Perhaps always invert but after approaching the problem from the front.
I would apply the 80/20 rule from both directions (so in theory perhaps spending up to 40% effort) to get the best chance of success. And really, you can't invert without first knowing enough about the problem you're trying to solve.
This follows the way my brain works too. Dog into the problem -> invert -> learn more about problem -> uninvert -> map problem in much greater detail -> reinvert -> discover truth -> solution
I would apply the 80/20 rule from both directions (so in theory perhaps spending up to 40% effort) to get the best chance of success. And really, you can't invert without first knowing enough about the problem you're trying to solve.