The thing that is missing from most of these government initiatives (in the US at least) is a mechanism and willingness to revisit and iterate. That's how we do software these days, but social programs are developed and shipped in a full waterfall cycle, then politicians just spend decades arguing about whether it should be cancelled due to some flaws or kept exactly as-is. What's missing are real data-driven attempts to close the feedback loop and fix the issues to achieve the stated objective.

Measuring success is good, but we also need to exercise the ability to make small updates and continue measuring, instead of just using that data to argue on the side of "throw it all out!" or "we can't touch it, it's essential!"

I think giving all schoolchildren free food is self-evidently a good thing and "verifiable measure of success" requirements are just a way to manufacture BS justifications for taking food away from children from famously unreliable policy outcome studies.

These programs are pitched along with statistics, so maybe it’s on us to demand the methods of those statistics continue with the program.