The worst thing you can do is unstructured interviews. Unfortunately, that's exactly what almost every company does: they create ad-hoc committees of workers and have them each design an interview, often on the fly, for each candidate.
Everything you do to add structure and remove degrees of freedom from interviewers will improve outcomes. Every company that hires more than one tech worker in a year should be working on this problem, and hopefully documenting their result.
Our structured interview process differed from Google in that we (a) completely standardized interviews and (b) designed the questions in those interviews to minimize the need for free-form follow-up questions ("Tell me about a time..." and then digging into the answer to mine conclusions). We tried to flip the script: instead of having the interviewer ask followups of the candidate, we put candidates in a position to ask lots of questions of the interviewer, within a framework that generated lists of "facts" that we could record and compare to other candidates.
Example: "I am going to describe a trading system in broad strokes. I know lots of things about it works, and I want you to ask me enough questions so that you can roughly diagram it on a whiteboard, capturing its components." Then, later, "I want to have a conversation where we rank these components in order of sensitivity to latency". The interviewer captures the answers, but the back-and-forth is dominated and mostly directed by the candidate.
The problem I see with the Google strategy is that it combines interviewer-directed free-form questions with an open-ended and somewhat fuzzy question that really evaluates less the candidate's ability to do the work and more their presence of mind during the hostile, stressful interview process.
Things I positively responded to in this article:
* The notion that we come to snap judgements about candidates within minutes of the interview and are then hostage to confirmation bias for the rest of the interview
* The effectiveness of combinations of evaluation techniques; for us, a combination of work-sample testing and scripted interviews were extremely effective.