By my last year at Matasano we were literally discarding the results of phone screens. We did them largely as a pro-forma exercise, and because they made team members more comfortable with the process as a whole.

We virtually never selected out candidates based on phone screens. We had a work sample process that kicked in after phone screens, and that cost us almost nothing to run. So we never had an incentive to prevent a candidate from going through that process.

Unlike phone screens, the work sample results were strongly predictive. You could bomb phone screens, ace work samples, and end up coming in for in-person interview.

(We did "nudge" candidates we felt wouldn't do well on the work sample tests, solely out of concern for not wasting their time, but anyone who wanted to proceed was able to).

Phone screens are a waste of time.

Spolsky did a really good job of documented the best practices within the framework of unstructured interviews as they were practiced 10 years ago. The problem isn't Spolsky's tactical suggestions; it's that the strategy they're a part of is being discredited.