Power BI interview questions, DAX included.

1. Power BI Desktop, the service, reports and dashboards: what’s the difference?Easy

Desktop is the Windows app where you model data and build reports; the service is the cloud where you publish, share, refresh and secure them. A report has one or more pages on one semantic model (today’s name for a dataset). A dashboard is a single page of tiles pinned from reports; it lives only in the service and can’t be filtered.6

2. What is DAX, and where do you write it?Easy

DAX (Data Analysis Expressions) is the formula language of Power BI semantic models. You write it in measures, calculated columns, calculated tables, row-level security rules, visual calculations and DAX queries.

Total Sales = SUM ( Sales[SalesAmount] )                -- a measure
Line Value  = Sales[Quantity] * Sales[UnitPrice]        -- a calculated column, row by row

3. Calculated column or measure?Easy

A calculated column is computed row by row at refresh and stored, so it costs memory: use it for what you slice or filter by. A measure is computed at query time for each cell and stores nothing: use it for anything you aggregate. On an axis or a slicer, a column; in Values, a measure.

Price Band        = IF ( Sales[UnitPrice] >= 2000, "Premium", "Regular" )   -- column
Avg Selling Price = DIVIDE ( [Total Sales], [Units Sold] )                   -- measure

4. SUM or SUMX?Easy

SUM adds up one column. SUMX is an iterator: it evaluates an expression row by row, then adds the results. Multiplying two SUMs is the classic wrong answer.

Revenue       = SUMX ( Sales, Sales[Quantity] * Sales[UnitPrice] )        -- right
Revenue Wrong = SUM ( Sales[Quantity] ) * SUM ( Sales[UnitPrice] )      -- wrong

5. COUNT, COUNTA, COUNTROWS or DISTINCTCOUNT?Easy

COUNT counts non-blank numbers, dates and strings; COUNTA also counts TRUE/FALSE. COUNTROWS counts a table’s rows and is Microsoft’s pick for that. DISTINCTCOUNT counts distinct values, BLANK included, and its totals don’t add up: an order in two categories counts once in each and once overall.7

Order Lines = COUNTROWS ( Sales )
Customers   = DISTINCTCOUNT ( Sales[CustomerID] )

6. DIVIDE or the / operator?Easy

DIVIDE returns BLANK, or your alternate result, when the denominator is zero or BLANK. Microsoft recommends it whenever the denominator could be zero, and / for constants; blank results also drop out of visuals.7

Avg Order Value = DIVIDE ( [Total Sales], [Orders] )   -- BLANK for a store with no orders

7. What is a star schema, and why not a snowflake?Easy

A star schema has fact tables (events you summarise: sales, stock) and dimension tables (what you filter by: date, product, store). A snowflake chains dimensions: Product → Subcategory → Category. Microsoft says one flattened table generally wins: fewer tables, shorter filter chains, and hierarchies can’t span tables.8

8. Explain cardinality and cross-filter directionEasy

Cardinality says which side holds unique values: one-to-many (the usual), one-to-one or many-to-many. Cross-filter direction says which way filters flow: Single, from the one side, or Both, which can slow queries and make paths ambiguous, so use it only where needed. Only one relationship between two tables can be active.9

9. What is Power Query, and should you merge or append?Easy

Power Query is Power BI’s data-preparation tool; each click becomes M code. Merge joins queries on matching columns, like a SQL join (left outer, right outer, full outer, inner, left anti, right anti); append stacks queries with the same columns, like UNION ALL. Append twelve monthly files; merge in each store’s region.10

10. Slicers or the Filters pane?Easy

Both filter data. A slicer sits on the page for readers to change; the Filters pane holds filters for one visual, one page or all pages. Slicers suit everyday choices (region, month); the pane suits fixed rules (leave out test stores).

11. Drill-down or drill-through?Easy

Drill-down moves down a hierarchy inside one visual: FY → quarter → month. Drill-through jumps to another page, filtered to what you right-clicked, such as one city’s detail page; add the field to that page’s drill-through well.

12. Which visual would you choose for each question?Easy

Give the reason, not the name: a line for a trend, sorted bars to compare categories, a stacked bar for parts of a whole, a scatter for two measures, a card for one number, a table for exact figures.

