Decades ago, as my father explained to me, ARPA (no "D" at that time) was happy if 1% of their projects went all the way through to successful deployment. If they had a higher success rate it would mean they weren't aiming high enough.