My answer was - First, I look at the expected solution that I am supposed to achieve and understand it thoroughly. Second, I find out the past approaches that were tried to solve the problem and understand why they failed. Third, I break the problem into sub-problems so that it becomes manageable to tackle. Fourth, I try to understand if there exists an approach that has previously seen to have worked on a similar if not the same problem (Literature Review). And finally, I try to design my solutions based on the data I have at hand and the understanding of the domain/problem gained from the literature review.