I'm not sure what kind of data you would use to drive it. A lot of the most important aspects of education aren't readily measurable: the ability to write, the network connections you make (to fellow students, professors, alumni, donors, etc), capacity to work in teams, and so on. You risk the same things you get with tech firms using leetcode and puzzles to select... well, to select those who are good at leetcode and puzzles.

You can measure it in terms of career success, but even there it's very hard to deal with the selection bias. Those who succeed are those who succeed.

I'm a big fan of Signum University, which is an online-only university dedicated to the soft skills that have to be taught personally, rather than a MOOC. It can't develop connections like Harvard and Stanford but it can develop those skills that make the hard-to-measure differences between those who succeed and those who are merely very good at taking tests. It remains, however, impossible to factor out those connections, which seem to make the biggest difference.