This actually brings up a third alternative... start with the industries, as you suggest, but then, as part of your hypothesis-resolution (see http://swombat.co

This actually brings up a third alternative... start with the industries, as you suggest, but then, as part of your hypothesis-resolution (see http://swombat.com/2011/1/7/how-to-evaluate-and-implement-startup-ideas-using-hypothesis-driven-development http://swombat.com/2011/1/7/how-to-evaluate-and-implement-st...), try to create delivery channels into those industries that you like. Then follow the one where you best succeed into building delivery channels. Best of both worlds?