In the context of startups, my understanding is that it's not enough to identify a small initial market (e.g. Julius Caesar impersonators), ideally it should have some reasonable potential to be huge in the future.