PL-300 Practice Test

1. A Fabric lakehouse holds a 900-million-row Delta table that a Spark job reloads every 15 minutes. Analysts want query performance comparable to Import mode, and new data should appear without scheduling full data refreshes. Which storage mode fits this requirement?

• A. Direct Lake mode on the lakehouse Delta tables
• B. A live connection to a semantic model published from Power BI Desktop
• C. Import mode with incremental refresh configured to run on a 15-minute schedule
• D. DirectQuery mode against the lakehouse SQL analytics endpoint for every table

2. A Direct Lake on SQL semantic model is built on a Fabric warehouse. One table is based on a SQL view rather than a Delta table, and Performance Analyzer shows that visuals using that table run noticeably slower than the rest. What explains the slowdown?

• A. Views force the VertiPaq engine to disable column compression for every table in the semantic model
• B. Queries against the view fall back to DirectQuery because Direct Lake on SQL can't load a view directly
• C. The view's rows are being imported into memory on every query, which triggers a full model refresh each time
• D. The capacity's Direct Lake row-group guardrail was exceeded, so the whole model switched to Import mode

3. A DirectQuery report over Azure SQL Database works for most pages, but a table visual that lists individual transactions for a whole year returns an error instead of data. The same visual works when filtered to one month. What is the most likely cause?

• A. Table visuals in DirectQuery mode can show at most 10,000 rows per page unless drillthrough is configured
• B. The semantic model exceeded the 1 GB size limit that applies to DirectQuery models on shared capacity
• C. The visual's query returned more than 1,000,000 rows, which exceeds the DirectQuery intermediate result limit
• D. Azure SQL Database rejects any DirectQuery statement that scans more than 100,000 rows in a single query

4. Finance keeps its budget in an Excel workbook on SharePoint and wants a report that reflects edits within minutes without waiting for a scheduled refresh. An analyst proposes connecting to the workbook in DirectQuery mode. What should the analyst tell them?

• A. DirectQuery is available for Excel workbooks as long as every sheet is first converted to a formatted table
• B. The Excel connector supports Import only, so the data must be imported and refreshed on a schedule
• C. DirectQuery works for Excel workbooks only when the file is stored in OneDrive for Business
• D. The workbook can be used in DirectQuery mode once it is registered on a standard-mode data gateway

5. A composite model has a 2-billion-row Sales table in DirectQuery mode and a 50-row Product table that is imported. A visual that combines Product attributes with Sales measures is slow, and Model view shows a limited relationship between the two tables. Which change should the analyst make?

• A. Set the Product table's storage mode to Dual so it can join Sales at the source or be served from the cache
• B. Convert the Sales table to Import and enable incremental refresh with a 30-day refresh window
• C. Replace the relationship with a TREATAS filter inside every measure that uses the Product table
• D. Set the relationship's cross-filter direction to Both so imported Product filters are pushed to Sales

6. A 40-table semantic model was built entirely in Import mode. The source database now changes every few minutes and stakeholders want visuals to show current values, so the analyst tries to switch the existing tables to DirectQuery in Model view. What will happen?

• A. The switch succeeds, but every measure that uses time intelligence is converted into a calculated column
• B. The tables can't be switched: storage mode can move from DirectQuery to Import or Dual, but not from Import to DirectQuery
• C. Each table can be switched from the Properties pane in Model view, and the model keeps its calculated columns and measures unchanged
• D. The switch succeeds only after 'Assume referential integrity' is enabled on every relationship in the model

7. A query merges customer records from an internal SQL Server database with exchange rates from a public web API. Refresh fails with 'Formula.Firewall: Query is accessing data sources that have privacy levels which cannot be used together.' What is the correct fix?

• A. Set the SQL Server source to Private and the web API to Public, then keep the merge inside a single query
• B. Turn on 'Ignore the Privacy levels' (Fast Combine) in the published semantic model's settings in the service
• C. Assign compatible privacy levels, such as Organizational, to both sources so they may be combined
• D. Change the merge to an inner join so no data from the SQL Server source has to be sent to the web API

8. A Power Query parameter named Region of type Text controls which rows are loaded, and a second parameter of type Any holds a connection string. After publishing, the model owner wants to change both values from the Power BI service without republishing. What will they find?

• A. Neither parameter can be edited in the service; parameter values are fixed at the moment the .pbix is published
• B. Only parameters bound to a slicer can be edited in the service, and only by users holding Build permission
• C. Both parameters can be changed under the semantic model's Parameters settings and take effect at the next refresh
• D. The Region parameter can be edited in the service, but the Any-type parameter is not supported there

9. Each month the operations team drops a new CSV file with an identical column layout into a SharePoint document library, and the report must include every file automatically. Which connection approach requires the least ongoing maintenance?

• A. Import the newest CSV with the Text/CSV connector and replace the Source file path each month
• B. Connect with the Web connector to each file's URL and append the new query to the model every month
• C. Load each CSV into its own dataflow and merge the dataflows inside the semantic model
• D. Use the SharePoint folder connector and Combine files so new files are picked up on refresh

10. Column quality in Power Query Editor reports 0 percent errors for an Amount column, yet the load fails with conversion errors on 37 rows of a 2-million-row table. Why did the profile miss them?

• A. Column quality reports empty cells as errors, and these 37 rows contained blanks rather than error values
• B. Column quality only inspects columns with a Text data type, so numeric conversion errors are never counted
• C. Profiling runs over the first 1,000 rows by default, so errors further down the table are not sampled
• D. Errors are counted only after Close & Apply, so the preview always shows 0 percent until the model loads

11. Column distribution for an OrderID column with 10,000 rows shows 9,990 distinct values and 9,980 unique values. Which conclusion is correct before using OrderID as the one side of a relationship?

• A. Ten OrderID values appear more than once, so the column can't be the one side until duplicates are resolved
• B. Ten rows are null, because Distinct counts non-null values while Unique counts every row in the column
• C. The column is already a valid key, because Distinct and Unique only differ when the 1,000-row profile sample is incomplete
• D. Twenty OrderID values are duplicated, because Unique counts each occurrence of a repeated value

12. After a data type change, 12 rows in a 500,000-row Sales query show Error in the Quantity column because the source contains the text 'n/a'. The business wants those orders kept but treated as having no quantity. Which transformation meets that requirement?

