in my experience a/b testing shazam just has a way, way bigger database. It would easily identify really obscure stuff like 1970s Arabic psych rock or disco even if it wasn't even available in US online stores, and wasn't ever sold here. Same goes for gracenote, and the other systems that tried the same shtick
the database is the golden goose here, not the algorithm. it's similar to what made grooveshark more fun than spotify