I agree that better data reporting, with focuses on goals and outcomes, much as has been done in the Aid sector, could help. I also think that having a better education environment, in the form of good advising for support, and broader class type / learning options, may help students complete school once they're there.
However, in a tie-back to the earlier article [1] on the disconnect between boss / worker expectations, I think we also need a much earlier focus on individual achievement and learning. On setting goals, and then largely giving students freedom in how they approach and attack those goals. That is what college students need to be successful, and that is what the job sector identifies as the signal of good, high performing workers.
Many of the students that we send to college are not mentally ready to be successful at college. In fact, many go without any clear idea of why they're going, besides the advertised 4+ year party, or family saying "go". In many ways, I don't think that's their fault. Part of it is, for not taking charge of their lives, but for many, college is one of the first times where you seriously have to set your own goals. High school teachers took care of it, parents took care of it. We need to remedy that by providing more, low-risk opportunities for them to develop those skills while they still can work with a net.
As a corollary, I also don't think we should make it easier for students who aren't ready to attend. It does a disservice to them, and due to the way our current college loan system is set up, it often ends up burying them for the rest of their lives under non-transferable debt. (often times at the scale of a low-end house) For those who it didn't, many times we get C to D folks out the other end, who muddled by, and drift on to something. Maybe the act of attending enriches them, and ultimately makes them better people, but for a lot, I feel like higher education has become an expensive trap that we're not properly preparing them for.
[1] http://www.insead.edu/facultyresearch/research/doc.cfm?did=4...