• A. Keep errors on the Quantity column so that only the 12 affected rows are isolated in a separate query for review
• B. Replace errors on the Quantity column with null from the Any column group on the Transform tab
• C. Remove errors on the Quantity column from the Reduce rows group on the Home tab
• D. Change the Quantity column's data type to Text so the 'n/a' values load without producing any errors

13. A weekly refresh starts failing with "Expression.Error: The column 'Cust_Name' of the table wasn't found" after the source team renamed a field in the CSV export. Which action resolves the failure without losing the downstream steps?

• A. Edit the step that references Cust_Name so it uses the new column name and let the later steps flow from it
• B. Set the column's privacy level to Organizational so the mismatched header is ignored during the refresh
• C. Turn off 'Use first row as headers' so the renamed column falls back to a generic Column1 name
• D. Delete the query and rebuild it from the CSV so that the automatically generated steps use the new column names

14. A financial statement exported from Excel lists the Department name only on the first row of each block, with blank cells beneath it until the next department starts. Which Power Query transformation gives every row its department?

• A. Unpivot the Department column so that the blank cells become attribute-value pairs
• B. Fill Down on the Department column so each null is replaced by the last non-empty value above it
• C. Group By the Department column using the All rows operation, then expand the nested tables afterward
• D. Replace Values on the Department column, replacing null with the text of the previous row

15. A Customer table contains 'Contoso', ' Contoso', and 'Contoso ' as separate values. After load, a relationship to the Sales table fails with a duplicate-value error, yet a visual groups only two of the three variants together. What explains this, and what fixes it?

• A. The engine is case-sensitive and keeps all three; apply Format > Lowercase in Power Query before loading
• B. The engine keeps all three variants intact; enable Assume referential integrity on the relationship
• C. The engine trims trailing but not leading spaces; apply Format > Trim in Power Query before loading
• D. The engine trims leading but not trailing spaces; apply Format > Clean in Power Query before loading

16. A CSV export begins with two title lines, and the real column headers appear in the third row. When the file is loaded, Power Query names the columns Column1 through Column8. Which sequence of steps produces correct headers?

• A. Use First Row as Headers, then Remove Top Rows with a value of 2
• B. Transpose the table, remove the first two columns, then transpose it back
• C. Remove Bottom Rows with a value of 2, then Use Headers as First Row
• D. Remove Top Rows with a value of 2, then Use First Row as Headers

17. A Revenue column is imported as Decimal number. Totals in a card sometimes show values such as 1,204,555.99999998 instead of 1,204,556.00, and a RANKX measure over the column gives inconsistent ranks. Which data type change addresses both symptoms?

• A. Change the column to Fixed decimal number, which stores exactly four decimal places without floating error
• B. Change the column to Percentage format, which rounds the displayed values to two decimal places
• C. Change the column to Text so the source values are displayed exactly as they were entered
• D. Change the column to Whole number, which stores up to 19 digits and removes any floating-point imprecision

18. In Power Query a CallLength column is typed as Duration and a Logged column is typed as Date/Time/Timezone. After Close & Apply, an analyst is surprised by the data types shown for these columns in Table view. What happened?

• A. CallLength became Decimal number and Logged became Date/Time, because the model converts both types on load
• B. CallLength became a Whole number of seconds and Logged kept its timezone offset in a separate hidden column
• C. Both columns were loaded as Text, because Duration and Date/Time/Timezone exist only in Power Query
• D. Both columns failed to load, because the Binary, Duration, and Date/Time/Timezone types are unsupported

19. An analyst groups a 3-million-row Orders query by CustomerKey to compute the sum of Amount and a count of rows, but also needs each customer's original order rows nested so they can be expanded later. Which Group By aggregation provides the nested rows?

• A. All rows, which outputs the grouped rows as a table value in the new column
• B. Count distinct rows, which keeps one nested table per distinct customer
• C. Median, which retains the underlying rows so the middle value can be recalculated
• D. Count rows, which stores the row count together with a reference to the source table

20. A long table has Country, Date, and Value columns. Pivoting the Date column with the Don't aggregate option returns an error for several countries while other countries pivot correctly. What is the cause?

• A. Don't aggregate is only permitted when the pivoted column has a Text data type, not a Date type
• B. The Date column contains more than 100 distinct values, which exceeds the number of new columns a pivot can create
• C. Pivoting requires a numeric Value column, and Don't aggregate switches the column to Text
• D. Some country and date combinations have more than one Value row, so a single cell can't hold them

21. A planning spreadsheet lists the twelve months down the first column and one department per column header, with budget amounts in the cells. Analysts want one row per department with the months as fields for a matrix. Which transformation delivers that layout?

• A. Transpose the table, then promote the first row so the month names become the column headers
• B. Unpivot all of the department columns so that every amount becomes its own attribute-value row
• C. Pivot the month column and aggregate with Sum so that the departments become the rows
• D. Merge the department columns into a single delimited column and split it by rows

22. A sales extract has Country followed by one column per month, and the source adds a new month column with each refresh. The query must always produce Country, Month, and Amount rows without anyone editing the steps. Which unpivot choice should be used?

• A. Select the month columns and choose Unpivot Only Selected Columns
• B. Select the Country column and choose Unpivot Other Columns
• C. Select every column and choose Unpivot Columns to preserve the order
• D. Select the Country column and choose Pivot Column with Don't aggregate

23. An events table from a SQL database stores each event's payload as a JSON string in a text column, and the report needs the payload's status and region fields as regular columns. Which Power Query approach is appropriate?

• A. Parse the column as JSON so each value becomes a record, then expand the status and region fields
• B. Use the Web connector to send each JSON string to an API that returns the fields as columns
• C. Change the column data type to Text and use Column From Examples to extract the two words
• D. Split the column by the ':' delimiter and rename the resulting fragments to status and region by position

24. Before loading, an analyst must list every Sales row whose CustomerKey has no matching row in the Customer query, so the missing customers can be reported to the source team. Which merge join kind returns exactly those rows?

• A. Full outer join, which returns every row from both of the queries
• B. Left outer join, which returns all Sales rows plus matching Customer rows
• C. Left anti join, which returns only the Sales rows without a match
• D. Inner join, which returns only the rows that match in both tables

25. Three regional order files are appended into one query. The East file's discount column is named 'Disc', while the other two files use 'Discount'. What does the appended result contain?

