The only way you can truly be sure of your hiring process is to pick a random sample of people who fail, and hire them anyway.
I see a couple of issues with this:
- Could it get you in legal trouble?
- Are you capable of evaluating who are the best employees AFTER they are hired?
- Only a large corporation would have the resources to gamble on hiring a random sample of people who failed their interview process.
Here's one half-hearted way to do it. Pick a random sample of resumes you reject, and give them phone interviews anyway. Pick a random sample of people who fail your phone interview, and give them an on-site interview anyway.
However, how can we conclude anything from a procedure that only examines the hired population and none of the unhired?