13. Row context or filter context?Medium

Filter context is every filter on a calculation: slicers, the visual’s rows and columns, the Filters pane, CALCULATE’s filters. Row context is the current row, in calculated columns and iterators. It filters nothing by itself, so the first column below shows the grand total on every row:

-- Calculated columns in the Product table
Sales (wrong) = SUM ( Sales[SalesAmount] )                 -- the grand total, every row
Sales (right) = CALCULATE ( SUM ( Sales[SalesAmount] ) )    -- each product's own sales

14. What does CALCULATE do?Medium

It evaluates an expression in a modified filter context. A filter argument adds a filter if that column isn’t filtered yet and overwrites it if it is; KEEPFILTERS intersects instead. Arguments can be Boolean expressions, tables (often from FILTER) or modifiers: REMOVEFILTERS, ALL, ALLEXCEPT, KEEPFILTERS, USERELATIONSHIP and CROSSFILTER.2

South Sales      = CALCULATE ( [Total Sales], Store[Region] = "South" )
-- A slicer on West still shows South: the filter was overwritten

South Sales Kept = CALCULATE ( [Total Sales], KEEPFILTERS ( Store[Region] = "South" ) )
-- Now West shows blank: South and West is nothing

15. ALL, ALLEXCEPT, ALLSELECTED or REMOVEFILTERS?Medium

REMOVEFILTERS (or ALL used as a filter) clears filters from columns or a table, whatever the slicers say. ALLEXCEPT clears every filter on a table except the columns you name. ALLSELECTED drops the visual’s own filters but keeps slicers, giving the visual’s total. When you only mean “remove filters”, Microsoft recommends REMOVEFILTERS.2

% of All Stores   = DIVIDE ( [Total Sales], CALCULATE ( [Total Sales], REMOVEFILTERS ( Store ) ) )
% of Visual Total = DIVIDE ( [Total Sales], CALCULATE ( [Total Sales], ALLSELECTED ( Store ) ) )
Region Sales      = CALCULATE ( [Total Sales], ALLEXCEPT ( Store, Store[Region] ) )  -- on each store's row

16. Why do you need a date table, and how do you mark it?Medium

Classic time intelligence needs a marked date table (Table tools → Mark as date table): unique dates, no blanks, no missing days. Build it with CALENDAR or Power Query, and turn off auto date/time.11 Microsoft now recommends calendar-based time intelligence (preview in October 2026): you tag columns as year, quarter, month or week, so fiscal and week calendars work.12

17. Year to date for India’s April–March financial yearMedium

TOTALYTD’s year end defaults to 31 December, so pass 31 March. The string follows the file’s locale: "3-31" in a US-locale file, "31-3" in a day-first one, so check one month by hand. A calendar needs no string.13

Sales FYTD = TOTALYTD ( [Total Sales], 'Date'[Date], "3-31" )            -- classic

Sales FYTD (calendar) = TOTALYTD ( [Total Sales], 'FY Calendar' )    -- calendar-based, preview

18. Year-on-year growth: SAMEPERIODLASTYEAR or PREVIOUSYEAR?Medium

SAMEPERIODLASTYEAR shifts the dates in context back a year, like DATEADD(dates, -1, YEAR): October 2026 becomes October 2025. PREVIOUSYEAR returns all of last year, whatever month you’re on. Keep last year in a variable and use DIVIDE, so a new store shows blank, not an error.13

Sales LY = CALCULATE ( [Total Sales], SAMEPERIODLASTYEAR ( 'Date'[Date] ) )

Sales YoY % =
VAR SalesLY = [Sales LY]
RETURN
    DIVIDE ( [Total Sales] - SalesLY, SalesLY )

19. Write a running totalMedium

Take the cell’s last date, clear the date filters and keep every date up to it (ALLSELECTED starts from the reader’s first selected date). The newer alternative, a visual calculation, runs DAX on the visual’s own data: simpler, but not reusable in other visuals, dashboards or exports.15

Running Sales =
VAR LastDateShown = MAX ( 'Date'[Date] )
RETURN
    CALCULATE (
        [Total Sales],
        FILTER ( ALL ( 'Date' ), 'Date'[Date] <= LastDateShown )
    )

Running sum = RUNNINGSUM ( [Total Sales] )     -- the visual calculation

20. Import, DirectQuery, Dual or Direct Lake?Medium

Often asked in IT-services rounds. A composite model mixes modes, table by table.1617

Worth knowing: Direct Lake on SQL endpoints can fall back to DirectQuery (for SQL views); on OneLake it never does.

21. What is query folding, and why does it matter?Medium

Query folding is Power Query translating your steps into the source’s language, such as SQL, so the source does the work; CSV and Excel files don’t fold. Put filters, removed columns and joins first, and check with View Native Query where available. Incremental refresh depends on it.18

22. Set up row-level security, static and dynamicMedium

Write each role’s DAX filter in Desktop (Modeling → Manage roles), publish, add people or groups in the service, and check with Test as role. Static rules fix a value; dynamic ones use USERPRINCIPALNAME(). RLS restricts only Viewer permissions, roles add up, and it hides rows; object-level security hides columns.4

