Large Dimensions in SSAS Tabular #ssas #vertipaq

After many years of helping several companies around the world creating small and large data models using SQL Server Analysis Services Tabular, I've seen a common performance issue that is underestimated at design time. The VertiPaq engine in SSAS Tabular is amazingly fast, you can have billion of rows in a table and query performance are incredible. However, in certain conditions queries made over tables with a few million rows are very slow. Why that?