• A. Separate Disc and Discount columns, with null in whichever one a source file lacks
• B. A single Discount column, because Append matches columns by their position rather than by name
• C. Only the columns that exist in all three files, with Disc and Discount both dropped
• D. An error, because Append requires every query to have identical column names and column order

26. A Sales fact table identifies products by the combination of ProductCode and Plant, and the Product dimension has the same two columns. The analyst can't create a relationship on both columns at once. Which preparation step enables a valid relationship?

• A. Create a many-to-many relationship on ProductCode and a second one on Plant, both set to bi-directional
• B. Add a column in Power Query that merges ProductCode and Plant into one key in both tables
• C. Enable 'Assume referential integrity' so the engine matches on both columns automatically
• D. Mark the Product table as the date table so the model can use composite keys for the relationship

27. Incremental refresh is configured with RangeStart and RangeEnd. The filter step is [OrderDate] >= RangeStart and [OrderDate] <= RangeEnd, and after a few refreshes some orders dated exactly at midnight on a partition boundary appear twice. What is the fix?

• A. Add Remove Duplicates as the final step so the duplicated boundary rows are dropped at load time
• B. Use >= RangeStart together with < RangeEnd so a boundary row satisfies only one partition's conditions
• C. Change both parameters to the Date data type so that partition boundaries no longer contain a time part
• D. Enable 'Only refresh complete days' so that rows on a partition boundary are excluded from every partition

28. A Customers query is referenced by three other queries that shape it for different fact tables. The base Customers query is only an intermediate step, yet it appears in the model and adds 400 MB. What should the analyst do?

• A. Convert the base Customers query into a custom function so it is only executed when the others invoke it
• B. Set the base Customers query's privacy level to Private so its rows are excluded from the model
• C. Delete the base Customers query and paste its steps into each of the three referencing queries
• D. Clear Enable load on the base Customers query so it stays available for reference but isn't loaded

29. A Sales table has an integer OrderYear column. When a report author drags it into a table visual it appears as a summed total such as 40,480 instead of listing the individual years. Which column property should the modeler change?

• A. Set the column's default summarization (Summarize by) to Don't summarize
• B. Set the column's data type to Text so the years are treated as categories
• C. Set the column's Data category to Year so the visual sorts it chronologically
• D. Hide the column and expose a measure MAX(Sales[OrderYear]) for the visual instead

30. A map visual plots a City column, but several cities such as Paris and Springfield land in the wrong country, and a GeoCode column holding values like AL is ambiguous between Alabama and Albania. What should the modeler configure?

• A. Enable bi-directional filtering between the Geography table and the map's fact table
• B. Set the Data category of the columns (for example City, State or Province, Country) so the map geocodes them correctly
• C. Create a calculated column that concatenates the city, state, and country and plot the concatenated text as the latitude field
• D. Change the City column's Summarize by setting to Count so each city is plotted only once

31. A bar chart uses a MonthName column from the Date table and shows April, August, and December first instead of January, February, March. The Date table also contains a MonthNumber column. What is the correct fix?

• A. Convert MonthName to a Date data type so the axis becomes continuous
• B. Create a hierarchy with MonthNumber above MonthName in the Date table
• C. In Column tools, set Sort by column on MonthName to MonthNumber
• D. Change the visual's sort to ascending by MonthName in the visual header

32. Reseller Sales has OrderDate, ShipDate, and DeliveryDate keys that all relate to one Date table. Users want a single matrix that shows orders placed in March 2026 broken down by the month in which they were shipped. Which design supports this?

• A. Set the ShipDate relationship as active and the OrderDate relationship as inactive, then switch between them in the visual's Format pane
• B. Set all three relationships to bi-directional so the Date table can filter Reseller Sales on any key
• C. Keep one Date table and write every measure with USERELATIONSHIP over the inactive ShipDate relationship
• D. Add a second date table, such as Ship Date, related to ShipDate so both dates can filter the matrix at once

33. A Customer table relates to Sales with a single-direction relationship. One measure must count the distinct customer countries for the products in the current filter context, but the modeler must not enable bi-directional filtering on the relationship because of its performance and slicer side effects. Which measure achieves this?

• A. Different Countries Sold = CALCULATE(DISTINCTCOUNT(Customer[Country]), CROSSFILTER(Customer[CustomerKey], Sales[CustomerKey], BOTH))
• B. Different Countries Sold = CALCULATE(DISTINCTCOUNT(Customer[Country]), USERELATIONSHIP(Customer[CustomerKey], Sales[CustomerKey]))
• C. Different Countries Sold = CALCULATE(DISTINCTCOUNT(Customer[Country]), ALL(Sales))
• D. Different Countries Sold = CALCULATE(DISTINCTCOUNT(Customer[Country]), TREATAS(VALUES(Sales[ProductKey]), Customer[CustomerKey]))

34. Customers can hold several bank accounts and accounts can be jointly owned by several customers. Transactions are recorded per account, and the report must slice transaction amounts by customer. How should the modeler relate Customer and Account?

• A. Create a direct many-to-many relationship between Customer and Account on the customer name column and set it to filter both ways
• B. Add a bridging table with one row per customer-account pair, related one-to-many from Customer and from Account
• C. Merge Customer into Account in Power Query so every account row repeats each of its owners' details
• D. Relate Customer directly to Transaction on AccountID with cross-filter direction set to Both

35. Mark as date table fails for a Calendar table with the message that the column must contain unique values, no nulls, and contiguous dates. The table was built from the distinct OrderDate values found in Sales. What is the underlying problem?

• A. The table was loaded from Power Query, and Mark as date table only accepts DAX calculated tables
• B. The table contains a MonthNumber column, and only a single-column table can be marked as a date table
• C. The table has no rows for days without sales, so the date values are not contiguous from beginning to end
• D. The Date column is stored as Date/Time, and Mark as date table accepts only columns stored with the Date data type

36. A model with 45 date columns across its Import tables is larger than expected, and Model view shows none of the extra tables that would explain the size. The modeler already has a proper marked Date table. Which Power BI Desktop option should be turned off, and why?

• A. Autodetect new relationships, because it creates a hidden bridge table for every date key that it detects
• B. Time intelligence quick measures, because they store a hidden date table in every table they touch
• C. Relative date filtering, because it caches a hidden calendar per date column for the current user
• D. Auto date/time, because it creates a hidden calculated date table for each date column in Import tables

37. A DirectQuery model over SQL Server relates a 3-billion-row Sales table to a Product dimension. Every ProductKey in Sales exists in Product, and query traces show OUTER JOINs that slow the visuals. Which relationship setting improves query efficiency?

