ERP. Forget about being fully SaaS but be SaaS enough so that what you build to scale in the cloud can also scale out of the cloud. Edit: what I mean is that most companies interested in SaaS are actually more confident to host their own little version of your solution for their needs.
Spend time studying OpenERP because it's succesful and you'll have to compete against it. Their platform is a great case study: it's powerful and modular while being difficult to take over. Their code base is a mess and yet they have developer's traction. There's a huge market, no matter what you plan (a platform or a solution).
Good point about the distinctions between Saas vs ERP. Customers don't want to buy software or services; they want to buy a solution to their problem (and preferably a solution with a positive ROI).
Spend time studying OpenERP because it's succesful and you'll have to compete against it. Their platform is a great case study: it's powerful and modular while being difficult to take over. Their code base is a mess and yet they have developer's traction. There's a huge market, no matter what you plan (a platform or a solution).
If you go for a platform, avoid small projects.