I think the ideal solution is to spend a lot of time in the analysis phase to load your brain up with the correct context, but then be ready to throw out the overengineered solution and just build what feels right.

Don't fall prey to sunk cost fallacy. Just because you spent hours researching a PhD level topic doesn't mean you now have to use it in your project, if it's not quite the right application.