Obviously, there's no guarantee that a former inmate won't commit another crime even if the rehabilitative measures he experienced while doing time were exemplary. However, there's also no guarantee that in the university example a graduate will find a well-paid job. Still, there are universities that have successfully implemented this business model.
There's a whole industry that's predicated on this kind of uncertainties: The insurance industry. While there's no guaranteed outcome risks and probabilities can be calculated and factored into the price of a service.
The question perhaps is: What value do we as a society allocate to a former convict likely not committing crimes in the future anymore but contributing to society instead?
Considering the alternative, i.e. a former convict not only being a liability to society but also being highly likely to commit crimes in the future, I'd that service should come with quite a high price tag.
What is hard is not just finding good metrics (that cannot be gamed), it is finding good metrict (that cannot be gamed) that can also provide profits to good players.