Later life success over median replacement value seems like a nice bonus structure for retirement.

E.g. statistically normalize cohort (income, etc.), measure their median outcomes, and then see how the teacher's cohort does

It'll be dirty, given the small sample size and number of variables, but it'd be a nice bonus +/- to retirement.

I'd also hazard prior student reviews, sampled when they turn 30 & 40 would be very interesting.