Data Model: Dimensions to fact table columns

As is well established, best practice is to have a Fact table with columns that are your 'facts' for each row. What is the suggested approach to when those columns may have dimensions that are unique to them? For example certain groups of columns may always 'go together' and in visualizations it may be desirable to filter or organize the data based on these dimensions.

• should the single large fact table be split into multiple fact tables based on these dimensional relationships?
• I can imagine a scenario where you pivot the facts, so that you end up with a large table of Key:Value pairs, and then you could have additional columns to define the relationship(s) to the dimensions.
  • This however would be very in-efficient, and could cause issues in 'large' datasets.
• Is there another approach I'm missing?