I don't understand what it is about, what it excels, what the real uses are, and why there's someone giving time to make this scheme thing.
Same thing could be said about any language. Tell me again why the world really needed Java? (hint: it didn't invent bytecode). Anyway, scheme's got a well-documented history, did start with a pretty specific purpose (implementing the Actor model), found its way to other purposes (teaching general computing), and by virtue of having been around longer than most people around here have been alive, isn't really concerned with judgements on its merits made by you or me. That said, there's probably more scheme implementations than useful scheme apps in the wild.
Specifically what it's good at is being a lisp implementation that at least until recently was pretty simple (R4RS was just 50 pages, the common lisp hyperspec is over 1000 pages). Even in that small spec, it demands things lisp doesn't, like tailcall elimination and continuations. Some point to that as holding scheme back, but most schemers would claim that they're essential to allow any scheme implementation to express things the essential "scheme way". So go read or watch SICP and realize you're probably not going to be writing a general fixpoint function in java or common lisp anytime soon. Of course you
can write one in Haskell, I dare say even more elegantly, but I'd still say scheme is a tad more approachable.