Microsoft Power BI Data Analyst (PL-300) - Practice Questions¶

15 questions for PL-300 prep. Most modeling and DAX questions assume a star schema and turn on understanding filter context.

Cert page: exams/azure/pl-300/

Question 1¶

Scenario: A 4-billion-row fact table lives in Azure Synapse. Current-day data must be accurate to the minute; five years of history is queried far more often and performance is a priority.

A. Full DirectQuery on the whole table, so that every visual always shows the latest minute of data from the source B. Import with incremental refresh only, set to refresh the most recent partition every minute through the gateway C. A composite model: history in Import, today in DirectQuery, shared dimensions in Dual D. Full Import with hourly refresh through the gateway, plus a larger capacity so that the refresh finishes quickly

Answer

**Correct: C** **Why:** A composite model satisfies both requirements at once. Dual mode on the shared dimensions is what prevents an unnecessary DirectQuery round trip for the imported queries. Full DirectQuery meets currency and fails performance. Import with any refresh schedule cannot be minute-accurate.

Question 2¶

Scenario: A refresh over a 200-million-row SQL table takes four hours. An index column is added as the second Power Query step, with filters and group-by afterwards.

A. Schedule the refresh overnight so the four hours fall outside business hours B. Switch the table to DirectQuery so that nothing needs to be refreshed C. Increase gateway memory so the mashup engine can hold more of the 200 million rows in RAM while it processes them D. Move or remove the index step so filters and grouping fold to the source, and check with View Native Query

Answer

**Correct: D** **Why:** Adding an index column breaks query folding, so every subsequent step processes locally on the full dataset. Reordering so folding-breaking steps come last restores source-side filtering. More memory treats the symptom. DirectQuery changes the interaction model to solve a refresh problem. Scheduling overnight leaves the four hours in place.

Question 3¶

Scenario: Profit Margin is created as a calculated column with (Revenue - Cost) / Revenue, then placed in a card with average aggregation. Finance reports the value is wrong at every level except individual rows.

A. Change the visual aggregation to sum B. Use AVERAGEX over the fact table to average the row-level margin C. Replace it with a measure: DIVIDE( SUM(Revenue) - SUM(Cost), SUM(Revenue) ) D. Add a second calculated column for the weighted average

Answer

**Correct: C** **Why:** Averaging a row-level ratio gives the average of ratios, not the ratio of totals. Ratios must be measures so aggregation happens before division. `DIVIDE` also handles division by zero. Summing percentages is worse. `AVERAGEX` computes an average of ratios, which is a different, usually unwanted, number here.

Question 4¶

Scenario: 200 regional managers must each see only their own region. An employee table maps each manager's email to their region.

A. Object-level security on the region and fact tables, with a role per manager that hides the other regions' rows from that manager B. A slicer defaulted to the user's region and saved in the report C. One dynamic RLS role filtering [Email] = USERPRINCIPALNAME(), with the security table related through to the facts D. 200 static row-level security roles, one for each manager, each with a hard-coded filter on that manager's own region and maintained by hand

Answer

**Correct: C** **Why:** Dynamic RLS with `USERPRINCIPALNAME()` scales to any number of users with one role. Static roles are unmanageable at 200. A slicer is not security, because users can change it. Object-level security hides tables and columns, not rows.

Question 5¶

Scenario: A SAMEPERIODLASTYEAR measure returns blank for every row. The model has a date column on the fact table and no separate date table.

A. Rely on Auto Date/Time, which Microsoft recommends for models with several date columns B. Create a contiguous date table, mark it as a date table, relate it to the fact, and use its date column C. Wrap the measure in CALCULATE with additional filters on the fact date column, to force the prior-year range for every row in the visual D. Change the fact date column's data type from Date/Time to Text

Answer

**Correct: B** **Why:** Time intelligence functions require a marked date table with a complete, gap-free date range. Blank results almost always mean the date table is missing, unmarked, or has gaps. Auto Date/Time creates hidden per-column tables that bloat the model and do not provide the marked-date-table behavior these functions expect.

Question 6¶

Scenario: A finance report must reach 3,000 employees across four departments, each seeing only their department's pages, while five authors keep editing without disrupting consumers.

A. Create four separate workspaces, one per department B. Share the report directly with all 3,000 users C. Publish an app with four audiences, keeping authors in the workspace D. Give all consumers Viewer access to the workspace and use RLS roles so that each department only sees its pages

Answer

**Correct: C** **Why:** Apps with audiences are the governed distribution mechanism, and publishing separates what consumers see from work in progress. Direct sharing is unmanageable at that scale and exposes edits immediately. Workspace Viewer access shows unpublished changes and offers no audience segmentation. Four workspaces duplicate the model and its maintenance.

Question 7¶

Scenario: Which workspace role can create and edit content and schedule refresh, but cannot publish an app?

A. Contributor B. Member, who can edit content and schedule refresh but is restricted from publishing the workspace app C. Viewer, who can edit content and schedule refresh in the workspace but is not allowed to publish an app D. Admin, who can edit content and schedule refresh but needs a Member to publish the workspace app for them

Answer

**Correct: A** **Why:** Contributor is exactly this boundary and it is directly testable. Admin and Member can both publish apps. Viewer can only consume, and is the only role to which row-level security applies.

Question 8¶

Scenario: An author tests row-level security in the workspace and sees all rows despite being assigned to a role. Why?