• A. Assume referential integrity, so the generated queries use INNER JOIN instead of OUTER JOIN
• B. Cross-filter direction Both, so the source database can prune rows from either table first
• C. Cardinality One to one, so the engine can skip the join and read from Sales directly
• D. Make this relationship active, so the source query can use a hash join instead of a nested loop

38. A report has a slicer on Product[Color]. The measure Red Sales = CALCULATE([Sales], Product[Color] = "Red") keeps showing red sales even when the slicer is set to Blue. The requirement is that the measure returns blank whenever Red is excluded by the slicer. Which change is correct?

• A. Red Sales = CALCULATE([Sales], FILTER(ALL(Product), Product[Color] = "Red"))
• B. Red Sales = CALCULATE([Sales], KEEPFILTERS(Product[Color] = "Red"))
• C. Red Sales = CALCULATE([Sales], ALLSELECTED(Product[Color]), Product[Color] = "Red")
• D. Red Sales = CALCULATE([Sales], REMOVEFILTERS(Product[Color]), Product[Color] = "Red")

39. A bar chart shows Sales by Category, and a slicer on Product[Category] lets users pick a subset of categories. A percentage measure must show each visible category's share of the selected categories only, so the bars always sum to 100 percent. Which measure is correct?

• A. Share = DIVIDE([Sales], CALCULATE([Sales], ALL(Sales), ALL(Product)))
• B. Share = DIVIDE([Sales], CALCULATE([Sales], ALL(Product[Category]), ALL(Product[Color])))
• C. Share = DIVIDE([Sales], CALCULATE([Sales], ALLSELECTED(Product[Category])))
• D. Share = DIVIDE([Sales], CALCULATE([Sales], REMOVEFILTERS(Product)))

40. The company's fiscal year ends on June 30. A year-to-date sales measure must reset on July 1 rather than January 1, using the marked Date table. Which measure is correct?

• A. YTD Sales = TOTALYTD([Sales], 'Date'[Date], "7/1")
• B. YTD Sales = CALCULATE([Sales], DATESYTD('Date'[Date]), 'Date'[FiscalYear] = YEAR(TODAY()))
• C. YTD Sales = TOTALYTD([Sales], 'Date'[Date], 6)
• D. YTD Sales = TOTALYTD([Sales], 'Date'[Date], "6/30")

41. A report page is filtered to April 1 through April 10, 2026. A measure must return sales for the entire previous quarter (January 1 through March 31), not just the ten matching days shifted back. Which expression returns the full quarter?

• A. CALCULATE([Sales], PARALLELPERIOD('Date'[Date], -1, QUARTER))
• B. CALCULATE([Sales], SAMEPERIODLASTYEAR('Date'[Date]))
• C. CALCULATE([Sales], DATESBETWEEN('Date'[Date], MIN('Date'[Date]) - 90, MAX('Date'[Date]) - 90))
• D. CALCULATE([Sales], DATEADD('Date'[Date], -1, QUARTER))

42. A line chart by month must show rolling 12-month sales, so that the point for March 2026 sums April 2025 through March 2026 for every month on the axis. Which measure is correct with a marked Date table?

• A. Rolling 12M = CALCULATE([Sales], PREVIOUSYEAR('Date'[Date]))
• B. Rolling 12M = CALCULATE([Sales], DATESINPERIOD('Date'[Date], MAX('Date'[Date]), -12, MONTH))
• C. Rolling 12M = CALCULATE([Sales], DATESBETWEEN('Date'[Date], MIN('Date'[Date]), MAX('Date'[Date])))
• D. Rolling 12M = CALCULATE([Sales], DATEADD('Date'[Date], -12, MONTH))

43. Management wants the average revenue per customer, defined as the mean of each customer's total revenue rather than the mean of individual order lines. Sales has one row per order line with a CustomerKey column. Which measure is correct?

• A. Avg per Customer = AVERAGEX(Sales, Sales[Revenue] / DISTINCTCOUNT(Sales[CustomerKey]))
• B. Avg per Customer = DIVIDE([Total Revenue], COUNTROWS(Sales))
• C. Avg per Customer = AVERAGEX(VALUES(Sales[CustomerKey]), [Total Revenue])
• D. Avg per Customer = AVERAGE(Sales[Revenue])

44. Sales stores Quantity and UnitPrice per order line. A measure defined as Revenue = SUM(Sales[Quantity]) * SUM(Sales[UnitPrice]) returns totals far larger than the invoiced amounts. Which measure computes revenue correctly?

• A. Revenue = SUM(Sales[Quantity]) * AVERAGE(Sales[UnitPrice])
• B. Revenue = CALCULATE(SUM(Sales[Quantity]) * SUM(Sales[UnitPrice]), ALL(Sales))
• C. Revenue = PRODUCTX(Sales, Sales[Quantity] + Sales[UnitPrice])
• D. Revenue = SUMX(Sales, Sales[Quantity] * Sales[UnitPrice])

45. Executives want a 'typical order value' figure that is not distorted by a handful of very large orders. Sales has one row per order with an OrderValue column. Which single-aggregation measure best meets that requirement?

• A. Typical Order = AVERAGE(Sales[OrderValue])
• B. Typical Order = STDEV.P(Sales[OrderValue])
• C. Typical Order = MAX(Sales[OrderValue])
• D. Typical Order = MEDIAN(Sales[OrderValue])

46. An Inventory table holds a stock count per product per day, but counts are recorded only on days when a stocktake happens. A measure must show the most recent recorded stock at month end even when the last day of the month has no row. Which measure is correct?

• A. Stock = CALCULATE(SUM(Inventory[Qty]), LASTDATE('Date'[Date]))
• B. Stock = LASTNONBLANKVALUE('Date'[Date], SUM(Inventory[Qty]))
• C. Stock = CLOSINGBALANCEMONTH(SUM(Inventory[Qty]), 'Date'[Date])
• D. Stock = SUMX(VALUES('Date'[Date]), SUM(Inventory[Qty]))

47. A report author who is new to DAX needs a running total of Sales Amount by month, wants the generated DAX to be visible and editable afterward, and does not want to write the formula from scratch. Which feature fits?

• A. A visual-level Show value as > Running total setting, which writes a hidden measure into the model
• B. A what-if parameter, which generates a GENERATESERIES table and a SELECTEDVALUE measure for the total
• C. A quick measure using the Running total calculation, which creates a measure whose DAX can be edited
• D. A calculated column created from the Running total template, which stores the cumulative value on each row of the table

48. Customers must be filtered with a slicer on age bands such as 18-24 and 25-34, derived from a BirthDate column. The analyst first writes the banding logic as a measure but can't add it to the slicer. Why, and what should be created instead?

