You measure what is easy to measure, then try to map that to forward returns. That’s why VCs have portfolios.

If it’s hard and/or expensive to measure, it’s likely better to just grab another startup instead.