advice on data architecture

Hello Community  -   Been watching all of the various videos on the import modes, composite models, DQ, etc etc.   We are in the process of building out our corporate data structure and I have some questions for the more experienced folks out there.  

Sort of a two part question: 

Our source data is generally ERP or CRM and we have dataflows connected to both.   We have various departments like sales, engineering, manufacturing, etc.   

 1.   Does it make sense to have datamodels/datasets at the fact table level?  For example, a dataset specifically built around the sales table, another one built around the orders table, etc.     In that example, they would likely contain the same dimensional tables (customers, products, date table, etc), only the fact tables would differ.    And if so, if creating reports that needed to use both, what is the best way of bringing them together?   Keeping in mind, we are asked to add modifications that require new measures/columns fairly frequently so connecting purely via a live connection does not really work for that purpose.  

2.  Our business does not require "live" to the minute reporting.   A table being refreshed twice a day for us is perfectly fine.   And we have less than 10 million rows of data.    Would a composite model make sense as we do (or will) likely have a need to augment our ERP and CRM data with other sources.  

3.   Does it make sense to have "organizationa'" data sets?  In other words, datasets that are specific to certain departments  (and possibly these departments would need to use one of the data sets that another department is also using).   For example, our Quality department has reports that require the sales and orders tables.  

Any guidance is appreciated!