The typical interview (be it a phone screen, or an onsite interview) consists of solving a problem. Some call these problems “puzzles”. They are usually non-real world problems that aim to verify your algorithmic skills and your computer science knowledge. The simple ones include recursion, binary search, basic data structures (linked list, hasthable, trees). The more complex ones require red-black trees, Dijkstra, knowledge of NP-completeness, etc. If you are on the phone, you write the code in a shared document. If onsite – you write it on a whiteboard. So, these puzzles should verify your computer science and algorithm skills. But let’s step back a little and see the picture from another angle.
- what you do on these interviews is something you never, ever do in real life: you write code without using any compiler or debugger. You do that in a limited time, with people watching you / waiting for you on the line. But let’s put that aside for now. Let’s assume that writing code without being able to run it is fine for interview purposes.
- the skills that these puzzles are testing are skills that the majority of developers have never needed. Most people are writing business software, and it does not require red-black trees. What was the last time you used recursion in your business software? So the last time you’ve done anything like that is in college. And many of these problems are really simple if you are a freshman, you did them as a homework just the other day. But then it becomes a bit more tedious to write even things as simple as a binary search. Because you just didn’t do it yesterday. Of course you will be able to do it, but for a little more time, so that you can remember, and for sure by using a compiler. (By the way, the puzzles at facebook were really simple. I didn’t do them perfectly though, which is my bad, perhaps due to interview anxiety or because I just haven’t done anything like that for the past 3 years)
- the skills tested are rarely what you will do in your daily work anyway. Even in these cool companies like Google and Facebook, there are still pretty regular projects that require coding to APIs, supporting existing code, etc. I don’t think you will be allowed to tweak the search engine in your first week, no matter how great you did on the interview
- interview preparation is suggested and actually required before these interviews. Exactly as if it is a college exam. But that’s dumb – you don’t want people to study to match your artificial interview criteria. You want them to be…good programmers.
- focusing on these computer science skills means these companies will probably miss good engineers that are simply not so interested in the low-level details.
Btw, here’s an excerpt from my feedback after my first phone interview with Facebook:
On the other hand, I don’t think having 1st year CS homework problems on interviews for senior developers is a great idea. One thing is – most people (including me) haven’t done this since university, and it looks a bit like trivia questions rather than actual programming.
The problems outlined above are what I don’t like about these types of interviews. And that’s obviously because I don’t like solving these sorts of puzzles. I just don’t like them, they are not interesting for me. You could argue that in addition to your daily job, you can participate in programming competitions (like TopCoder) in order to keep your algorithm skills trained.
I’ll give a short story about my high-school years. There were two student competitions – one was about exactly these types of programming puzzles – you are given a number of them for a fixed period of time, and you must submit a solution that covers as many of the pre-defined (but unknown to you) test cases as possible. The other competition was about creating a piece of software at home, and then presenting it in front of a jury. I was a top-competitor in the latter, and sucked quite a lot in the former.
Why? Because I hated solving useless, unrealistic problems for the sake of solving them. I liked building software instead. I would probably be good at solving puzzles if I liked them. I just don’t. And these are not two levels of skill – one who can solve complex algorithmic puzzles (superior), and one who can’t, therefore he builds some piece of software (inferior). These are two different types of skills. And both of them are very useful in the process of creating good software. One writes the low-level stuff, the other one designs the APIs, the architecture, the deployment scheme, manages abstraction in the code. So, to get back to the question what I do now in addition to my daily job – I build stuff. I’ve worked on a few personal projects that I enjoyed. Way more than I would’ve enjoyed a TopCoder competition.
Unfortunately these cool companies are hiring primarily the TopCoder-type of people. Which probably works for them, because they have a lot of candidates and they can afford a lot of “false-negatives”. But many smaller companies adopt these interview practices, and so they fail to get the best technical talent. The best article on software engineer interviewing I’ve read appeared just a few weeks ago. Jeff Atwood advised how to hire a programmer, and I completely support his approach.
And my problem with interviews is that they don’t actually verify if you can do real programming work. And obviously my problem is that I don’t like low-level and algorithmic stuff, so I wouldn’t be able to work for cool companies like Google and Facebook.
Important note: I’m not saying you should not know what computational complexity is, how a hashtable works, or how to write recursion. You should, because that is basic stuff that you need in order to be able to write good code. But focusing too much on these things is what I consider irrelevant to day-to-day programming. (And for the trolls: I wouldn’t have passed the 2 phone screens if I was a complete dumbass who can only write websites in PHP and thinks a hashtable is some sort of furniture)