I was talking the other day with some RPG design nerd friends (let's be clear, they're nerds about a lot of things, this was just the overlap we were talking about) about items and what makes items in RPGs fun. We've all gotten to the "because they let you do things you couldn't otherwise do" part, even if we're sure there's something else we're missing.
I pointed out that in the real world, technology (usually) gets invented to do a thing we already do better in someway, and then we figure out the new things it allows us to do. W said I was looking at broad technology, like computers, where they were looking at Joe PC's L33t MaGIc Haxxzor Rig. which got us to looking at the forest to figure out how to implement the trees, "Needs moar tree.", and the design failures of writing the forest, i.e. interchangeable and uninteresting items.
All of which has me thinking about the approach to RPG design issues. Is it better to start from the tree level: what do you want X to do in your game? Or is it better to start from the forest level: what makes X fun? How does that integrate and impact the other aspects of the system? How does X actually work in the real world and how are we importing it into the game?
... As you might be able to tell from having more forest questions, when approaching an abstract question like this, I default to 'forest' mode.
That said, my personal opinion is it depends. I know, real useful that. But I don't think this is something you can look at in isolation. A systemic, big picture approach is probably going to work better for a more narrative heavy system, one with more abstraction up at the systemic level. A deep dive into individual components, a more 'tree' approach if you will, is going to work better in systems where you want the difference between different items of the same type matter to game play.
Also, in my ideal RPG design scenario, you have multiple perspectives. Even if you're the sole designer for a product, being able to bounce thoughts and ideas off of someone who approaches things from a different perspective (writing groups are great guys) is going to get you a stronger product.
That's the whole idea behind play testing, isn't it? Hand off your project to someone who only knows what's on the page (instead of what's in your head) and see if it works.