[woofie11]

Yup. The stupid brain version of a typo.

* Setting the bar at 2 std dev means you need to interview 20 candidates at your hiring cut-off to find one you believe is qualified. The other 19 qualified candidates get rejected by chance.

* Setting the bar at 1 std dev means you need to interview 3-4 candidates to find one who you believe is qualified. The other 2-3 are rejected by chance.

Of course, most candidates aren't right at cut-off, so it's not quite 1:20, but it is pretty random. A good model for hiring, from the candidate side, is that you interview. If you pass, you role a die, just like in an RPG.

Unless you're coming in through a side door or a backdoor, rejections are just part of life. People take it personally, and fault the interviewer for a bad process, but interviewing is fundamentally a very noise measurement. It's the only way we've found to get good employees at scale:

* You set an unreasonably high bar.

* You interview a lot of people.

* Occasionally, qualified candidates pass by chance, but the majority are rejected.

* If you set the bar high enough, unqualified candidates will rarely pass by chance, by virtue of how Gaussian curves look

[stillblue]

I'm still confused. Standard deviation of what? What is being measured here?

[woofie11]

Some measure of employee quality, as evidenced in the interview. To make this simple, let's take an exam instead of an interview:

* Let's say I was hiring someone to do calculus.

* I want applicants who typically score at least a 90% on a calculus AP. The exam has a std. dev. of 3 points.

* I get an applicant, and I give them an AP exam.

* They score a 91. What do I do?

That means they are 1/3 of a std. dev. above my cut-off. They might typically score 88 and got a little bit lucky, or even 85 and got very lucky. Or they might typically score a 94, and got unlucky.

If I hire that person, although they're probably qualified (assuming a uniform distribution of applicants), I'll get a lot of bad hires.

To avoid that, I set the cut-off at 93% or 96%. This means I intentionally reject most people who meet my cut-off. On the other hand, if I hire someone, I can be pretty confident they're qualified.

The cut-off needs to be high in part since the distribution of applicants isn't uniform. Most applicants are unqualified morons. Qualified people will apply to a few places, and are quickly hired. Unqualified people will apply over, and over, and over, everywhere they can.

[stillblue]

>Unqualified people will apply >over, and over, and over, >everywhere they can

Given this is true, won’t it be the case that you’d hire more people who were unqualified but repeatedly got better at interviews as opposed to the qualified one(someone who always gets 91 or 92 in your example)

You’ll actually end up hiring folks who are not good at what they do but people who got better at interviews.

[woofie11]

As a point of fact, that's kinda what happens at big firms. No one has yet figured out a better model. There isn't a magic oracle to point you to perfect employees.

That's not really how it happens, though. You get people who are really good at whiteboard coding and little invert-a-tree type exercises, with no other skills. Most don't do this through interviewing over and over, but by practicing online. You end up with big companies full of people with nothing better to do with their time.

A lot of the people who interview over and over really have no idea what they're doing wrong. Employers don't give feedback, aside from "Thank you. We'll let you know. [ghost]," or at best "Thank you. We filled the role." A half-dozen interviews helps, but from then, it's rapidly diminishing returns.

I like startups, where you can do all sorts of side and back hiring channels, but none of those scale above maybe 100 people, and usually not past a dozen.

[papaf]

Certainly not clarity of written communication.