1 Getting the actual real requirements nailed down
2 Designing the system to run in the real world accounting for all those edge cases and falilure modes.
That makes me wonder if a business-analyst or product-manager role should involve a technical interview where you spec out a mock product to see how well you gather requirements.
If by technical interview, you mean the common process of making someone do a simulacrum of the job they will do over the course of days, in a few overly tense hours, then no I can't possibly recommend that.
If on the other hand you are saying, "should I be evaluating product specification and requirements gathering when hiring business analysts, or product managers" then I would say of course! What else would be evaluating them on?
I wouldn't say that coding is the easy bit (if it were I wouldn't have seen so many terrible code bases in my life). I'd say that requirements gathering and systems design are also crucial difficult parts of the process of developing software and any developer acquisition methodology should take them into account as well.
In your experience. Not to be flippant, but if you are developing software in an environment that has lots of changing requirements, there are ways to write that code and developers who are good in those situations.
Similarly, there are environments where static requirements are cornerstone for success. Certain methodologies deal with this environment better than others and certain developers will be more successful than others in these environments.
Above a certain threshold, yes. But I have seen so much utterly crazy hacked-together crap, which was clearly not caused by changing requirements, but rather either incompetence or negligence. I'm guessing that so have you.
> Getting the actual real requirements nailed down
Wherein the programmer plays the role of social worker. Seriously though, the skill sets are very similar. The client is typically the ultimate source of the requirements, but it's never a simple case of asking the client "what should I build? More often than not, the client doesn't know what s/he wants, wants something that will actually hurt him/her, is actually seven different people who want opposite things, wants the roses simultaneously painted white and red, etc. Which is why I find requirements collection the most challenging part of the job.
It is the central problem in the software development community right now that we have no good objective way to measure software quality. Without that measure we can't judge software methodologies (thus the TDD flame wars) and we can't judge software developers (thus the interview flame wars).
So in each case we come up with proxies for this measure. Having done the interview side of the fence a lot, my strong intuition is that nearly everything people try to measure in developer acquisition strategies either don't correlate or negatively correlate to hiring good developers.
At this point, the only advice I give people about this is figure out a repeatable, measurable process for your organization and measure every single filter you use during hiring. Test the measures on just hired applicants, applicants a few months out, and applicants a few years out. It is time consuming and fraught with problems but there is no magic bullet out of this morass.
Yeah I'm completely fed up with this whole GitHub or GTFO mindset. Many good programmers:
A.) Are great in their dayjobs and don't wish to spend their free time writing Open Source. Maybe they have families. Maybe they have non-coding hobbies.
B.) Work on private repos all the time.
Whiteboarding lets me prove raw problem solving IQ without being negatively judged for lack of twitter followers or technical blog posts.
> Whiteboarding lets me prove raw problem solving IQ without being negatively judged for lack of twitter followers or technical blog posts.
Portfolios allow devs to prove raw problem solving IQ without being negatively judged for lack of documentation or a debugger.
There are developers who aren't good at whiteboarding interviews because they don't think that way and work better when they're at a laptop and can refer to other source code and documentation and can run their code continuously as they write it.
With those devs, they may not look impressive in a whiteboard interview, but you can look at their GitHub portfolio and draw conclusions about their programming ability.
An ideal interview should be holistic and include both off-the-cuff code and polished projects.
"without being negatively judged for lack of documentation or a debugger"
A good judge at a whiteboard won't mark you down for uncertainty about interfaces, which partially ameliorates this, for those cases where you have a good judge...
I'm also tempted to say that needing a debugger to lay out a mostly correct solution to a small problem is weird - can you elaborate on how that might be applicable?
In any event, I agree with your ultimate conclusion - holistic is probably a better way to go.
I have walked out of technical interviews having spent 40 minutes struggling to write code on a whiteboard, sat down in front of a laptop, and coded a working solution in 5 minutes.
For my current job I was provided with the specifications for a simple app to display images from a flickr rss feed and asked to code however I felt was appropriate. It was interesting and fun, and vastly less stressful than any whiteboard test.
It was also a great indicator of what working at the company was actually going to be like.
