Hi, most time dimensions are setup using a base Date field in the fact table, and they have plenty of issues for time analysis as it is. However my fact sales records have the time aspect assigned by pre-calculated periods, because depending on various
factors, monthly final invoices are all raised on varying days (usually 2nd friday of month but can change). The monthly period is therefore not a straight calendar month. Probably a very common scenario.
So, the invoicing system already assigns the year (ie 2009, 2010, 2011) and monthly period (1, 2, 3 ... 12 with 3 representing march, even though that might represent 13th march to 9th april) and I want to use those as Time dim so we can do YTD, growth-on-prev-year
It looks like its best to setup 2 Dimensions to link to 2 DataColumns/Attributes in the Fact table (say, FYear and FMonth, both integers). That way I can assigned attribute names like March to key column 3. If I combined them into 1 dimension with both fields
making up a single key column, would have to either repeat the month names or link it to another Star schema I believe.
I can use the Add Business Intelligence wizard to make the Dimensions into Time ones instead of regular but I'm still not totally sure if this is the best structure/method and once done, how to use the YTD calcs to show in the cube browser (and my MDX knowle
View Complete Post