I can't imagine the difficulty to accurately measure the success or failure of long-tail HR hiring processes like phone screens. The success or failure of a candidate post-hire has so many variables it must be very hard to attribute them to signals present in a screen. I imagine most of the data points are derived from signals found in successful candidates, and then trying to find them in an assessment or screen.
Its really hard, and I hope the negative tone of my comment does not suggest I don't respect the problemset and the people willing to throw themselves at it.