Good point. Sometimes people do. I generally try to stay away from the data format because I want to focus on the main algorithm and 45 minutes isn't a lot of time. I don't want to test people's ability to read and parse a file, I'm more interested in learning how they think. So I generally say something along the lines of imagine you're given some sort of file iterator that reads a row at a time, and it returns an object that has timestamp, customerid, pageid, to abstract that part of the problem.