i made this same point recently using the nba for illustration: https://news.ycombinator.com/item?id=20848444
I numbered each resume without reading it and drew lots. We had 6 people interview ~30 people for 10 slots. Each person had 15 minutes to talk. Candidates were out of the running if 3 people said “no”. We stopped when we had 10 people.
It worked out pretty good, but made HR angry. We ended up with 2 bozos, but the rest were pretty great.
I worked at another place years ago where we would mentor call center agents to become DBAs and Unix people. Those folks were all awesome, and a few had GEDs only. One of them called me to thank me for the opportunity, he just became a CIO!
Personally, if you have the time, you can mentor and get better results from a marginally qualified, engaged candidate than a person with the right credentials.
You would think the tech industry, with its endless graveyard of technically-superior-but-undermarketed or under-supported products would recognize this.
Random somehow seems even more unfair and arbitrary.
It sounds like a great opportunity for an experiment. Say you have 20 slots, pick 10 using the GPA/nitpick method, and 10 at random, and then compare the results 5 or 10 years later.