[mulmen]

Those common definitions are priceless. On top of not repeating yourself you also avoid questions like “why does carlineng’s report say it is week 43 but mulmen’s report shows week 44? This is even worse if we both calculated revenue but I did SUM(payment) and you used SUM(price*qty).

I do think there’s a potential for performance benefits too because you can sort the fact and dimensions in ways that make sense for each one. Calendar is sorted by week number, product is sorted by a category. Fact is sorted by a date and locale, etc. since you don’t materialize the full projection more expensive but flexible sorting like interleaving becomes possible.

It’s probably academic but it’s not as hard as people make it sound and I think there are real benefits over these wide tables with simple sorting.

Plus if you use a tool like Looker it will write the reporting SQL for you! Generating star schema queries is easy once the tables are in place.