Moreover, your theory or story should be grounded in what is already known, other theories or stories, etc. It should not be a matter of your consulting your imagination (sometimes called a "stylized theory"), or if it is such, make sure we understand why your imagined mechanism or variables make sense. If you advocating an abstraction that you believe gets at what is really going on, you've got to convince me about it in two ways: why you believe it is a good account; the empirics support it better than some other account.
If you are doing a theoretical dissertation, be sure your theory comes out of what is already accepted, perhaps rejecting it. Ideas you just thought up usually don't work well.