Education largely does not work for most people who enrol. Most college students do not finish, and something like a third of university students do not either. And even among those students who receive the credential, many are probably able to slide through without actually learning much.
Anecdotally I think you'll find education has a pareto distribution: the top few percent of students learn enormously, the top 20% learn a decent amount, the next 20% or so squeek through with marginal benefit, and the bottom half does not even finish.
The reason I get annoyed is that I think these new, data-driven forms of education shoot themselves in the foot by gathering data. If they just shut their ears and pretended that everyone was learning based on the grades/assessments like universities and colleges do, they might be just as successful.
In other words, the proper comparison is not whether the education is most likely to succeed, but how likely it is to succeed in comparison to other forms. Sort of like those quit smoking aids which reportedly double your success rate of quitting--in a notoriously difficult task like quitting smoking, going from a 3% success to 6% is profound.