About Sparsity

Harry Sun

Hi Team,

I am reading the Best Practice Screen, which suggests that version / scenario can be set as Dense to improve speed.

https://community.board.com/discussion/16440/sparsity

Just for the learning purpose (not means I will change it), we should review the folloing Sparsity? Thanks

Edward

This is my understanding so please someone correct me if I am wrong but,

It’s not always black and white and when in doubt it's better to lean towards sparse over dense.

Consistency across cubes is more important. A dimension marked as dense in one cube should ideally be dense in all cubes to avoid unnecessary increase in number of sparse structures (visible in the cube analysis section). Especially important to avoid different sparse structures in dataflow & calculation steps.

I would use a dense when it is a small dimension with few members that are frequently populated.

• Example: Version – if typically all versions (e.g., Plan, Forecast, Actual) are populated.

Another way I find helpful to think about it is if users typically filter by a dimension on screen (e.g., selecting one version or scenario at a time), and then running a calculation, it may be a good candidate for dense.

If it helps in your example I would also consider Business Unit as a dense dimension (but hard to judge without seeing the underlying data and in general more context).