I find this advice from John Kay’s Obliquity very useful:

When faced with a task that daunts you, a project that you find difficult, begin by doing something.  Choose a small component that seems potentially relevant to the task.  While it seems to make sense to plan everything before you start, mostly you can’t: objectives are not clearly enough defined, the nature of the problem keeps shifting, it is too complex, and you lack sufficient information.  The direct approach is simply impossible.