Steven Johnson writes:
And then I start writing, fully aware that there are large blind spots of material that I’m going to have to track down while I’m in the middle of the book. This makes it vastly easier to get started, of course, but it has an additional benefit that has become clear to me over the years: when you’re researching in media res, the new ideas or details or stories that you stumble across are much more useful to you, because you can immediately see the slots where they belong.
I’m currently writing a book about sailing and clouds. Several books about those topics sit on the shelf behind me. The further into the book I get the more I am aware of my need for research, but as Steven says, I now know the specific “slots” I need to fill.
As Beth Revis said about outlining, the question is not if you should research, but when.