PL-300 domain - 35% of the exam

Model the Data

Model the Data is 35% of the PL-300 exam. These are the objectives it covers, each with practice questions, with every answer explained.

The domain in numbers

78
Practice questions
3
Objectives
9
Easy
49
Medium
20
Hard

Where people go wrong

  • Expecting one Date table to serve two roles at once

    The wrong answers try to make a single Date table filter OrderDate and ShipDate together: a second active relationship, a bidirectional cross-filter, or USERELATIONSHIP. Only one relationship between two tables can be active, and cross-filter direction controls propagation, not which role applies. USERELATIONSHIP switches the role inside one measure at a time. To filter both roles in one visual, duplicate the Date dimension so each role has its own active relationship to Sales.

    See it in a sample question
  • Expecting CALCULATE filters to intersect by default

    The wrong answers assume a CALCULATE filter argument combines with a filter already on the same column, or that KEEPFILTERS discards the argument or clears filters. By default the new filter overwrites the existing one, so Blue Revenue inside a visual filtered to Red returns Blue sales only. KEEPFILTERS is what intersects the new filter with the existing one. Clearing filters is the job of REMOVEFILTERS or ALL.

    See it in a sample question
  • Using FILTER and Boolean arguments the wrong way round

    The wrong answers misjudge when a CALCULATE filter needs FILTER. A Boolean filter argument cannot reference a measure, so a test on [Profit] needs FILTER over VALUES, and neither KEEPFILTERS nor ALL lifts that restriction. Where a Boolean expression is enough, as for Red product sales in an Import model, passing a table expression is the pattern the documentation says to avoid: FILTER over Product iterates row by row, with or without ALL, and CALCULATETABLE also produces a table expression.

    See it in a sample question
  • Picking a date-set function where an aggregator is needed

    The wrong answers choose a time-intelligence function that returns the wrong kind of result or the wrong period. DATESYTD and DATESMTD return a table of dates and do not aggregate an expression, while TOTALYTD and TOTALQTD evaluate one. TOTALMTD sums across the month-to-date range instead of returning the closing-day value that CLOSINGBALANCEMONTH gives, and OPENINGBALANCEMONTH reads the first date rather than the last.

    See it in a sample question
  • Treating calculated columns as free of storage cost

    The wrong answers give calculated columns the behaviour of measures: no stored result, evaluation on demand, or no effect on size. A calculated column is evaluated at data refresh and its result is stored in the model, so it increases model size on a large fact table. Calculated columns compress less efficiently than Power Query computed columns and lengthen refresh, so the guidance prefers a custom column in Power Query.

    See it in a sample question

Objectives in this domain

Sample questions from this domain

Free sampleModel the Datamedium

How many active relationships can exist between any two tables in a Power BI semantic model at the same time?

  • AUp to three active relationships when the table pair is a role-playing dimension.
  • BUnlimited active relationships as long as their cardinalities differ from each other.
  • CExactly one active relationship is allowed per pair of related tables. Correct
  • DTwo active relationships when one is single direction and the other is both.
Only one active relationship is permitted between any two tables; extras must be inactive. Microsoft guidance is explicit that there can only be one active filter propagation path between two model tables. Additional relationship paths between the same pair must be marked inactive, and inactive relationships are only engaged at calculation time using USERELATIONSHIP.

Why A is wrong: Role-playing dimensions can have multiple relationships, but only one is active; the rest must be inactive.

Why B is wrong: Different cardinalities do not unlock multiple active relationships between the same two tables.

Why C is correct: Correct. Microsoft guidance is explicit that there can only be one active filter propagation path between two model tables.

Why D is wrong: Cross-filter direction does not unlock multiple active paths between the same two tables; only one active relationship is permitted.

Free sampleModel the Datamedium

A visual is already filtered to Product[Color] = "Red". The measure Blue Revenue is defined as CALCULATE(SUM(Sales[Sales Amount]), Product[Color] = "Blue"). What does Blue Revenue return for that visual cell?

  • ABlank, because CALCULATE intersects the visual filter Red with the argument filter Blue, leaving no rows.
  • BSum of Sales[Sales Amount] for Red products, because the existing visual filter is preserved.
  • CSum of Sales[Sales Amount] for Blue products only, because CALCULATE overwrites the existing filter on Product[Color]. Correct
  • DSum of Sales[Sales Amount] across every color, because adding any filter clears the existing filter on the table.
By default CALCULATE overwrites any existing filter on a column targeted by its filter argument; wrap with KEEPFILTERS to preserve the prior filter. When a filter argument targets a column that is already in the filter context, the standard CALCULATE behaviour is to overwrite the existing filter on that column. So the prior Red filter is replaced by the Blue filter argument, yielding the Blue-products sum.