[Region] = "South"                          -- static role, filter on Store

Store[Region]                               -- dynamic role, filter on Store
    IN CALCULATETABLE (
        VALUES ( UserRegion[Region] ),
        UserRegion[Email] = USERPRINCIPALNAME ()
    )

23. Two dates, one active relationship: report sales by ship dateMedium

Only one of Sales’ two date relationships can be active; ShipDate’s is inactive (dashed). USERELATIONSHIP switches it on inside CALCULATE for one measure. With row-level security on those tables it can error, so secured models use a separate Ship Date table.19

Sales by Ship Date =
CALCULATE ( [Total Sales], USERELATIONSHIP ( Sales[ShipDate], 'Date'[Date] ) )

24. Rank products by sales. What about ties?Medium

RANKX ranks the current value among an expression’s values over a table. Order defaults to descending; ties default to Skip (1, 2, 2, 4), and Dense gives 1, 2, 2, 3. Rank over ALLSELECTED to respect the slicers, and hide the meaningless rank on the total row.20

Product Rank =
IF (
    ISINSCOPE ( 'Product'[Product] ),
    RANKX ( ALLSELECTED ( 'Product'[Product] ), [Total Sales], , DESC, Dense )
)

25. Make a visual’s title follow the slicersMedium

SELECTEDVALUE(column, alternate) returns the value when exactly one is filtered, and the alternate otherwise; it’s short for IF(HASONEVALUE(…), VALUES(…), …). Write a text measure and set the visual’s title with fx. A related trap: VALUES can add a blank row for unmatched keys; DISTINCT doesn’t.20

Sales Title =
"Sales in " & SELECTEDVALUE ( Store[City], "all cities" )
    & ", " & SELECTEDVALUE ( 'Date'[FY], "all years" )

26. Why use variables in DAX?Medium

A variable is calculated once, so nothing is evaluated twice: in Microsoft’s year-on-year example, the version with a variable ran “in about half the query time”. Variables also read better, are easy to debug (return one to inspect it) and replace the old EARLIER function.14

Sales YoY % =
VAR SalesLY = [Sales LY]
RETURN
    -- DIVIDE ( [Total Sales] - SalesLY, SalesLY )
    SalesLY        -- return the variable while you debug, then switch back

27. When do you need a gateway, and how often can a model refresh?Medium

A gateway lets the service reach data it can’t reach directly, such as an office SQL Server; cloud sources need none unless mixed with one that does, and Microsoft prefers the enterprise (standard) gateway to personal mode. Scheduled refresh: 8 a day on shared capacity, 48 on Premium, PPU or Fabric; limits of 2 and 5 hours per refresh; schedules pause after two months without visits.3

28. Workspace roles, apps and sharingMedium

Admins manage a workspace and its people; Members add people with lower roles and allow resharing; Contributors create and edit content; Viewers only view. Share widely through an app, a packaged read-only set of reports, rather than workspace access.21

29. Why does a measure in a calculated column differ from SUM?Hard

In a calculated column, SUM ignores the row and returns the grand total. A measure reference acts as if wrapped in CALCULATE, so the row becomes a filter: context transition. It filters on every column, so identical duplicate rows each show their combined total, and it’s costly inside iterators over big tables.2

-- Calculated columns in the Product table
Sales via SUM       = SUM ( Sales[SalesAmount] )               -- grand total on every row
Sales via measure   = [Total Sales]                            -- this product's sales
Sales via CALCULATE = CALCULATE ( SUM ( Sales[SalesAmount] ) ) -- the same as the line above

30. Show the top 3 products in each regionHard

A visual-level Top N filter ranks across the whole visual, not per region, so do it in the measure: in a matrix of Region then Product, TOPN runs in each region’s context and KEEPFILTERS keeps those three. TOPN returns every row tied at third place, so the product name breaks ties.22

Sales Top 3 in Region =
VAR Top3 =
    TOPN ( 3, ALLSELECTED ( 'Product'[Product] ), [Total Sales], DESC, 'Product'[Product], ASC )
RETURN
    CALCULATE ( [Total Sales], KEEPFILTERS ( Top3 ) )

31. Customers and accounts are many-to-many. Model joint accountsHard

Microsoft’s own example: keep Customer and Account as dimensions, add a bridging table (AccountCustomer, one row per pair), relate them one-to-many, and let one relationship filter both ways so customer filters reach Transaction. Customer balances become non-additive, since joint holders share an account. Don’t relate the dimensions directly as many-to-many. To keep the model single-direction, switch direction inside one measure:23

Customer Balance =
CALCULATE (
    SUM ( 'Transaction'[Amount] ),
    CROSSFILTER ( AccountCustomer[AccountID], Account[AccountID], Both )
)

