Calendar-based time intelligence in Power BI and Fabric Tabular models (available since September 2025) introduces subtle pitfalls when working with custom calendars. Four key issues are explored: (1) DATEADD() behaves differently across months of unequal length, causing shifted PY values between leap and non-leap years due to the 'Distance from Parent' mechanism; (2) a workaround using -12 MONTH granularity instead of -1 YEAR resolves the leap year shift; (3) weekly calendar calculations can produce incorrect results in certain data model setups with no clear root cause, requiring a full model rebuild; (4) mixing weekly and monthly logic in the same date table is blocked by the engine's consistent categorization requirement. Weekly calendars work correctly for PY comparisons because all weeks are equal length. The feature is still in Preview, and multiple date tables may be needed to handle complex calendar scenarios.
Table of contents
IntroductionSetup of the CalendarsPrevious Months and different month lengthsWhat happens with the previous year?Weekly calculations – Head scratchingMixing weekly with monthly logicWeekly calculations – That’s interesting!ConclusionReferencesSort: