Power BI and Fabric Tabular models received Calendar-based Time Intelligence in September 2025, offering new possibilities for date-based analysis. However, the author discovered several unexpected behaviors that practitioners should understand before implementing these calendars in production environments.

The most significant issue appears when comparing previous year values across leap years. When comparing 2024 (a leap year) to 2025, the PY measure shifts dates by one day due to DAX's "Distance from Parent" calculation, which uses the YEAR as the parent in the calendar hierarchy. This causes incorrect daily values while maintaining accurate monthly totals—a subtle but dangerous inconsistency. The author recommends using DATEADD with -12 MONTH instead of -1 YEAR as a workaround.

Weekly calculations present additional complications. When the hierarchy is collapsed, all PY values appear identical; they only display correctly when expanding down to the date level. One solution involved rebuilding the data model entirely, though the underlying cause remains unclear. The author advises using separate date tables for weekly versus monthly calendars to avoid interference between calendar types.