX questions, especially for mid-level roles. Use stories from your own projects where possible; a real example is more convincing than a textbook answer.
How it works
- Scenario question asked
- Clarify the symptom and who is affected
- Pick the diagnostic tool
- Is the root cause in the model? Yes: Fix relationships, keys or measures. No: Fix refresh, security or source.
- Verify the numbers or speed again
- Structured, credible answer
Syntax
Answer structure (scenario):
1. Clarify the symptom and scope
2. Diagnose with a tool (Performance Analyzer, View as, Refresh history, View Native Query)
3. Fix the root cause
4. Verify and prevent (tests, documentation)Example
Q: A manager in Pune sees all regions' data. Why?
A: Check that RLS roles exist and she is assigned in the Service; if she is a workspace Member or Contributor, RLS does not apply to her, so give access via the app as a reader instead.
Q: Sales by City shows the same total in every row.
A: The relationship Customers → Sales is missing or inactive, or the filter direction is wrong. Fix the relationship; Jaipur should show ₹2,700.
Q: Refresh of 5 years of Sales takes 2 hours.
A: Confirm folding, remove unused columns, then add incremental refresh (archive 5 years, refresh 10 days).Key points
- Star schema and single-direction filters are the safe defaults
- Same total on every row usually means a missing relationship
- RLS issues: role membership and workspace roles
- Slow refresh: folding, fewer columns, incremental refresh
- Answer scenarios as diagnose → fix → verify