That's a question that gets asked a lot on the mailing list and in the irc channel (#concatenative on Freenode). Here, I'll try to give my answer to that question.
My view is that languages and computer things in general should be judged on what they are now, not what they can become in the future. Things like Bitcoin, the Julia programming language, LLVM gets a silly amount of attention due to the hyped promises on that they will someday be awesome.
Let's take ClojureScript as an example. The promise is that you will someday be able to write your websites in Clojure and use the cljs compiler to compile to Javascript. Well, last I checked, the javascript files it generated was hundreds of kilobytes large and the compiler took something like 30 seconds to run.
The developers are probably working on fixing that (or they might even already have, in the latest git version), but currently it is not useful tool for writing web sites.
If we apply the same critical mindset when judging Factor's usefulness, what do we get? Well here are some things Factor can do for you today:
Easy to contribute to In contrast to, say, Python where you will have a hellish time trying to get your package accepted into the standard library. To be fair, CPython is also (currently) a much more polished software than Factor is which is one of the reasons getting your contributions accepted in the latter is easier. :)
But if you write Rosetta code examples, library bindings (why not write a binding for libxml2? it would be very useful) or something else, it's almost certain your contributions would be merged.
Writing simple games Factor has bindings for opengl and directx so you could easily write some simple games. SDL is not wrapped which is a flaw. It would be really nice to be able to use that library too.
Note also that the deployment mechanism means you can bake all your code into a simple executable file which is easy for distribution.
And some points Factor is weak at:
Web programming Furnace is nice, but it's no Rails killer. It is severly lacking in documentation which I think makes it kind of impractical.
Hard to learn Both because the language is so different from everything else, but also due to the lack of newbie tutorials. I think it's roughly as hard as Haskell.
Non-paycheck-helping Languages like VB.Net, SQL and PHP are good to learn to get employed. Factor, Clojure and Erlang are mostly useless for that purpouse.
Not a good first language Someone who doesn't know programming would be much better of starting with Python and not Factor.
Uncertain future At the moment, there is only I and two other developers regularily pushing stuff to Factor. If we lost interest, there would be no one to take over. Slava Pestov, who is the languages inventor, isn't active enough anymore.