> I have walked out of technical interviews having spent 40 minutes struggling to write code on a whiteboard, sat down in front of a laptop, and coded a working solution in 5 minutes.
This is how I bombed a DevOps role interview at Twilio, but was picked up as a VP of Engineering elsewhere. +1.
my twilio interview was the worst I've ever had in my life. Flew coast to coast, aced the tech, aced the preso, then was rejected on the basis of a perceived lack of motivation. In other words, the fact that I didn't already use twilio and sing its praises like a good koolaid guzzler was the only reason.
Agree with OP. Coding under pressure is a horrible way to measure good coders, because it's an artificial construct.
Whoa. "The guy didn't seem motivated" was the standard empty excuse 10 years ago. They are supposed to say "you wouldn't be a good fit as this time" as the standard empty excuse these days. They need to get with the times.
Companies reading my parent comment should take note - abstruse technical questions on a whiteboard are fairly ineffective.
For people looking for alternatives to whiteboarding or phone screens where the candidate writes code in the blind, check out https://coderpad.io
It lets you write and execute code with your interviewee in real-time, and provides a much more native programming experience easily and over the browser. Some rather large companies have started using us exclusively in the in-person technical interview, even buying Chromebooks especially for the application.
Disclaimer: I am the guy who makes CoderPad and am obviously biased.
If you ask me to "code" something on the whiteboard, that simple request is more an indication of your competence or cluefulness, or rather lack thereof, than the result would be an indication of my qualities.
The proper reaction to such a request would be: "So, you guys do all your work here on whiteboards? That's seems unusual, I used laptops or workstations at all my previous jobs."
What if I told you before you came in that we'd be coding on whiteboards. That my expectations were calibrated properly for that, and that the problems would be "white-board" sized.
I've done this -- had people talk to me about their anxiety or problems with it, and put them at ease. For some, I have recommended that they just go practice of an hour or so with a friend -- it's really not that hard to code "whiteboard" level code if you code every day and practice it a little. It's such a common tech interview style, that if you are looking for a job, it's worth working a little at it.
Frankly, in a code-editor, my expectations are much higher -- I don't even require you get any framework class or method name right (or even perfect syntax) -- but the compiler will. It gets in the way of the essence of the question -- which is more about collaboration.
So instead of not doing a stupid thing you still do it but just warn people in advance and even tell them to practice doing that stupid thing, even if it will not be a part of their job?
I don't think it's stupid, but I do think you shouldn't be surprised by tech interviews.
This is for a screen to make sure you are a programmer. If you can't code up a four line function without an editor, there are going to be a lot of jobs you might like that you won't get. Ditto with calling strangers stupid.
Not all our work, but we definitely use code sketched out on whiteboards as a communication tool. Are you claiming that you can't reason about algorithms or communicate them effectively if you're not in front of a computer?
I don't give a flying fuck about minor syntax errors, forgotten API, or anything a compiler can catch. If you forget an API, I either give it to you or ask you to make up something reasonable. (Chances are, I can't remember it either). But if you can't describe an algorithm to me in front of a whiteboard, I don't want you on my team. And if you can't translate it to pseudocode that resembles the language you're going to be working in, I also don't want you on my team.
You bring up an important point. If you liked the job & atmosphere, don't give up! If you feel you bombed on something, research it as soon as you get home and do what you can to impress them.. send an email back with working code (or better answers).
I totally failed an interview because I was panicked when writing some simple code on a whiteboard. The interview started with three interviewers in the room but only one was talking to me. The other two remained silent the entire time. The talking interviewer asked me to solve a simple problem on a whiteboard, and I to my own surprise just couldn't do it. I'm already nervous just meeting new people who don't know me as a person, and I'm trying to perform my craft outside of my usual environment while being judged.
I use a whiteboard at work to pseudo code a solution, and the writing looks like chicken scratch but even my co-workers get what I'm expressing to them. Then I open up a console and start typing out my idea to see if I was right. That's how I normally solve some throw away piece of code. I don't go into a manager's office and call some other random people and in detail explain to them my idea to a solution.
