Question

An analytics engineer manages a 12 GB Import-mode semantic model in Fabric. Query performance is slow and the model consumes excessive memory. Using VertiPaq Analyzer, they discover that a single DateTimeStamp column in the fact table (holding transaction timestamps to the second) accounts for over 40% of the model's total size. The engineer needs to reduce model size while preserving the ability to analyze trends by day. What is the most effective change?

Accepted Answer

Split the DateTimeStamp into a lower-cardinality Date column and, if time is not analytically required, remove the time component

Answer

Split the DateTimeStamp into a Date column and a separate Time column, keeping both at full precision

Answer

Switch the fact table to DirectQuery storage mode to avoid storing the column in memory

Answer

Enable the large semantic model storage format on the workspace so the column compresses better