Free DP-600 practice test: 30 questions with answers

Measures are evaluated in the filter context of each visual, so a SUM measure responds to slicers and filters. Calculated columns store row values, hard-coded numbers do not change, and text concatenation is not a total.

Calculated columns are stored per row in the table and can be used for filtering and slicing. Measures are aggregated at query time, visual filters are display-level, and themes are presentation settings.

TOTALYTD computes a year-to-date total over a date column, using the date context. RELATED fetches a value from a related table, CONCATENATE joins text, and ISBLANK tests for blank values.

Calculation groups apply reusable time or other calculation items to a base measure, which reduces duplication. Printing, separate models, and removing the date table do not reuse calculations.

A dimension row relates to many fact rows, which is one-to-many from the dimension to the fact table. Many-to-many, one-to-one, and unrelated describe different structures.

Row-level security roles apply DAX filters that restrict rows for each user or group. Slicers can be cleared, passwords do not filter rows, and hiding pages does not restrict the underlying data.

Hiding a table keeps it available for calculations while removing it from the field list in reports. Deleting breaks calculations, random renaming confuses users, and text conversion does not hide anything.

Calculated columns are computed during refresh and stored, while measures are evaluated at query time, so moving logic can reduce refresh cost when appropriate. Adding columns, disabling refresh, and images do not solve the cost.

DirectLake reads Delta tables in OneLake directly for near-real-time results without an import copy. Import mode copies data into the model, and printed or manual exports are not live.

Incremental refresh loads only new or changed partitions, which cuts refresh time for large tables. Full history reloads, removing dates, and text-only keys do not address the performance need.

Relationships require matching data types and unique dimension keys to work correctly. Images, deleting tables, and comments do not produce a valid relationship.

A left outer join keeps all sales, and an explicit unknown member handles missing products transparently. Inner joins drop sales silently, deleting without records loses data, and duplication inflates totals.

Using the view as the source means refreshes read current table data through the view. A one-time copy, text file, or manual paste will not reflect future changes.

A Dataflow Gen2 transforms data once and writes the output to a table that many models can use. Printed scripts, bookmarks, and visual themes do not provide reusable prepared data.

Partitioning by a frequently filtered column enables pruning, and compaction keeps file counts healthy. Text dates, removing the column, and duplicating per date reduce performance or increase maintenance.

Stable unique surrogate keys give reliable relationships between dimensions and facts. Names with duplicates, changing random numbers, and missing keys break the model.

Deployment pipelines promote items through stages with controlled deployment and optional approvals. Manual copies, spreadsheets, and printed notes are error-prone and do not enforce stage control.

Git integration versions item definitions and supports pull-request review. Screenshots, printed definitions, and chat histories are not version control.

Failure notifications alert owners as soon as a refresh fails. Monthly checks, user reports, and turning off refresh do not provide timely detection.

Build permission and workspace roles scoped to the modelling group limit who can build on or edit models. Broad edit rights, shared passwords, and public datasets violate least privilege.

The lineage view shows dependencies from sources through transformations to reports, which supports impact analysis. Printer queues, calendar planners, and theme editors do not show data dependencies.

Restricting export through settings or policies limits data leaving the platform. Personal email, public links, and disabled auditing increase exposure.

Reviewing capacity metrics and rescheduling heavy work reduces contention without losing functionality. Turning off reports, deleting models, and ignoring the problem do not address the cause.

Natural-language features such as Q&A or Copilot let users explore semantic models conversationally. A printed dictionary, static image, and emailed request are not interactive exploration.

SQL endpoints allow ad hoc SQL exploration directly over lakehouse and warehouse data. Visuals without sources, paper, and images cannot run queries.

Key influencers and decomposition tree visuals help identify and break down drivers of a metric. A total in a text box, a printed map, and a table without calculations do not reveal drivers.

Hierarchies with drill-down allow users to move from year to month to day in the same visual. Images, separate reports per day, and removing dates prevent that exploration.

Fewer visuals load faster, and complex measures over large tables take longer to compute. More visuals add work, the data model strongly affects performance, and high-cardinality columns in visuals are often expensive.

A change measure comparing the two quarters, ranked by product, answers the question directly. Counting products, alphabetical lists, and images do not identify the largest increase.

A matrix shows values across two categorical dimensions in rows and columns, which is compact and clear. Pie charts with many slices, single cards, and gauges do not show two dimensions well.