• A. Measures can't contain IF logic; create a calculated table that lists the bands and relate it to Customers on the BirthDate column
• B. Slicers accept only Text columns from the fact table; create the band as a column in Sales rather than in Customers
• C. Measures are evaluated only at refresh; create a quick measure so the band is materialized when the model loads
• D. Measures return a single value per filter context and can't be placed in a slicer; create a calculated column with the band

49. Sales managers want to drag a slider from 0 to 30 percent and see projected revenue after that discount recalculated in every visual on the page. Which modeling feature delivers both the slider and the value the measure uses?

• A. A what-if numeric range parameter, which creates a parameter table, a slicer, and a selected-value measure
• B. A field parameter, which creates a calculated table of measures that the slicer switches between in each visual
• C. A Power Query parameter of type Decimal Number that report readers change in the Filters pane
• D. A calculation group with one calculation item per discount percentage from 0 to 30

50. After a Time Intelligence calculation group is added in Model view, report authors notice that numeric columns no longer show the summation icon and can't be dropped straight into a visual's Values well. What caused this change?

• A. Adding a calculation group switches the model into DirectQuery mode, where implicit aggregation of columns isn't supported by visuals
• B. The calculation group's precedence property was set higher than the Values well allows for implicit aggregation
• C. Calculation groups convert every numeric column to Text, so columns must be wrapped in a measure before use
• D. Adding the calculation group enabled Discourage implicit measures, because calculation items apply only to explicit measures

51. A Targets table has a Region text column and monthly target amounts, but no relationship to the Region dimension because the modeler wants to avoid a relationship on a text column. A measure must return targets for the regions selected on the Region dimension. Which measure works?

• A. Target = CALCULATE(SUM(Targets[Amount]), CROSSFILTER(Region[RegionName], Targets[Region], BOTH))
• B. Target = CALCULATE(SUM(Targets[Amount]), FILTER(Targets, Targets[Region] = SELECTEDVALUE(Region[RegionName])))
• C. Target = CALCULATE(SUM(Targets[Amount]), TREATAS(VALUES(Region[RegionName]), Targets[Region]))
• D. Target = CALCULATE(SUM(Targets[Amount]), USERELATIONSHIP(Region[RegionName], Targets[Region]))

52. A 1.8-billion-row Import fact table stores one row per point-of-sale scan, but every report aggregates to store, product, and day. The model exceeds the capacity's memory limit. Which data-reduction technique addresses the size while keeping the reports correct?

• A. Convert the fact table's numeric columns to Text so the VertiPaq dictionary can compress them further
• B. Add a calculated column with the day-level total so visuals read the column instead of summing the rows
• C. Replace the integer keys with the descriptive Product and Store names so fewer lookup tables are needed
• D. Pre-summarize the table in Power Query with Group By to one row per store, product, and day before loading

53. A modeler needs a FullName column built by concatenating FirstName and LastName in a 40-million-row Customer table imported from SQL Server. Where should the column be created to get the best compression and refresh performance?

• A. As a Power Query custom column, or in the source query, so it is compressed like a regular loaded column
• B. As a measure using CONCATENATEX, because measures occupy no storage and are computed at refresh time
• C. As a calculated table that duplicates Customer with the extra column so the original table stays unchanged
• D. As a DAX calculated column, because VertiPaq compresses DAX calculated columns better than columns loaded from Power Query

54. Performance Analyzer shows a card visual with DAX query 40 ms, Visual display 30 ms, and Other 4,800 ms on a page that contains 32 visuals. What does the Other duration represent, and what is the appropriate fix?

• A. Time spent rendering web images and geocoding; remove the map and image visuals from the card's interactions
• B. Time the on-premises data gateway spent transferring the query result set to the service; move the semantic model to a Fabric capacity
• C. Time the visual spent waiting for other visuals to complete or doing background preparation; reduce the number of visuals on the page
• D. Time spent by the storage engine scanning the model in memory; add an aggregation table over the fact table

55. In DAX query view an analyst writes DEFINE MEASURE Sales[Margin %] = DIVIDE([Profit], [Sales]) followed by EVALUATE to test it against the model. The test looks right, and the measure must now exist in the semantic model. What is the most direct way to add it?

• A. Copy the EVALUATE result into a calculated table, which stores the measure definition permanently
• B. Publish the .pbix, because DEFINE MEASURE blocks are written into the model when the file is published
• C. Run the query a second time, because a DEFINE MEASURE that executes twice is persisted automatically
• D. Use the Update model with changes button so the query-scoped measure is added to the model

56. Performance Analyzer flags a slow measure defined as Red Sales = CALCULATE([Sales], FILTER('Product', 'Product'[Color] = "Red")). The result is correct, and existing Color filters may be replaced. Which rewrite improves performance without changing the result?

• A. Red Sales = CALCULATE([Sales], FILTER(VALUES('Product'[Color]), 'Product'[Color] = "Red"))
• B. Red Sales = CALCULATE([Sales], FILTER(ALL('Product'), 'Product'[Color] = "Red"))
• C. Red Sales = CALCULATE([Sales], 'Product'[Color] = "Red")
• D. Red Sales = SUMX(FILTER('Product', 'Product'[Color] = "Red"), [Sales])

57. A category manager wants to see how the sales rank of five product categories changes from month to month, with the top-ranked category always drawn on top for each month. Which visual is designed for this?

• A. A ribbon chart, which shows the highest-ranked category on top in each period and connects the rank changes
• B. A stacked column chart, which orders the segments by value within each month's column from largest to smallest
• C. A line chart with one line per category, which plots each category's sales value across the months
• D. A 100% stacked area chart, which shows each category's share of total sales in each month

58. A matrix must color each cell's background using a hex color that a DAX measure returns, such as #C00000 when a KPI is below target and #107C10 when it is above, so the color logic lives in one place. Which conditional formatting style should be used?

• A. Rules, with one rule per color that compares the KPI measure to the target value
• B. Field value, with the measure that returns the color codes selected as the field
• C. Data bars, with the positive and negative bar colors bound to the measure
• D. Color scale, with the measure supplying the minimum and maximum colors

59. A report author wants an on-canvas text summary that updates when readers change slicers, can be scoped to the whole report, one page, or selected visuals, and whose tone can be steered with a prompt. Which feature should the author add?

