Today I found myself needing to load a .csv file into powerpivot.
This .csv file has around 11M rows in it, 45 columns, fully denormalised, and is around 3,4G.
So, to test and validate Microsoft’s claims that the Vertipaq engine does a decent job at compression, I loaded the .csv file as-is directly into a PowerPivot model.
And yes, Microsoft were right. The same data in PowerPivot compressed to a 140M .xlsx file!!
That’s a compression ratio of 4.1%!!