Michael proposes concentrating on business desktop instead of consumer one, but I think some smaller desktop niches could be worth going after too.
For example I've understood Linux to have fairly good support for pro audio production (assuming one is willing to invest a bit of effort). That kind of content creation "niches" could be good target; sound effects, music, videos, animations, radio, leaflets, document processing etc.
Linux "marketing folks" could then construct success stories out of successful migrations from costly & proprietary software workflows to free ones, what to expect i.e. what are the pros & cons compared to proprietary offerings etc.