• A. The Copilot pane's report summary, which is saved into the report as a static text visual for readers
• B. A Q&A visual configured with the question 'summarize this report' so the generated answer text updates when slicers change
• C. A Narrative with Copilot visual, which summarizes the report, a page, or selected visuals and accepts custom prompts
• D. A text box with dynamic values mapped to measures, which recalculates the numbers on every refresh

60. In Power BI Desktop, an author selects 'Suggest content for this report' in the Copilot pane, but Copilot won't generate a page for a semantic model in which implicit measures were disabled after a calculation group was added. What is the reason?

• A. Copilot requires at least one custom visual in the report before it can suggest content
• B. Copilot builds pages only for models connected live to Analysis Services, not for Import models
• C. Copilot page creation requires the model's Q&A feature switch to be turned off first
• D. Copilot can't create report pages for semantic models that have implicit measures disabled

61. A design team requires that every chart's series colors follow the corporate palette across 30 reports, without editing each visual. Which report theme JSON property sets those colors?

• A. "dataColors", an array of hex values used for data series in visuals
• B. "background", which sets the fill applied to every visual's data points
• C. "tableAccent", which sets the accent color that visuals use for all series
• D. "foreground", which sets the default series color and the text color

62. Every page of a 12-page report must exclude cancelled orders, and report readers must not be able to remove or change that filter, although they should still see that it is applied. How should the author configure it?

• A. Add a slicer on order status to the first page and set its default selection to exclude cancelled
• B. Add the filter under Filters on all pages and lock it in the Filters pane
• C. Add the filter under Filters on this page on page 1 and enable Sync slicers for the rest
• D. Add the filter to each visual under Filters on this visual and hide the Filters pane

63. A relative date slicer set to 'Last 1 day' shows different rows for a reader in Sydney than the author expects at 9 AM Sydney time. What explains the behavior in the published report?

• A. Relative date slicers evaluate 'today' in the time zone of the capacity's home region
• B. Relative date slicers cache the author's local date when the report is published to the service
• C. Relative date slicers and filters are evaluated against the current time in UTC, not the reader's local time
• D. Relative date slicers use the time zone offset stored in the semantic model's Date/Time/Timezone columns

64. A report consumer with only read access wants a plain-language summary of the current page's visuals, respecting the slicers they applied, and cannot add visuals to the report. Which Copilot capability serves this?

• A. The Narrative with Copilot visual, which the consumer adds to the page while in reading view of the report
• B. Prep data for AI, which lets consumers write AI instructions that summarize the semantic model
• C. The Q&A visual's Copilot suggestions, which require Build permission on the semantic model
• D. The Copilot pane's report summary, available in the Power BI service without write permission on the report

65. Compliance requires that readers of a report can export the aggregated numbers they see in visuals but never the row-level underlying data. Where does the report author enforce this?

• A. In Report settings (File > Options and settings > Options), by allowing export of summarized data only
• B. In the semantic model settings, by turning off Build permission for every reader of the report
• C. In each visual's Format pane, by disabling the Export data option in the visual header for every visual
• D. In the workspace settings, by setting the license mode to Pro so exports are capped at 30,000 rows

66. A matrix shows monthly Sales Amount. The author wants a cumulative column that adds up the rows in the order they appear in the matrix, without creating a model measure. Which visual calculation does this?

• A. Running total = SUMX(ALL('Date'[Month]), [Sales Amount])
• B. Running total = RUNNINGSUM([Sales Amount])
• C. Running total = MOVINGAVERAGE([Sales Amount], 12)
• D. Running total = [Sales Amount] - PREVIOUS([Sales Amount])

67. A 'Reset view' button must return the page to its default layout, restoring hidden visuals and spotlight state, but it must leave whatever slicer selections the reader has made untouched. How should the bookmark be configured?

• A. Set the bookmark to Selected visuals and include only the slicers on the page so their state is preserved
• B. Enable Current page on the bookmark so it re-opens the page with the reader's slicer state
• C. Clear the Data option on the bookmark and keep Display, so it applies visibility without filters and slicers
• D. Clear the Display option on the bookmark and keep Data, so it applies slicers without visibility changes

68. When readers hover over a bar for a product, a small page with three visuals filtered to that product should appear instead of the default tooltip. Which configuration is required?

• A. Create a page, hide it from navigation, and set the bar visual's drillthrough target to that page
• B. Create a page, add the product field under Filters on this page, and set the bar visual's Tooltip type to Default
• C. Add the three visuals to the bar chart's Tooltips field well so they are shown when the bar is hovered
• D. Create a tooltip page (Tooltip canvas size, Tooltip slider on) with the product field in Tooltip fields

69. Selecting a region on a map currently shades the non-matching bars in a column chart while keeping them visible. Stakeholders want the column chart to show only the selected region's bars. How should the author change the interaction?

• A. Turn on Edit interactions, select the column chart, and choose the Highlight icon on the map
• B. Set the column chart's Data colors to transparent for categories outside the selection
• C. Turn on Edit interactions, select the map, and choose the Filter icon on the column chart
• D. Turn on Edit interactions, select the map, and choose the None icon on the column chart

70. A report gains new pages every sprint, and a navigation bar on every page must show a button per page with the correct names and order without the author editing buttons each time. Which feature meets this?

• A. A drillthrough button whose destination list is bound to the page names table
• B. A page navigator, which automatically syncs its buttons with the report's pages, names, and order
• C. A bookmark navigator, which creates a button for every bookmark in the Bookmarks pane
• D. A set of blank buttons with Page navigation actions, one per page, copied onto every page of the report

71. In a table visual sorted by Number of customers, a reader wants ties broken by Total revenue so that rows with the same customer count are ordered by revenue. How can the reader add the second sort column?

• A. Right-click the Number of customers header and choose Sort by another column > Total revenue
• B. Select the Total revenue header twice so it becomes the primary sort and the previous sort is kept
• C. Hold Shift and select the Total revenue column header to append it to the sort order
• D. Drag the Total revenue field into the Sort by bucket in the visual's Format pane

72. A page contains a 'Details' overlay made of a shape, a title text box, and four visuals that must be shown or hidden together by a single bookmark and moved as one unit when the layout changes. What should the author do?

