Sample question 1 — Model the data
A utility company loads five years of meter readings into a Power BI semantic model, and the dataset grows by about two million rows each month. Full refreshes now exceed the gateway timeout. The source is an Azure SQL database, and the team wants only recent partitions to refresh while historical partitions stay untouched. Which requirement must the team satisfy for incremental refresh to work as intended?
- A. The date column used for partitioning must be marked as a date table in the model
- B. All transformations applied before the RangeStart and RangeEnd filters must fold
- C. The semantic model must use DirectQuery mode for every table in the dataset
- D. The gateway must be installed on the same machine that hosts the Azure SQL database
Answer: B. Option B is correct because incremental refresh relies on query folding to translate the RangeStart and RangeEnd filters into source queries, so any transformation that breaks folding prevents partition pruning and forces full reads. Option C is incorrect because incremental refresh works with import mode, and forcing DirectQuery on every table changes the architecture without enabling partitioning. Option D is incorrect because the gateway only needs network access to the source; co-locating it with the database is not a prerequisite. Option A is incorrect because marking a date table supports time intelligence, not the partition filtering that incremental refresh performs.
