A Closer Look at Much-Maligned Technical Interviews

They may not be perfect, but they do work.

Is there a relationship between the result of a technical interview (for Google, Microsoft, Facebook and so on) and the performance of the employee? originally appeared on Quora, the knowledge-sharing network where compelling questions are answered by people with unique insights. You can follow Quora on Twitter, Facebook, and Google Plus.

Is there a relationship between performance in a technical interview and employee performance? I believe so, but it's a very hard thing to study.

The Silly Google Study

Google tried to study this, but the study was fundamentally flawed. What Google did was attempt to correlate interview scores with performance review score. Sounds fair, right?

Not at all. There were a number of flaws with this.

First, the interview score does not totally represent your interview performance—ditto for your performance review score. The actual description heavily influenced how the committee sees you. There is a strong correlation between how you performed and your score (hopefully), but it's not perfect. That will mask some of the correlation between interview scores and performance review scores.

Second, and more importantly, Google did not hire at random and then see how those people did. Of course not. No company wants to take that risk.

Instead, they looked at those who got hired. To get hired, a candidate must perform very well. When my hiring committee voted to hire someone, they nearly all had an average score of between 3.1 and 3.3.

So what Google concluded was that in this very high and very narrow band of scores, interview scores are not predictive of job performance scores. Gee, shocking.

This is like wanting to study if athletic skill correlates with life expectancy (sounds plausible!). But, oops, the only data you have is that of Olympic medal winners. You don't find a correlation therefore… nothing. You can't conclude anything from that, other than that minor changes in phenomenal athletic skill don't correlate with a statistically significant longer life.

So let's just dismiss this study…

Why I Believe It Works

There are two reasons I believe it works.

First, virtually every top tech company has built their engineering teams with this process: Microsoft, Google, Facebook, Amazon, Dropbox, Airbnb, etc. There are little deviations, and some companies do a better job with this process than others, but the types of questions are essentially the same.

I am not saying: They're famous, so they must know what they are doing. No no.

What I'm saying is that if it doesn't correlate reasonably with job performance, then you must believe that engineers hired at random could build systems of this complexity. I find this highly unlikely.

Second, I do interview coaching for startups going through acquisitions. This has allowed me to interview the same person several times and also talk to their manager about how they are doing. My perception of their strengths and weaknesses seems to match up reasonably well with their actual strengths and weaknesses. When I say that someone is really smart, they are—and this process does a reasonably good job at identifying that.

What I mean by “works”

It's not perfect by any means. There are also times when I have a sense that someone is bright (that is, they do have the skills that I'm trying to hire for), but they are so nervous that I can't get enough data to feel confident about hiring them. They get rejected, and that sucks.

The existence of some false negatives and positives though does not mean a process is totally flawed.

It's about whether it works overall, and I believe it does.

What does/doesn't work:

• The technical interview process is designed to identify people who are smart and have the aptitude to become good coders.

• It also can somewhat evaluate a candidate's ability to work on a team, although many processes can identify that.

• It does not identify skill with a technology. If you want that, you'll need to add that in.

• It does not identify work ethic. Good luck finding a process that does.

• It is relatively efficient, scalable, and easy to implement. That matters too.

Again, it is not perfect. But when implemented well, it's pretty good.

Lots of companies, however, don't implement it well. They ask easy questions or common questions or just stupid questions. I've done lots of interviewer training. There are some interviewers who know what they're doing, and a whole lot who don't.

This question originally appeared on Quora. More questions on Quora:

* Job Interviews: What is a typical software engineering interview with you like?

* Preparation: How much does software engineering interview preparation actually help?

* Life Lessons: What words should you never say when being interviewed for a job?

Photo Credit: Klaus Vedfelt/Getty Images