32. A report is slow. How do you find and fix the cause?Hard

Measure first: Performance Analyzer splits each visual’s time into DAX query, DirectQuery, visual display and other; run the slowest query in DAX query view. Then fix the model (drop unused columns and rows, cut granularity and cardinality, remove needless bi-directional filters), the DAX (variables; Boolean filters rather than FILTER over a table, which Microsoft calls more efficient) and the page (fewer visuals).24

33. How does incremental refresh work?Hard

Create Date/Time parameters with the reserved, case-sensitive names RangeStart and RangeEnd, filter the fact table with them, set a policy (keep five years, refresh ten days) and publish: the service partitions the table and refreshes only recent partitions. The filter must fold. It runs on Pro, PPU, Premium and Embedded; a real-time DirectQuery partition needs Premium, PPU or Embedded, and you can’t download the .pbix afterwards.25

// Power Query (M). OrderDate must be Date/Time, like the parameters;
// >= at one end and < at the other, so no row lands in two partitions
#"Filtered Rows" =
    Table.SelectRows ( Source, each [OrderDate] >= RangeStart and [OrderDate] < RangeEnd )

34. Stock doesn’t add up over time. How do you measure it?Hard

Stock adds up across products but not over time: a month’s stock is its last day’s. That’s a semi-additive measure, a PL-300 skill. LASTNONBLANK takes the last date with data, so a month in progress still shows a value; CLOSINGBALANCEMONTH takes the month’s last date. BFSI rounds ask it about account balances.131

Closing Stock =
CALCULATE (
    SUM ( Stock[Units] ),
    LASTNONBLANK ( 'Date'[Date], CALCULATE ( SUM ( Stock[Units] ) ) )
)

35. What problem do calculation groups solve?Hard

Ten measures with five time views each means fifty measures. A calculation group defines each view once, with SELECTEDMEASURE() standing for whatever measure the visual shows; create it in Model view or TMDL view. It turns on “discourage implicit measures”, since items apply only to explicit measures, and gives every measure the variant data type.26

-- Calculation items in a group named "Time Intelligence"
Current = SELECTEDMEASURE ()
FYTD    = CALCULATE ( SELECTEDMEASURE (), DATESYTD ( 'Date'[Date], "3-31" ) )
LY      = CALCULATE ( SELECTEDMEASURE (), SAMEPERIODLASTYEAR ( 'Date'[Date] ) )

36. Let readers switch a chart between sales, units and ordersHard

Field parameters (Modeling → New parameter → Fields) make a small table of measures or columns. Put it in a slicer and the visual, and readers choose what it shows, axes included. Limits: no implicit measures, not as linked fields of a drill-through or tooltip page, and no AI visuals or Q&A.26

Metric = {
    ( "Sales",  NAMEOF ( [Total Sales] ), 0 ),
    ( "Units",  NAMEOF ( [Units Sold] ),  1 ),
    ( "Orders", NAMEOF ( [Orders] ),      2 )
}

37. Managers see their own sales and everyone’s below them. How?Hard

Dynamic RLS on a hierarchy: store it in Employee (EmployeeID, ManagerID, Email), turn it into a path with PATH, and keep rows whose path contains the signed-in user’s ID; the filter flows on to Sales.274

-- Calculated column: "112|14|3" means 3 reports to 14, who reports to 112
Org Path = PATH ( Employee[EmployeeID], Employee[ManagerID] )

PATHCONTAINS (                                  -- role filter on Employee
    Employee[Org Path],
    LOOKUPVALUE ( Employee[EmployeeID], Employee[Email], USERPRINCIPALNAME () )
)

38. Pro, Premium Per User or Fabric capacity: who needs which licence?Hard

Creating and sharing in a workspace needs Pro or Premium Per User (PPU), and so does viewing, unless the workspace is on an F64 or larger Fabric capacity (or a P capacity), where a free licence and the Viewer role do. PPU adds Premium features per user (48 refreshes a day, models over 1 GB) but isn’t a Fabric capacity; P SKUs are being retired for F SKUs.21

39. Copilot, visual calculations and DAX query view: what to know in 2026Hard

Copilot needs a paid Fabric (F2 or higher) or Premium (P1 or higher) capacity, not just Pro or PPU; on a capacity outside the US and EU data boundaries, as in India, an admin must allow processing outside the region. Microsoft warns that unprepared models give generic or inaccurate answers.5 PL-300’s skills list includes four Copilot tasks, visual calculations and DAX query view.1

40. Walk us through a dashboard you builtHard

It decides most second rounds: two minutes, five parts, then questions.

1. The question and reader: “Store managers needed to see which products would run out before the festive weekend.”
2. The data: sources, size, Power Query cleaning.
3. The model: star schema, date table, key measures.
4. The design: visual choices, drill-through, who sees what.
5. The result: the decision it drove, how you checked the numbers, what you’d change.