I also got the feeling they never saw my resume because I was asked if I had a github profile, and it's on my resume.
this is true but at the same time i think great programmers are also really resourceful. cutting off those resources in a standup can make it difficult to evaluate a candidate, especially if the programming question is something that only one party knows about. i think being able to figure out how to get the right answers is much more important than having the candidate know an arbitrary one that the interviewer asks.
I hear these complaints all the time, yet not many people complaining actually offer viable solutions for vetting a candidate's ability.
That said, I would say Github contributions and maybe paired with StackOverflow activity may be a good substitute for a technical interview when available.
The author said Github contributions and/or a specific programming project. My development group does a relatively simple tech interview + a programming project.
I think it's just a symtpom of an evolving industry. Studied solutions are better than viable solutions, and we may just be studying it without acknowledging that fact. Someone should put some actual researchers on this problem though.
We've recently given a huge amount of thought to this idea (how to technically assess applicants) and sought input from many sources. This is because we are building a curated online community of technical (job) profiles, and we have a very strong incentive program for people to sign up. (It's being announced any minute.) In fact, you could say our incentive program is too strong, and we are afraid of getting "CTO/senior web applications developer" with "20 years of Linux and AWS experience" who can't actually define what "ls" is or what "for" does (in any language).
So, what shall we do? How can we very strongly incentivize every qualified person to upload their profile, in a way that lets us curate people's actual abilities?
The suggestions this article makes, to demand githubs or complete projects done for demonstration purposes, have both been rejected. Sure, it may be a good strong signal. But then so is the signal of someone creating a complete application for your company (complete with branding) to actually A/B test on the front page, and then if it does well, for the canddiate to show up for a 3 month unpaid trial during which he or she must contribute as a full member of the team and can be dismissed at any time for no reason. I guarantee, anyone who passes that test would be a great candidate. The issue is that there are perhaps three people on the planet who would even consider applying to a company on those terms - and they're the original founders' siblings.
The problem is that it is not a reasonable burden on job applicants, and neither is a complete github profile, and neither is a complete programming project. It's just too much.
What is needed is a smaller burden that is a good, strong signal.
We think we've found one, but are not sure. (You can see it on our web site as soon as we announce.)
In the mean time, if anyone here has any breakthrough ideas in this space, we would be very interested.
I recently drove 3 hours to a technical interview. I had done plenty of screening, including screenshare while I wrote code to the guy's delight.
He invited me down for the technical interview. I came down a week later. His wife had just delivered a baby, so he was out. The guys in the room were a business level ex-marine who was there for god knows why, and a technical dude who clearly was pissed he had to sink to the level of doing the interview in the first place.
For the first 45 minutes, I aced everything he threw at me, and I noticed that he wouldn't drill in on anything that I clearly knew very well. He kept jumping from topic to topic, and eventually asked to do an extremely tricky SQL query, but write the code to do it in awk on the whiteboard. The job posting said nothing about awk, and I told him I didn't know awk that well. He then sensed i wasn't a command line master (the job posting said nothing about needing to be a fucking sysadmin or know advanced CL stuff) and hammered me on things that I had already said I didn't know.
2 things became clear:
He wanted to show off to his boss and make himself look like a badass while simultaneously making me look incompetent.
There was no fucking way I was going to accept a job working here if it was offered.
I didn't so much as get a phone call or an email to thank me for driving 6 hours round trip. Nothing. Which screams out to me that this company sucked.
Perhaps your comment was intended as a sarcasm, in which role it serves admirably. I'll please myself nonetheless to answer as though it were genuine: If you want to make absolutely certain that you never, ever hire the best candidate you can find for any of your positions, I can't think of a better way of doing so than to bring this sort of attitude to the hiring process.
Yeah, that pretty much sucks. I generally try and ask people about what they did and drill into whatever it is they know, but most interviews I've gone on the interviewer wants you to know the stuff the interviewer knows/is good at.
Absolutely, and in the United States there's a standard mileage rate that the IRS will let you use for computing a tax deduction (if it's for business), which companies often use as the amount to reimburse. For 2014 it's $0.56/mile. Obviously a company is under no obligation to reimburse you for travel expenses to an interview, but in the case of a long drive like that I would certainly ask. In fact, for a 3-hour drive each way I would probably also ask to be put up in a hotel room the night before the interview, reimbursed by the company, so that I wouldn't be tired from the drive on the day of the interview.
