look for an unoccupied niche that has potential (this and some other good tips can be found in the paper Picking a research problem — the critical decision which is primarily addressed to the researchers in biology and medicine but can be of interest to the other scientists too)
try to move beyond the subject of your Ph.D. thesis (if you have already defended one, indeed) or your postodoc (or your postdoctoral mentor, for that matter); more broadly, beyond your current area of research (see e.g. this post of Terence Tao). This has an extra benefit of reducing the risk of being scooped as discussed here.
do something you will enjoy doing and what you feel you can do
your work should rather open the way to new breakthroughs than close the whole subject down
The last three tips are somewhat of a common wisdom and can be found in a number of places; see e.g. the article Choosing a research topic by Richard Reis, which contains some further interesting thoughts on the subject.