Interview was straightforward except for one sticking point. A friend who interviewed for a similar position in Australia reported precisely the same experience and the same irritation. In the interview, a scenario was pitched in which an arbitrary question requires a numerical answer. You will have to "guess" a whole bunch of input data and then predict an answer. If your first instinct is to build a simple mathematical model to reduce the number of arbitrary guess-numbers you need to use you will be "wrong". This will not be explained to you but you may be told afterwards that you need to guess the entirety, i.e. simply start throwing random numbers down on a piece of paper and work out a single-use answer. The question will usually be something where you could get the answer off the internet in 5 seconds. It's a useless scenario and entirely contrary to the way an experienced analyst should work but there you have it. Welcome to IBM.