t's most likely a generic scheme (otherwise they'd promote the fact
that it covers different kinds of music) that resembles the processing
used for pop music on radio: a fast compression that maximizes
loudness and reduces dynamic range. It's like photos — you can make
most any snapshot "pop" as though removing a veil of smoke from the
image with a couple of steps.
It's an interesting service model for the Web, I wonder how it will
develop. Of course it won't instantly make a really crap recording
marketable, and if they ever pitch it as such they'll be frauds. Even
us performer-composer-recordist-engineers can benefit from a savvy
mastering engineer's efforts, but it's doubtful a simple algorithm can
do anywhere near as good a job as a skilled human can for specific
recordings.