Several other ideas in this list question the efficacy of traditional grades, but how are we to be sure one way or another?
The obvious answer is to control with randomized longitudinal studies that follow select students through their entire educational career. These studies let you make meaningful (though qualitative) comparisons across grades and institutions and are a source of ideas numerical approaches aren't designed to expose.
Instead of having longitudinal studies only as part of occasional, special-purpose investigations, build it into the core practice of operation an education institution. Building in on-going (but sparse) validation can keep grades relevant.