I recently flew in for a 4-hour on-site interview halfway across the continent, which included a live test of programmer skill which required knowledge of javascript, bash shell commands, and Google searching, along with exposure to the company's internal tools. Also featured was an abstract problem solving interview, and a short sales pitch for myself in front of the whole team.
After I left, I also did not get so much as a phone call or e-mail. Neither thank you nor followup. When I attempted to re-establish contact, it was like shouting into a black hole.
The feedback I received on site from the interviewers was neutral to positive, with one interviewer claiming that I was the only applicant to come up with the correct response to their abstract brainteaser.
If I take two whole days off from my existing job to come to you and indulge you in your cute little skill tests intended to prove my bona fides, you ought to have the decency to follow up.
So you're not the only person to be on the receiving end of this unacceptable behavior from interviewing companies.
In my case, I was pissed, but mainly due to opportunity cost rather than the trip itself.
By the way, you should email their founder and let them know. I doubt he/she is aware of this, and would be equally furious to know that their workers are burning bridges like this, because that's exactly what they did.
In my case, the company did actually burn a bridge. My much larger company was looking for a strategic partner with high levels of data science expertise on sensor based data. This particular company came into consideration, and I made sure to remove them from selection. Hell hath no fury like a geek scorned.
Founder? The company in question was founded in 1986 and swallowed by a multinational in 2000. If I'm going to tilt at windmills, I'd prefer to aim my lance at those that have not yet unseated me.
I think that last sentence is the key point here - it doesn't sound like somewhere that would be a very friendly environment in general. This kind of thing wouldn't happen where I work, but either way, the technical interview isn't really at fault.
This sums up the general problem with intervie(wer)s: most of them are wrongly focused on "catching the candidate on incompetence" instead of verifying how well will s/he perform on a given job.
During my last job hunt, I attended an interview at a mega corp. A interviewer walked in and then the interview started. Not sure why, As it went along, the guy seemed to get angry and angrier every time I answered a question. It was weird behavior.
But from what I know, running into such people is pretty common in interviews.
Here's what I've been doing. I'm running on an egregiously small sample size right now, but so far, so good.
My goal is to answer "Are you honest about what you know, excited and curious about what you don't, willing and able to learn, and driven to excel at the role that we're discussing?" Sometimes answering this requires code, but generally not a whole lot, and mostly I just use it to tease out where someone actually is vs. where they say they are. I don't want to hire a candidate who is best suited to make my problems go away tomorrow. I want, a year or three down the line, to say I built the best team possible.
Maybe I can say this because I don't think the "technical" part of what we're doing is the major challenge, though I think that most engineering teams are actually in the same boat as us. Google has far different problems than we do, so their hiring process is rightfully different than ours. And maybe I'm just getting it all wrong. We'll see what happens.
I would be more specific - Death to Interview By Whiteboard.
I haven't had many, but every whiteboard interview I've had has been miserable. At best I've just about managed to convey what I mean, at worst I don't even understand the question.
One particularly memorable exercise had the interviewer write out HTML on the board and ask me to write out the CSS to turn it into a dropdown menu. I didn't even know where to start, because I couldn't construct a mental model of what was going on where. Because outside of the white board interview I never have to.
A much better technical interview that I do not with death upon is one that put me in front a computer which was rigged up to a projector that the interviewer could see. We talked through my solution to an exercise as I constructed it, and I was free to switch between the browser and editor as I wanted. You know, like I might in a normal, actual working day.
An even even better interview was one that left me in a room for 45 minutes with a task. After that time I sat down with the interviewer and talked through what I did.
"Look at github contributions and stackoverflow posts"
That's fine for startup-ish hires, but in most of those cases, isn't the candidate being sought out/referred/etc? Almost all of my work is on non-public SCM systems, and frankly, my job doesn't leave a ton of time to post to *overflow sites, and when I go home, I'm home, and spend it with my hobbies and family.
I also consider myself a really great developer. Especially in the more corporate .net world. When I review candidates, I find that a simple (at a PC, with resharper installed) coding test, with a prebuilt solution, needing only an implementation of a method done, is a great filter for competency. And after all, rough competency is the most I hope to get out of a coding test. The far more important part is always the stuff like how well I think they'll fit in a team, how their approach to solving problems in general is, etc.