A. The role's DAX expression is wrong, for example a filter on the wrong column, so it returns every row to the tester B. Workspace Contributors or higher bypass RLS; test with View As or a real Viewer C. RLS requires a Premium capacity, so roles assigned on Pro workspaces are ignored when the report is opened D. The relationship direction is wrong, so the security table's filter never reaches the fact table in the model

Answer

**Correct: B** **Why:** RLS applies to the Viewer role. Anyone with Contributor or above sees all rows in the workspace regardless of role membership, so validation must use View As in Desktop and then a genuine Viewer or app consumer in the service.

Question 9¶

Scenario: A model built from one 60-column flat spreadsheet is 800 MB, slow, and filters inconsistently. What is the correct fix?

A. Split into fact and dimension tables, add a date table, and drop unused high-cardinality columns B. Add calculated columns for the common measures so results are precomputed C. Switch to DirectQuery so that the 800 MB file is no longer loaded into the model and the filters are evaluated at the source D. Reduce the number of visuals per page and turn off interactions between them, so that fewer queries run on each click, hoping the page then feels faster

Answer

**Correct: A** **Why:** Star schema resolves the ambiguous filtering and the repetition driving size. The largest size win usually comes from removing high-cardinality columns such as transaction IDs and timestamps. Calculated columns increase size. DirectQuery moves the problem to the source. Fewer visuals does not fix a model problem.

Question 10¶

Scenario: Which relationship configuration should generally be avoided because it introduces ambiguity?

A. Single-direction one-to-many from dimension to fact, with the dimension on the one side and filters flowing down to the fact B. Bidirectional cross-filtering between two tables C. An inactive relationship activated with USERELATIONSHIP inside the measure that needs the alternative date D. A one-to-many relationship from a date table marked as a date table to the fact table's date column

Answer

**Correct: B** **Why:** Bidirectional filtering creates ambiguous filter paths and can produce wrong results elsewhere in the model. The usual alternative is single direction plus `CROSSFILTER` inside specific measures where both directions are genuinely needed.

Question 11¶

Scenario: Incremental refresh cannot be configured on a table. What is the most likely cause?

A. The report is published to a workspace that has no deployment pipeline assigned B. The parameters are not named exactly RangeStart and RangeEnd, or folding does not reach the source C. The dataset is not on a Premium capacity, because incremental refresh is only offered to Premium workspaces D. There is no date table in the model, so the service cannot work out which partitions to create for the refresh

Answer

**Correct: B** **Why:** Incremental refresh requires two date/time parameters named exactly `RangeStart` and `RangeEnd`, a filter using them, and query folding so the source can partition. A model date table is unrelated to the parameter requirement.

Question 12¶

Scenario: One report page renders in 25 seconds while others are instant. The model is Import mode and reasonably sized. What is the first step?

A. Switch the model to DirectQuery so the slow page queries the source B. Increase the capacity size to the next SKU C. Split the page into three pages with fewer visuals each D. Run Performance Analyzer to see whether the time is in the DAX query, visual display, or other

Answer

**Correct: D** **Why:** The exam expects the diagnostic step before the fix. Performance Analyzer tells you whether the cost is in the query, the visual, or elsewhere, which determines whether you rewrite a measure, change the model, or reduce visual cardinality. Splitting pages may help, but only once you know why.

Question 13¶

Scenario: Which storage mode allows a dimension table to serve both an imported fact table and a DirectQuery fact table without forcing a source round trip for the imported queries?

A. Dual, which behaves as Import or DirectQuery depending on the query B. Direct Lake C. Import, which holds the dimension in memory but forces a source query when it serves the DirectQuery fact table D. DirectQuery, which always queries the source and so serves the two fact tables with no imported copy of the dimension

Answer

**Correct: A** **Why:** Dual lets the engine decide per query, acting as Import when serving imported facts and as DirectQuery when serving DirectQuery facts. That is precisely the problem Dual exists to solve, and it is why composite models depend on it.

Question 14¶

Scenario: What does a deployment pipeline rule do?

A. Restricts who can deploy between stages, using workspace roles B. Schedules deployments so each stage is promoted automatically C. Validates the semantic model and blocks promotion when any DAX measure fails D. Swaps data source parameters and connection details so each stage points at the right data

Answer

**Correct: D** **Why:** Rules are the mechanism ensuring production content points at production data after promotion. Access control between stages is governed separately by workspace permissions, and rules do not schedule or validate.

Question 15¶

Scenario: A report must be accessible to keyboard and screen reader users. Which combination is required?

A. A mobile layout, which reorders visuals for phone screens so that the report is readable for every kind of user, whatever device they open it on B. Alt text on visuals, a logical tab order from the Selection pane, enough contrast, and not relying on color alone C. Alt text on visuals and a high-contrast theme D. Exporting the report to PDF each week for assistive technology users

Answer

**Correct: B** **Why:** Accessibility on this exam consistently covers all four: alt text, tab order, contrast, and not conveying meaning through color alone. A mobile layout addresses screen size, not accessibility. PDF export is a different output, not an accessibility control.

Scoring guide¶

13-15 correct (85%+): Ready. Spend remaining time predicting DAX results before running them.
10-12 correct (65-80%): Review filter context and the star schema material, which underpin most wrong answers.
Below 10: Build the models listed in the practice plan. PL-300 rewards hands-on work more than any other exam in the Microsoft data track.