Initial online test: 1) With six individual profiles of different people and six shopping lists, we were asked to match characters to lists. 2) 10 online transactions with limited data were given, and we were asked to predict whether the transaction is fraudulent or real. 3) The longest part, we were asked across 5,000 transactions, fraudulent and real, to give input on the data. It's important to do pivot tables (although excel algorithms could also do the trick) and remember\know statistical intuitions - the law of large numbers, average vs. median, the importance of std. deviations, etc.