• A. Give the six objects the same tab order value so that the bookmark treats them as a single navigable element
• B. Place the six objects on a hidden page and reference that page from the bookmark's Current page setting
• C. Set identical Layer order values on the six objects so they are hidden together when one is hidden
• D. Select the six objects, group them from the Selection pane, and toggle the group's visibility in the bookmark

73. A summary report and a detail report are in the same workspace, and users must right-click a customer in the summary and drill through to the detail report's customer page. The drillthrough entry doesn't appear in the service. Which condition is most likely missing?

• A. Cross-report drillthrough isn't enabled in the target report's settings, or the field names and data types don't match
• B. The two reports use different themes, so the drillthrough filter can't map the customer field between them
• C. The target report's customer page isn't hidden from navigation, and cross-report drillthrough targets must be hidden pages
• D. The summary report needs Build permission on the detail report's semantic model for the entry to appear

74. Regional managers want to swap the measure on a chart's axis or change its visual type for their own use, without edit permission and without affecting other readers. What must the author enable, and how do readers keep their version?

• A. Publish the report to an app; app users save changes as app-level bookmarks
• B. Enable Edit interactions for readers; they keep changes through persistent filters
• C. Enable Personalize visuals in Report settings; readers save their changes as personal bookmarks
• D. Grant readers the Contributor workspace role; they keep their changes with File > Save a copy

75. A line chart with five series is distinguished only by color. A color-blind stakeholder can't tell the lines apart, and the report must be made accessible without changing the theme. Which visual setting should the author apply?

• A. Increase the line stroke width so the color difference is more visible
• B. Add alt text describing which color represents each series
• C. Turn on markers and assign a different marker shape to each series
• D. Turn on data labels for each series and increase the label font size

76. A DirectQuery report must refresh its visuals only when new rows arrive, by polling a count of orders every 10 seconds rather than re-running every visual on a timer. The workspace is on a Fabric capacity. Which automatic page refresh setup is correct?

• A. Fixed interval set to 10 seconds, which re-queries only the visuals whose underlying data changed
• B. Change detection with two measures, one for row count and one for last order time
• C. Scheduled refresh of the semantic model every 10 seconds from the workspace settings
• D. Change detection with a measure such as a count of OrderID, checked every 10 seconds

77. A histogram must show how many orders fall into each $50 band of OrderValue (0-50, 50-100, and so on). OrderValue is a numeric column in the Sales table. What should the author create?

• A. A New group on OrderValue with Group type Bin and a bin size of 50
• B. A New group on OrderValue with Group type List, adding one group per band manually
• C. A measure that divides OrderValue by 50 and rounds down, placed on the X axis
• D. A field parameter on OrderValue with 50 as the step between values

78. Users typing 'revenue by territory' into a Q&A visual get no result because the model's fields are named SalesAmount and Region. Which Q&A setup step fixes this without renaming the fields?

• A. Turn on the Q&A feature switch in the tenant settings so field names are inferred
• B. Add revenue and territory as synonyms for the SalesAmount and Region fields in Q&A setup
• C. Add a calculated column named revenue and a calculated table named territory to the model
• D. Enable Row labels for the Sales table so Q&A can match any typed word to its columns

79. On a clustered column chart of average delivery time per carrier, the analyst must show each bar's uncertainty using upper and lower bound measures, and add a line at the 90th percentile of the values. Where are both of these added?

• A. A calculation group with items for upper bound, lower bound, and percentile
• B. The visual's Tooltips field well, which draws bounds as an overlay on hover
• C. The Analytics pane, which provides Error bars and a Percentile line for that visual type
• D. The Format pane's Data labels card, which supports upper and lower bound labels for each column

80. An analyst wants Power BI to flag unexpected spikes and dips in daily website sessions and explain the likely contributing factors. Which visual and feature combination supports this?

• A. A scatter chart of sessions by day with automatic clustering enabled in the Analytics pane
• B. A decomposition tree with Explain the increase applied to each branch
• C. A card visual with a data alert configured on the published dashboard tile
• D. A line chart with a date axis, with Find anomalies enabled in the Analytics pane

81. A team of analysts must create and edit reports in a workspace but must not publish or update the workspace's app, and the workspace admin has not delegated app updates. Which role should they be assigned?

• A. Contributor, which can publish and edit content but can update the app only if the admin allows it
• B. Member, which can edit content but is blocked from app publishing unless the admin promotes it
• C. Viewer, which can edit reports it has been granted Build permission on but can't publish apps
• D. Admin, with the 'Allow contributors to update the app' setting turned off for the workspace

82. One app must serve both the sales and finance teams: sales should see three pages and finance a different set of content, without the team maintaining two apps. How should the app be configured?

• A. Publish the app twice under different names and assign each team to one copy
• B. Share the workspace with both teams and use report-level filters per team
• C. Apply row-level security roles named Sales and Finance to the app's reports
• D. Create two audiences in the app and show or hide content per audience

83. A report is shared with an external partner through direct access. The author hid several sensitive columns in the report and assumed the partner can't see them. Which statement about the partner's access is accurate?

• A. Hidden fields are removed from the shared semantic model copy that the service creates for each share
• B. The partner can see the hidden fields only if they are also granted Build permission on the model
• C. Sharing the report also grants read access to its semantic model, so hidden fields aren't protected; use RLS or OLS
• D. The partner sees only the fields placed on visuals, because sharing applies the report's field list as a security filter

84. Row-level security limits regional managers to their own region, and the roles have members assigned. One manager who was given the Member workspace role still sees every region. Why, and what is the fix?

• A. RLS applies only to users with Viewer permission; Admin, Member, and Contributor bypass it, so assign the manager Viewer
• B. RLS only applies once the semantic model is certified; ask an authorized reviewer to certify it
• C. Member role users need the role's DAX filter re-validated with Test as role before it takes effect
• D. The manager belongs to two roles and the roles are combined with AND logic; remove the second role assignment

85. Two hundred sales reps must each see only their own accounts in a shared report, and a UserAccess table maps each rep's sign-in name to account IDs. Which approach avoids creating and maintaining 200 roles?

• A. One role with a DAX filter such as [UserEmail] = USERPRINCIPALNAME() on the UserAccess table
• B. One role per rep with a static filter such as [RepName] = "Ana Souza", assigned individually
• C. A single role with no filter, relying on the Viewer workspace role to restrict rows automatically
• D. A Power Query parameter for the rep's email that each rep edits in the service before viewing

86. A modeler assigns members to an RLS role in the Power BI service by adding the Microsoft 365 group that the sales team uses for Teams and email, but the assignment is rejected. Which group types can be added?

