You have to design your data tables in an operational system to support operations, in particular, to not get corrupted, which tends to lean towards
https://en.wikipedia.org/wiki/Database_normalization
but means you have to write queries with joins to get answers and many people find that difficult. Tooling to provide a better view for analytics is an interesting question.
As for statistics I think anybody making decisions should know a little about them. Myself I am a fan of "nonparametric" methods because they only make me learn one probability distribution so if I was stuck on a desert island (with pencil and paper) I could compute my own tables for methods like