How do you write this article and not once reference I/O Psychology or the literature that examines how well various tests predict job performance? (e.g. Schmidt and Hunter, 1998)
I swear this isn't witchcraft. You just analyze the job, determine the knowledge and skills that are important, required at entry, and can't be obtained in a 15 minute orientation, and then hire based on those things. It takes a few hours worth of meetings. I've done it dozens of times.
But really what all that boils down to is get someone knowledgeable about the role and have them write any questions and design the exercises. Don't let some dingleberry MBA ask people how to move Mt. Fuji or whatever dumb trendy thing they're teaching in business school these days.
The cool thing about it is that the core of it is really just one page.
There's a page in there with a list of types of tests and their respective r values, which is a number between zero and one that explains how well a given type of test predicts job performance based on this gigantic meta analysis the researchers ran. Zero means there's no relationship between the test and job performance and one means the test predicts job performance perfectly.
Generally you want something better than .3 for high stakes things like jobs. Education and experience sits at ... .11 or so. It's pretty bad. By contrast, skills tests do really well. Depending on the type they can go over .4. That's a pretty big benefit if you're hiring lots of people.
That said it can be very hard to convince people that "just having a conversation with someone" isn't all that predictive at scale. Industry calls that an "unstructured interview" and they're terrible vectors for unconscious or conscious bias. "Hey, you went to the same school as me..." and now that person is viewed favorably.
Seriously this stuff is WELL STUDIED but for some reason the MBA lizards never care. It's maddening.