• A. Individual users only; group assignments are not supported for row-level security roles
• B. Security groups, distribution groups, and mail-enabled groups; Microsoft 365 groups aren't supported for RLS roles
• C. Microsoft 365 groups only, provided the group is first synchronized to the Power BI admin portal
• D. Any Microsoft Entra group type, provided the semantic model is hosted in a workspace on a Premium or Fabric capacity

87. An analyst with a Fabric free license tries to apply a Confidential sensitivity label to a semantic model in a shared workspace and finds the option unavailable, although labels are enabled for the tenant. What is the most likely reason?

• A. Applying labels to items outside My workspace requires a Pro or PPU license plus Edit permission on the item
• B. Labels can be applied only in Power BI Desktop before publishing, never to items that already exist in the service
• C. Sensitivity labels apply only to reports and dashboards, so semantic models can't be labeled directly
• D. The workspace must be assigned to a Premium capacity before any of its items can carry a label

88. A data engineer has finished a well-tested semantic model and wants it flagged as endorsed for reuse today, but is not in the group the Power BI admin authorized for certification. What can the engineer do?

• A. Certify the semantic model, because certification is available to anyone with Build permission on the model
• B. Apply a Highly Confidential sensitivity label, which appears as an endorsement badge in the hub
• C. Request certification through the workspace's Contributor role, which grants certification rights
• D. Promote the semantic model, which any owner or member with write permission on the workspace can do

89. A department needs scheduled refresh for semantic models owned by several analysts against an on-premises SQL Server, plus a DirectQuery report on the same server. One analyst has installed a personal-mode gateway on their laptop. Why is that insufficient?

• A. Personal mode supports only Import refresh for its installing user and can't serve DirectQuery or shared data sources
• B. Personal mode can only connect to cloud sources such as Azure SQL Database, not to servers on the on-premises network
• C. Personal mode requires a Premium capacity, whereas the standard gateway works with Pro workspaces
• D. Personal mode gateways expire after 30 days and must be reinstalled by a tenant administrator

90. A semantic model in a workspace on a Fabric capacity needs data refreshed every 30 minutes during business hours, and the support team's alias must be emailed when a refresh fails. Which configuration is possible in the semantic model settings?

• A. Up to 24 scheduled refresh times per day, with failure notifications configured in the workspace's contact list
• B. Up to 48 scheduled refresh times per day, with additional failure-notification contacts entered in the refresh settings
• C. Up to 8 scheduled refresh times per day, because every workspace shares the same daily refresh cap
• D. Unlimited refresh times per day, but failure notifications can be sent only to the semantic model owner

91. A colleague edited a published report in the Power BI service, adding a new page. The original author then republishes the same .pbix from Power BI Desktop and confirms the replace prompt. What is the result?

• A. Publishing fails with a conflict error, because a report edited in the service can no longer be overwritten from Desktop
• B. A second copy of the report is created with a numeric suffix, leaving the service-edited report intact
• C. The semantic model and the report are replaced by the Desktop versions, so the page added in the service is lost
• D. The semantic model is replaced but the report keeps the service-added page, because reports merge on publish

92. A manager wants an email whenever a dashboard tile showing weekly returns exceeds 500. The tile was pinned from a clustered column chart. Why is the Manage alerts option unavailable, and what change enables it?

• A. Alerts work only on tiles in a Premium workspace; move the dashboard to a workspace on a capacity
• B. Alerts require an existing report subscription on the source report; create the weekly subscription first
• C. Alerts work only on streaming tiles created with Add tile; rebuild the tile from a streaming dataset
• D. Alerts work only on tiles pinned from gauge, KPI, and card visuals; pin a card showing the returns value

93. A deployment pipeline moves a semantic model from Development to Production. In Development the model reads DevSQL01, but in Production it must read ProdSQL01 without anyone editing the model after each deployment. What should be configured?

• A. A data source rule (or parameter rule) on the Production stage's semantic model that swaps the source
• B. A separate .pbix for Production whose queries point at ProdSQL01, published manually after each deployment
• C. A gateway data source named ProdSQL01 on the Production workspace, which overrides the model's source
• D. A sensitivity label on the Production model that redirects queries to the production server

94. A curated set of reports and dashboards must be distributed to 300 people with a simple, consistent navigation experience. Which distribution method fits best?

• A. Publishing a Power BI app from the workspace
• B. Emailing a weekly subscription snapshot to everyone
• C. Granting the 300 people the Viewer role
• D. Sharing a direct link to each report with everyone

95. A semantic model has been reviewed by the central data team and should be signposted to users as trustworthy and officially supported. What should be applied?

• A. A Confidential sensitivity label from Purview
• B. Certification through the endorsement settings
• C. A Promoted badge applied by the model's owner
• D. A Master data label set in the workspace settings

96. A dataset must be shared so other authors can build their own reports on it without duplicating the model. What should be configured?

• A. Read permission on the semantic model
• B. The Viewer role in the model's workspace
• C. Reshare permission on the semantic model
• D. Build permission on the semantic model

97. Row-level security has been defined with a role named RegionManager. What additional step is required before it takes effect for users?

• A. Publish the report to a workspace on a Premium or Fabric capacity
• B. Assign users or security groups to the role in the Power BI service
• C. Add the users' UPNs to a column in the RegionManager fact table
• D. Turn on the relationship's Apply security filter in both directions

98. Content in a workspace must carry a classification that follows exported files into Excel and PDF. Which capability provides that?

• A. Certified endorsement applied to every report in the workspace
• B. Row-level security roles applied to the underlying semantic model
• C. Workspace-level access restricted to a Microsoft Entra security group
• D. Sensitivity labels from Microsoft Purview Information Protection

99. Finance users should receive a formatted snapshot of a report in their inbox every Monday morning without opening Power BI. What should be configured?

• A. A scheduled refresh set to run each Monday at 7 AM
• B. A dashboard pinned in the Teams channel
• C. A report subscription with a weekly schedule
• D. A data alert on a KPI tile checked every 24 hours

100. A semantic model refresh must run four times per day, but the workspace is on a shared capacity. What limitation should the analyst plan around?

• A. Shared capacity blocks scheduled refresh for models that use an on-premises gateway
• B. Shared capacity permits scheduled refresh only for DirectQuery models
• C. Shared capacity enforces a two-hour refresh limit but places no cap on daily refreshes
• D. Shared capacity caps scheduled refreshes per day, unlike dedicated capacity