Microsoft DP-600: Implement and manage semantic models Practice Questions

MAXX iterates over all products in the current category (using ALLEXCEPT to remove the product filter while keeping the category filter) and returns the maximum monthly sales, which is then used as the denominator. This correctly computes the max within the category in the current filter context. RANKX finds rank positions, not the maximum value itself. Using MAX(FactSales[SalesAmount]) would return the max value in a column, not the maximum of a measure across products — it ignores the category boundary correctly. TOPN(1,...) can retrieve the top product but requires additional unpacking to get its measure value and is less efficient than MAXX for this pattern.

PREVIOUSQUARTER returns the set of dates in the quarter immediately preceding the last date in the current filter context, making it the standard function for this use case. DATEADD with QUARTER shifts the current period by one quarter, which is equivalent but PREVIOUSQUARTER is the idiomatic choice — DATEADD is also correct in behavior, but PREVIOUSQUARTER is the most direct answer. PARALLELPERIOD shifts by a complete period and is similar but typically used for same-quarter comparisons across years. DATESINPERIOD counts back one quarter from MIN(Date), which would return dates from the beginning of the selection minus one quarter, not necessarily the prior complete quarter, making it inaccurate for this purpose.

Dual storage mode allows a table to act as either Import or DirectQuery depending on the context of the query. When queried alongside an Import table, it behaves as Import (using the cached copy); when queried alongside a DirectQuery table, it behaves as DirectQuery. This makes it ideal for shared dimension tables in composite models. Import-only mode would cause issues when DimDate is used in a DirectQuery relationship chain. DirectQuery would slow down all queries, even those that could benefit from in-memory evaluation. Direct Lake is a Fabric-specific mode for Delta tables and cannot be mixed arbitrarily in this composite scenario.

Enabling large semantic model storage format allows the model to use the V3 format and store data beyond the default limits efficiently, reducing memory pressure per GB of data through better compression and paging. Removing unnecessary columns and rows in Power Query directly reduces the amount of data loaded into memory, which is the most impactful approach to reducing model size. Replacing calculated columns with measures reduces memory because calculated columns are materialized in-memory while measures are computed on demand — this is also beneficial but not among the top two most direct actions compared to the listed options. Query caching reduces CPU/query load but does not reduce memory footprint. Switching dimensions to DirectQuery would create a composite model and introduce query latency, violating the import-mode performance requirement.

REMOVEFILTERS(DimProduct) removes only filters on the DimProduct table, leaving all other filters (including DimDate[Year]) intact. The measure is mathematically correct: for each product row it divides product sales by total sales for that year, and for the category total row it divides category sales by total sales for that year — which does sum to 100% per year. If the percentages don't sum to 100%, the issue is most likely a visual setting such as 'Show items with no data', a cross-filter from another visual, or a slicer affecting only some products. REMOVEFILTERS does not touch DimDate, so the year filter is preserved. DIVIDE returning BLANK() for zero denominators would cause individual cells to be blank, not cause subtotals to be incorrect. VAR in DAX captures the filter context at the point of evaluation (not declaration) relative to the current row context, so at the category subtotal row CurrentSales correctly holds the category-level value.