Syntax memorization is no longer a measure of a good developer. All it shows is basic competence in a language.
This is a big problem in the industry, in order to really be seen as a good candidate you have to make your work a second part time job so you have something to show off publicly. If you work for a private company with private software, you could still be really good but looking at your github profile, many will dismiss you right away.
Of course the flip side is, you can probably get by with minimal outside work. Do some code katas, have an up to date best practices simple app in github, etc. I think as long as you have something in github that looks good it will do the job, even if you're not working extra hours daily on open source projects.
I do love the absurdity in the UK where recent employment contracts (last 5 years or so) often state that "all your everything are belong to us" (IP, copyright, creative output, thoughts, ideas etc), yet companies also want applicants who actively contribute to or release open-source software.
Having worked under such terms, I've never felt I could participate in open source without jeopardising the project.
Its weasel words for they want a noob, recent grad, or a contractor.
They don't want "You were doing exactly the same thing over there, now you'll be doing exactly the same thing over here", because they don't have the money in the budget for a real applicant.
Iambob doesn't account for the fact that interviewing candidates falls into a Simpson's Paradox for this instance.
You have two cohorts which comprise a majority of your applicants:
1. Unemployed, X% qualified. Has time.
2. Employed, 100-X% qualified. Does not have time.
X is small. If both candidates are qualified, tests for X also find how qualified they are so they can be compared.
His methods are objectively better, provided that both candidates have the same amount of free time. If you compare two unemployed engineers, you'll pick the right one. If you compare two employed engineers, you'll pick the right one.
If you compare an unemployed engineer vs an employed one, however, you'll probably pick the unemployed engineer, even if the employed one is better. Because he doesn't have as much time for your homework or to maintain a Github profile.
Thus, if you're comparing 30 engineers, half of whom are employed, you'll pick the most qualified unemployed candidate, rather than the objectively most qualified candidate.
That's why, even if the test is slightly worse, with an interview room only test you'll consistently pick the better candidate: a large, large segment of talented candidates are employed.
I would love a statistical analysis of Simpson's Paradox and technical interview methods if anyone's ever done one.
If you're employed, it's generally a lot easier to find the time to do a coding problem at home than it is to actually physically come in during a weekday and talk to people.
Take a sick day -- have a "doctor's appt". It's really not that big a deal. "Hey, I have to come right from work, do you mind if I don't wear a suit so I don't have to go home and change" -- it's a well-understood problem.
Surely, though, it's easier to just do a coding problem in the evening? No worries about coming up with a lie for your employer, no possibility of suspicion, and some people have to take sick days out of their vacation time sadly.
Some people are tired and not at their best after a full day of coding.
I talk to the applicant and try to work something out. I interview weekends and evenings if that's what they want. First interviews are always on the phone. Eventually, they have to come in, but by then, we are pretty sure it isn't a waste of time.
I can't code shit in the evening after work. Some times I take vacation days, intending on doing nothing other than working on personal projects during daylight hours.
Too many times I've been in a technical interview where they asked me to define design patterns, but never ask me when/how to actually use them. I usually decline their offers.
When I interview candidates, I look for a good foundation for what they will be working with like do they know the difference between class and struct and the implications of using them (day 1 kind of stuff).
Then I ask them about how they would go about solving a problem they were unfamiliar with. Google-foo is a skill that must be learned and honed. I don't care if you can spit off all sorts of acronyms, I want to know if you are capable of using common sense to solve a problem.
Last but not least, how do they stay up-to-date and relevant. Not looking for the 8 to 5 developers and not interested in those cutting-edge guys either.
Must have good understanding of the basics, principals and skills in problem solving. Telling a block of code that you have a masters in CS will NOT make the bug go away.
The term "senior developer" is an unfortunate one because it can mean a really good, skilled and experienced developer. But it usually means 20 years of experience (1yr exp * 20) or they've just been around so long that the "senior" part refers to age.
OP has the wrong assumptions. The live technical interview is already well-known to be flawed, because qualified people often fail. However, only qualified people can pass.
