In short: Work-based learning holds people when the mentoring layer is built on purpose, and whether it's working can be checked while it runs.
Apprenticeships, internships, and work-based learning succeed on the details. These pages cover what registered apprenticeship requires, the mentoring layer, designing placements that stick, and knowing whether the program works.
A grant evaluation runs across the whole award, from the first question asked to the funder's annual report, and the plan can change when the program does.
Some parts of evaluation a small team can genuinely do alone and some parts shouldn't be done alone, and knowing which is which protects both the budget and the credibility of the result.
Measuring the same things across sites is possible without erasing what makes each one different, and the design question is which few things everyone reports and what stays local.
Students choose and leave career paths for reasons that are knowable, and the choice between recruiting more and keeping the ones you've got is answerable with what a program already tracks.