Why A is wrong: By default, when the same column is already in the filter context, the CALCULATE filter argument overwrites it rather than intersecting.

Why B is wrong: CALCULATE only preserves the existing filter on the same column if you wrap the filter argument in KEEPFILTERS.

Why C is correct: Correct. When a filter argument targets a column that is already in the filter context, the standard CALCULATE behaviour is to overwrite the existing filter on that column.

Why D is wrong: Filters on other columns of the same table are preserved; only the Product[Color] filter is overwritten by the CALCULATE filter argument.

Free sampleModel the Datamedium

An author wants the most efficient CALCULATE filter argument to compute Red product sales. The model is an Import model. Which expression matches the documented best practice?

  • ACALCULATE([Sales], FILTER('Product', 'Product'[Color] = "Red")).
  • BCALCULATE([Sales], FILTER(ALL('Product'), 'Product'[Color] = "Red")).
  • CCALCULATE([Sales], CALCULATETABLE('Product', 'Product'[Color] = "Red")).
  • DCALCULATE([Sales], KEEPFILTERS('Product'[Color] = "Red")). Correct
Prefer Boolean filter arguments inside CALCULATE; only fall back to FILTER when the predicate is too complex for a Boolean expression. The documented best practice is to pass a Boolean expression as the CALCULATE filter argument whenever possible, because Import-model columns are stored in optimized column stores. KEEPFILTERS is used here to preserve any existing color filter; the key point is that the filter argument is a single Boolean expression on one column rather than a FILTER over a table.

Why A is wrong: FILTER on the whole Product table iterates row by row and is documented as less efficient than a Boolean filter argument when one is sufficient.

Why B is wrong: Wrapping the table in ALL inside FILTER also overwrites every Product filter, which changes the semantics and remains a row-by-row iterator.

Why C is wrong: Using CALCULATETABLE as a filter argument also produces a table expression, which the guidance explicitly says to avoid when a Boolean filter argument is enough.

Why D is correct: Correct. The documented best practice is to pass a Boolean expression as the CALCULATE filter argument whenever possible, because Import-model columns are stored in optimized column stores.

Free sampleModel the Datamedium

You want a measure that returns year-to-date Sales[Sales Amount] for the current filter context, using the 'Date'[Date] column. Which DAX expression is correct and idiomatic?

  • ADATESYTD(SUM(Sales[Sales Amount]), 'Date'[Date]).
  • BSAMEPERIODLASTYEAR(SUM(Sales[Sales Amount]), 'Date'[Date]).
  • CTOTALYTD(SUM(Sales[Sales Amount]), 'Date'[Date]). Correct
  • DPARALLELPERIOD(SUM(Sales[Sales Amount]), 'Date'[Date], 0, YEAR).
TOTALYTD aggregates an expression year-to-date; DATESYTD just returns the date table for that range. TOTALYTD evaluates the year-to-date value of an expression in the current context. The first argument is the expression to aggregate and the second is the dates column. DATESYTD only returns the date set and is typically used inside CALCULATE.

Why A is wrong: DATESYTD returns a table of dates for the year-to-date; it does not take an expression to aggregate.

Why B is wrong: SAMEPERIODLASTYEAR shifts dates back one year; it does not produce a year-to-date total.

Why C is correct: Correct. TOTALYTD evaluates the year-to-date value of an expression in the current context.

Why D is wrong: PARALLELPERIOD returns a parallel date column shifted by a number of intervals; it is not the YTD aggregator and its first argument is a dates column, not an expression.

Free sampleModel the Datamedium

When is a calculated column evaluated and where does its result reside in the model?

  • AEvaluated on demand at query time, results not stored, no model size impact.
  • BEvaluated in Power Query, results stored as a custom column in the source query.
  • CEvaluated when a visual selects a slicer value, results stored in the visual cache.
  • DEvaluated at data refresh, results stored in the model, increasing model size. Correct
Calculated columns compute at refresh, persist in the model, and add to model size. Calculated columns are computed during data refresh and the values are stored in the model. They evaluate in row context, so each row is computed independently. Because the values persist, calculated columns increase model size.

Why A is wrong: On-demand evaluation with no stored result describes measures, not calculated columns.

Why B is wrong: Power Query custom columns are written in M and computed during query load; calculated columns are written in DAX and computed during model refresh.

Why C is wrong: Visual-level recomputation describes measure behaviour, not calculated columns.

Why D is correct: Correct. Calculated columns are computed during data refresh and the values are stored in the model.

Other domains in this exam

See also the PL-300 cert hub, the study guide, and the cheat sheet.

Examworthy is not affiliated with or endorsed by Microsoft. Original, blueprint-aligned practice material only.