Reading through the comments there's a lot of defending 37signals. I think the poster is absolutely right though.
There was an interview with Marissa Meyer on The Charlie Rose show where they talked about recruiting talent for Google and Marissa was asked what her experience was. Since it's Google they had obviously tried to do all sorts of predictions on how well a new employee would turn out based on cover letters, interviews, resumes, etc. What she said was interesting. The only predictor of how well a new employee would do his job was his resume. Nothing else gave a statistically significant prediction.
So Marissa and I think the poster is right and Fried is wrong :-)
Isn't there a bit of self selection and typical Google pseudo-science going on with this approach? You can't get in the door of Google without a particular baseline standard resume. This is Fried's entire point, that he would actually consider a high school dropout whereas Google will not. It's a disagreement as to the relevance of a resume which is, for all intents and purposes, an abstraction of a person. Fried argues to skip over the abstraction and go straight to judging the person.
There is an entire group of people that the likes of Google simply will not look at and so it makes their results a little less substantive. Here are folks that would never get hired at Google in their 24 year old form:
-John Carmack
-Bill Gates
-Steve Jobs
-Larry Ellison
-Paul Allen
-Michael Dell
If Marissa Meyer truly believes she's found a statistically significant predictor, then I urge her to hire people without doing any interviews. She won't, because she knows it's more pseudo-science from Google. I wish more people would call them out on this kind of logic. They have a particular knack for marketing local maxima as the global kind.
Thanks. I took the liberty of transcribing the latter part of that section.
"We basically found that their background and references are the best predictor. You can't use them exclusively, but it's true: The best predictor of future performance is past performance. And that's what we really found out through the regression models. We also found that there were a few interviewers in the company who were very, very good. They were several standard deviations off, meaning that they could tell in an interview where or not someone was going to be good or not. In some cases, they would be aberrant and reach a different conclusion from the other interviewers, but they would be correct."
There was an interview with Marissa Meyer on The Charlie Rose show where they talked about recruiting talent for Google and Marissa was asked what her experience was. Since it's Google they had obviously tried to do all sorts of predictions on how well a new employee would turn out based on cover letters, interviews, resumes, etc. What she said was interesting. The only predictor of how well a new employee would do his job was his resume. Nothing else gave a statistically significant prediction.
So Marissa and I think the poster is right and Fried is wrong :-)