4 ms·

I get why companies want to use Leetcode for evaluation. It basically quantifies an applicant down to a number which makes hiring easier and more scalable. Ther

I get why companies want to use Leetcode for evaluation. It basically quantifies an applicant down to a number which makes hiring easier and more scalable. There's a lot less room for bias if every candidate can be boiled down to pass or fail by a single test. That being said, I was on the hiring committee at my last job and personally, I found the best way to hire engineers is to just chat with them about previous projects. I'd usually run through one or two, "just confirming you can code" whiteboard questions and then ask them to tell me about a project they are proud of that they had a large part in. If you ask questions that dig into the project ("How did you do X?", "Why did you do Y?", "Now that it's over, what would you change?") you can usually get a pretty good idea of their skill level and knowledge as a developer as well as some great insights into their personality. Obviously, that's much harder to scale and more subject to bias, but it worked for me.

bradleyjg 7y ago
In addition to being hard to scale and subject to bias, it's hard to know, even for you, if it was actually a good system or not. The best you can hope for is to have decent confidence it had a low false positive rate. Short of industry wide collaboration with some kind of longitudinal study, getting an idea of what does and doesn't work in terms of false negatives is nearly impossible.
- Hermitian909 7y ago
  I think you can tell if it's a good system for your company. As an individual employer all I care about are: 1. are my false positives low? 2. is the time it takes to fill a position low? If the answer to both of those questions is "yes" then the system is great for me (and I've seen the system described above provide affirmative answers to both these questions in a startup setting). Determining whether a hiring system is locally good is significantly easier than determining if it's good in general.
  - okaram 7y ago
    You can say if it's acceptable ... but not necessarily good The problem is that you don't know the cost of the false negatives; you don't know whether the people you didn't hire would have greatly improved your processes / code / business
    - hinkley 7y ago
      Software is a collaboration. Almost by definition, all the stuff I don’t know how to do well is at least as important as the stuff I do, even as full stack person. It’s probably not the stuff you know how to do that makes or breaks your company. Either someone else does it, or you “figure it out” and hopefully you know if you did it right before it’s too late. How do you hire someone to be good at stuff you aren’t already good at? There’s gonna be a lot of those in the false negatives pile.

bradleyjg 7y ago

In addition to being hard to scale and subject to bias, it's hard to know, even for you, if it was actually a good system or not. The best you can hope for is to have decent confidence it had a low false positive rate. Short of industry wide collaboration with some kind of longitudinal study, getting an idea of what does and doesn't work in terms of false negatives is nearly impossible.

Hermitian909 7y ago
I think you can tell if it's a good system for your company. As an individual employer all I care about are: 1. are my false positives low? 2. is the time it takes to fill a position low? If the answer to both of those questions is "yes" then the system is great for me (and I've seen the system described above provide affirmative answers to both these questions in a startup setting). Determining whether a hiring system is locally good is significantly easier than determining if it's good in general.
- okaram 7y ago
  You can say if it's acceptable ... but not necessarily good The problem is that you don't know the cost of the false negatives; you don't know whether the people you didn't hire would have greatly improved your processes / code / business
  - hinkley 7y ago
    Software is a collaboration. Almost by definition, all the stuff I don’t know how to do well is at least as important as the stuff I do, even as full stack person. It’s probably not the stuff you know how to do that makes or breaks your company. Either someone else does it, or you “figure it out” and hopefully you know if you did it right before it’s too late. How do you hire someone to be good at stuff you aren’t already good at? There’s gonna be a lot of those in the false negatives pile.