However, it's true that coding exercises that can be completed in an interview are arbitrary and abstract (by necessity, since you have no context.) In my opinion one should use live coding exercises to test fluency -- can you code at the level you are claiming? Debugging something might be a better test. Then pair that up with a short (short!) take home exam to test problem-solving.
The one thing that really has to go is whiteboard coding. Ask the candidate to bring a laptop or give them a used one.
OP also thinks that high-pressure situations don't occur with programming. But if you work on the web, sooner or later you will be debugging a tricky issue that's costing that company $X per hour while your all your managers up to the CEO hover behind your desk. Being able to communicate with non-technical people and inspire their confidence is also very important. Often more important than coding ability.
If only this were true. There is a whole class of people out there who are stellar face to face and not great for your specific technical/business situation.
Yeah. To be honest, there are cases where I've been that guy.
But past a certain level, it's your job to explain your own weaknesses too. I just had a technical interview the other week and they relied way too much on definitions and whether I'd heard of this and that. I happen to do great at those but I want to be sure I can be productive (the position uses Elixir, which I'm totally unfamiliar with) so I'm suggesting they give me a take-home exam.
The assumption that only qualified people can pass a technical interview is somewhat wrong.
I have had my fair share of false positives. And probably I have been one or two times the false positive. And this is something that everyone can relate I suppose.
Sometimes you just need luck - like the gotcha question being something you have debugged soon in a language you are not that competent at.
The one thing that really has to go is whiteboard coding.
A smart candidate will bring a notebook computer, loaded with suitable IDEs etc ready to go, without being asked to. Impresses people when you tell them to continue the interview while you write the requested code (multitasking, competent enough to spare nontrivial brain cycles answering unrelated questions).
Judging by the downvotes, this isn't something HN commenters find advisable. It's also something I hadn't thought of, and as I am currently looking for a new position, I wondered whether it was something I should consider doing.
Will some of those with opinions of this suggestion describe those opinions, so that I can better evaluate whether or not it's actually worth considering? Thanks!
As I'm responsible for a small development department, I regularly do interviews including written tests. If one of the candidates would ask me if they could do it on their notebook, I would probably say no, as it sort of defeats the purpose.
However, I don't think bringing your notebook to show off some of your work is a bad idea.
How does it defeat the purpose? We don't code on whiteboards, so just the mental transition of coding with a pen while standing up - under stress to boot - is not a good indicator of using proper tools the normal way.
Downvotes? Meh. Taking a notebook and writing the solution with sensible tools - while fielding unrelated questions from 4 interviewers - got me an offer.
In the age of ultraportable computers and free/cheap IDEs, why would a software engineer not take his toolbox, prepared to demonstrate his abilities on demand? Doing so got me two offers.
Well, I've seem a lot of false positives. The questions asked tend more to the mathematical side, So I've seem smart people but with little programming experience, that didn't knew how to write quality code, pass those interviews.
The problem is that problem solving with fancy algorithm's can even be part of the job, but are normally just a small fraction.
I've never seem dumb people going incredibly well in technical interviews, that's true, but I've seem a lot of bad programmers giving great answers to those abstract problems. And sometimes, they're knowledge wasn't actually a fit.
Github profile: How many repo's do they have? How active are they? Do they contribute to open source projects? And then, of course, the obvious code quality questions.
I thought most people came to the conclusion that this is a horrible metric by which to gauge competence?
Then again...
Now us Python developers, Ruby developers, Java developers, PHP developers, and (God help you) Perl developers, let us put aside our meaningless and largely annoying quibbles to unite under a single banner. For once, let's say who cares whether Django or Rails is the superior framework and just join our voices to call for an end to the modern technical interview.
The final paragraph shows what demographic the OP belongs to, so it's hardly a surprise they think that way.
I've tailored my learning and goals around the fact that projects are one way to get my foot in the door without a degree or experience. Otherwise, they just have to take my good word and burn time interviewing me. I thought this would be true regardless of development sector.
Once you're dealing with someone who has verifiable experience, then repo activity should be weighed much less (unless it's something really awesome, then chances are you're dealing with an awesome developer).
