[tbrownaw]

I'd rather see a usable sort key (item a), maybe with a short description.

Given that you have a position to fill, and a deadline for when you need it filled, and a large number of applicants, and other everyday work you need to keep doing... how many minutes each are you able to spend on a first-pass review of the applicants?

[randomdata]

As they saying goes, you get what you give.

There's nothing wrong with picking someone that way, but you're essentially selecting someone based on random chance. You'll probably find someone who is adequate at the job, and if that is what you want, who am I to judge?

But the recurring theme on HN is that people want to hire the best of the best. The best people are 10 times more productive, they say. It may be conjecture, maybe businesses don't want that kind of person at all, but selecting someone from a sorted list is almost never going to find you that person. For that, you're going to have to put in some effort.

Anyway, it doesn't matter how you do it. The benefit of the open market is that the one who does it right will find greater success with their hire.

[tbrownaw]

At some point, you need to decide that one candidate is better than the others.

So, you need some function that provides at least a partial ordering of your candidates (ie, a sort key).

Computing these sort keys is expensive. It can be optimized by (partial) precomputation. Maybe this is a letter grade from a school, or having/not having a degree, or a summary prepared by the candidate (a resume), or whatever.

You probably don't entirely trust whoever did that precomputation. So you'll want to check their work. You can't do this for every candidate since you still want time to sleep at night. So, you need to use that pre-computed sort key as a filter to throw out at least some portion of the applicants. But the ones that you don't throw out, you'll want to investigate further.

This could be reviewing their past work (github portfolio, etc). Or it could be in-person interviews, or phone interviews, or whatever.

If looking over a portfolio is less expensive than doing an interview, does it let you filter out enough candidates to be cost-effective?

But the recurring theme on HN is that people want to hire the best of the best. The best people are 10 times more productive, they say. It may be conjecture, maybe businesses don't want that kind of person at all, but selecting someone from a sorted list is almost never going to find you that person. For that, you're going to have to put in some effort.

The only way to find the best (or the "good enough") is from a (at least partially) sorted list. What varies is the quality of the sorting, and of the contents of the list. Given limited resources, how do you maximize the quality of the sorting (assuming somebody else is handling maximizing the quality of the list contents)?