Like others are reporting, the 1st round was similar CSVs with data around flights/airports/tickets/passengers. Questions were almost too basic, pure statistics, overview, light data manipulation (used pandas in a jupyter notebook). The most difficult question involved identifying that a discounted executive class ticket could (in some cases) cost less than an economy ticket... The goal seemed to be to identify a particularity in the dataset with business impact, in this case the inconsistency in ticket categories with impact in branding/reputation - without really needing to isolate other variables that could make the overlap a natural consequence like flight distance etc. I passed to next stage. A colleague of mine was interviewed a few days before with a slight variant dataset, but the last question was WAY more difficult, requiring a complicated self-join etc etc. He failed.