when to use a seperate month dimension?

Hi,

looking at various Kimball examples I notice that he often uses a seperate month dimension when there is also a day dimension that contains a month_key.

Could someone explain the value of using a seperate month dimension and if there are any specific situations in which it would be preferrable (or not) to use one?

Thanks,

Karl

The primary purpose is for clarity in the model. There is no ambiguity as to the grain of the measures if the fact table uses a month dimension. It doesn't affect conformance if the attribute values in the month dimension are identical to corresponding attributes in the date dimension.

Personally, I prefer to use a single date dimension table with additional natural keys for month, quarter and year. These NK columns would usually be null except for designated rows used to represent a particular period (such as the last day of each month to represent month values). I would then use role names in the fact table to clarify the time grain (such as month_date_key). For presentation purposes, I may sometimes create views (month view, quarter view, etc...) if it helps clarify things.