3 ms·

The problem is that there is almost no correlation between how candidates do in whiteboard interviews, and the ability to perform the job on a day to day basis.

by vfc1 8y ago

The problem is that there is almost no correlation between how candidates do in whiteboard interviews, and the ability to perform the job on a day to day basis. Also, and the article says it: you are being tested mostly for personality compatibility with the interviewer, which again has nothing to do with the ability to perform the job. This plus the fact that the interviewers are volunteers, creates overtime an effect where lots of people in the workforce have similar personality traits, which is not a great thing because it creates an imbalance in the normal personality balance that a group needs to work properly. Personality imbalances in groups usually lead to dysfunctional situations over time: too many sloppy people who just cowboy code until dawn, too many excessively cautious people that won't touch a line of code any more with fear of breaking something, etc.

blauditore 8y ago
> The problem is that there is almost no correlation between how candidates do in whiteboard interviews, and the ability to perform the job on a day to day basis. What else would you use as metric? Talking about technical topics? In my experience, there are people good at talking about those things, but miserably failing at the actual job. Introducing simple whiteboard programming tasks in addition has helped weeding those out and dramatically improved hiring success.
- vfc1 8y ago
  There are some very solid alternatives to this current system, which lets face it some of things that they make you do in an interview are borderline ridiculous and its a matter of lucj more than anything else. For the technical part, I would as screening value forms of unforgeable proof of technical competence, such as Github profiles, meaningful open source contributions, stackoverflow contributions. Next for the technical part, a homework assignment. Homework is much more close to the real world experience: get a task and would you have to figure it out. Then the technical interview would be to explain the homework. There the only goal is really to confirm that the person did the homework themselves rather than anything else. For the behavioral part, a full battery of psychological tests, to weed out psychopathic traits mostly. There are alternatives, the current interviewing system is not a good solution.
- stale2002 8y ago
  Sure, give someone a problem slightly harder than fizz buzz just to make sure they aren't a complete fraud. But the difference between someone who can solve an easy whiteboarding problem and a hard one is merely the fact that they practiced for months doing it. It is unlikely that solving the hard problem is much of a signal other then that.
GlennS 8y ago
Are you sure? I don't think the pseudocode whiteboard exercises are pointless, I think they're a useful filter. You can make them simple, and anyone remotely competent who isn't completely crippled by nerves should be able to zip through them. They don't have to be about esoterica. I just want to know that the candidate can attack and break down a problem, even if they don't know how to do all the bits immediately. And I want them to have heard of a hash table before. I've switched to 1-2 hour homework exercises on the advice of an experience colleague though.
- vfc1 8y ago
  The problem is, there is a huge disconnect between hash tables and the work we do every day, so why to use that as a criterion? Homework exercises seem like such a better way to go, its much closer to the actual work experience. And its easier for the interviewer, all they have to do is to assess if the person actually did the exercise themselves.
  - GlennS 8y ago
    Wait, what? What sort of programming do you do where you don't use hash tables every day?
    - vfc1 8y ago
      Although hash tables are used internally by libraries, they are not directly used by the application developer. When was the last time you had to hand code a hash table? Unless you are a library developer, its an example of a task and knowledge that will never be needed. If its something that is never used in that particular job, why ask it in the interview in the first place?
      - GlennS 8y ago
        Well, our disagreement is just a matter of miscommunication then. I never said anything about hand coding a hash table. I would like prospective candidates for programming jobs to know that there is a thing in their language of choice that lets them make fast lookup tables.
_pmf_ 8y ago
> The problem is that there is almost no correlation between how candidates do in whiteboard interviews, and the ability to perform the job on a day to day basis. > Also, and the article says it: you are being tested mostly for personality compatibility with the interviewer, which again has nothing to do with the ability to perform the job. For a majority of candidates, there is a correlation (both aspects). It might not the best or even a "good" way of determining suitability for a job (as, say, doing some actual domain work), but you cannot just deny that there is a correlation.
grogers 8y ago
> This plus the fact that the interviewers are volunteers Is this true at Google? I've never heard of a company where interviewing was voluntary, it's always been part of the normal job responsibilities.

by vfc1 8y ago

blauditore 8y ago

> The problem is that there is almost no correlation between how candidates do in whiteboard interviews, and the ability to perform the job on a day to day basis. What else would you use as metric? Talking about technical topics? In my experience, there are people good at talking about those things, but miserably failing at the actual job. Introducing simple whiteboard programming tasks in addition has helped weeding those out and dramatically improved hiring success.

vfc1 8y ago
There are some very solid alternatives to this current system, which lets face it some of things that they make you do in an interview are borderline ridiculous and its a matter of lucj more than anything else. For the technical part, I would as screening value forms of unforgeable proof of technical competence, such as Github profiles, meaningful open source contributions, stackoverflow contributions. Next for the technical part, a homework assignment. Homework is much more close to the real world experience: get a task and would you have to figure it out. Then the technical interview would be to explain the homework. There the only goal is really to confirm that the person did the homework themselves rather than anything else. For the behavioral part, a full battery of psychological tests, to weed out psychopathic traits mostly. There are alternatives, the current interviewing system is not a good solution.
stale2002 8y ago
Sure, give someone a problem slightly harder than fizz buzz just to make sure they aren't a complete fraud. But the difference between someone who can solve an easy whiteboarding problem and a hard one is merely the fact that they practiced for months doing it. It is unlikely that solving the hard problem is much of a signal other then that.

GlennS 8y ago

Are you sure? I don't think the pseudocode whiteboard exercises are pointless, I think they're a useful filter. You can make them simple, and anyone remotely competent who isn't completely crippled by nerves should be able to zip through them. They don't have to be about esoterica. I just want to know that the candidate can attack and break down a problem, even if they don't know how to do all the bits immediately. And I want them to have heard of a hash table before. I've switched to 1-2 hour homework exercises on the advice of an experience colleague though.

vfc1 8y ago
The problem is, there is a huge disconnect between hash tables and the work we do every day, so why to use that as a criterion? Homework exercises seem like such a better way to go, its much closer to the actual work experience. And its easier for the interviewer, all they have to do is to assess if the person actually did the exercise themselves.
- GlennS 8y ago
  Wait, what? What sort of programming do you do where you don't use hash tables every day?
  - vfc1 8y ago
    Although hash tables are used internally by libraries, they are not directly used by the application developer. When was the last time you had to hand code a hash table? Unless you are a library developer, its an example of a task and knowledge that will never be needed. If its something that is never used in that particular job, why ask it in the interview in the first place?
    - GlennS 8y ago
      Well, our disagreement is just a matter of miscommunication then. I never said anything about hand coding a hash table. I would like prospective candidates for programming jobs to know that there is a thing in their language of choice that lets them make fast lookup tables.

_pmf_ 8y ago

> The problem is that there is almost no correlation between how candidates do in whiteboard interviews, and the ability to perform the job on a day to day basis. > Also, and the article says it: you are being tested mostly for personality compatibility with the interviewer, which again has nothing to do with the ability to perform the job. For a majority of candidates, there is a correlation (both aspects). It might not the best or even a "good" way of determining suitability for a job (as, say, doing some actual domain work), but you cannot just deny that there is a correlation.

grogers 8y ago

> This plus the fact that the interviewers are volunteers Is this true at Google? I've never heard of a company where interviewing was voluntary, it's always been part of the normal job responsibilities.