If you're critical of someone lacking github repos but they have experience, then I think it becomes a symptom of "everyone trying to stand out"-itis. It's probably best to have different solutions for different skill/experience levels.
Just because it's hard to do technical interviews well doesn't mean that doing them well isn't worthwhile.
In my technical interviews, I look to talk through specific technical problems, and they're usually real-world bugs or design issues I've worked through in my job. There's not usually a lot of code involved, and the code isn't that tricky. It doesn't usually even rely on too much specific domain knowledge. It's about being able to take a broad technical question, formalize it a little into something concrete enough to be discussed, identifying constraints on the solution, working towards a solution, reasoning about performance, and discussing tradeoffs of this approach vs. others. In short, it's about technical reasoning. I don't care if we get to the right answer by the end if we've had a fruitful technical discussion.
Code samples are a good idea, but they're as deeply problematic as interviews. Most of the code from most of the best engineers I know is closed-source, not on github. 48-hour programming projects can tell you what someone can cobble together, but they don't say anything about the person's attention to longer-term concerns around design or code quality. Moreover, being able to code up a technical solution to a 48-hour problem is like basic literacy: I expect at least that, but I really want to find people who can make forward progress on the uncertainties associated with a 3-month project, at least.
A technical interview whiteboard code or data structure gotchas or mt. fuji questions is a bad interview. It sucks that most technical interviews are bad, but that means we should fix them not remove them.
Our team "technical interviews" by having a technology discussion with the applicant. One of the first lines of question is figuring out what they are most familiar with so we can discuss that particular thing, area, library, or whatever. If a person can't discuss what they are most familiar with in the high-pressure interview, I'm not sure they can discuss something they just learned about in a team design meeting either. It's also a great way for the candidate to figure out if he wants to work with us - something that is just as important as the reverse.
Quit making technical interviews a quiz show. Quit checking off boxes on your form. Quit with BAD technical interviews. But don't remove them entirely - that's just as dumb.
It is not the technical interview that is a problem; it is the interviewer. I sit quite often on both sides of the table and I have noticed that some interviewers are eager to show off their skills more than trying to learn about the candidate. Or they are auditioning bros to go clubbing or to invite to barbecues . I have not ever cleared a tech interview when I was interviewed by guys in twenties or early thirties. I have a 100% success rate when I am interviewed by people,all races and gender , over 40.
When I am conducting interviews, I place a high degree of importance on the candidates aptitude and approach to problem solving and I usually build up a pretty good team with candidates who the bros wouldn't hire.
Some people suggest looking at candidate's github repo; this would not work for enterprise software developers and most large companies have restrictions on what code a developer can claim as his/her own and publish.
I'd say that contra to what a lot of propaganda will tell you, being experienced matters.
In interviewing a diversity of experience is especially crucial. Further, being experienced at interviewing is super important, and that sort of experience is even harder to gain than other software experience.
Before I tech screen anyone, I have a short email or phone chat with them. I ask if they program every day and how much of a typical recent day was spent programming.
If it sounds like they are currently not full-time employed as a programmer (and not programming a lot), I explain that they will probably not do well on technical tests (mine or anyone else's) if they don't practice. I recommend they just find some sample questions online and practice them -- let me know when they think they are ready.
I explain it's just like playing an instrument -- you wouldn't audition without practicing beforehand, right? I also explain that it's very hard to tell the difference between someone who is rusty and someone who is not skilled -- I want them to be at their best.
To everyone, currently programming or not, I explain what the interview will be and everything about it that I can except for the questions. I am hoping they will self-select out if they know they can't program (or talk to me about it).
I also explain that I am not looking for people who know all of the answers, but I am trying to calibrate their resume and see what it's like to collaborate. All tech screens are conducted in the language of the applicant's choice.
The meta-question I am asking them: If I tell you a bunch of requirements and some guidelines for success, will you do the work necessary to succeed.
Otherwise, in what possible high-pressure situation will you ever be expected to solve an obscure data structures problem on a whiteboard. With no syntax checking or debugger. In front of three scrutinizing neck-beards. Answer: Never.
THIS. ALL OF THIS.
I recently had a technical interview where I was asked to write some Java code that printed out every 7th number. I could not, for the life of me, figure out how to do it. And I've been a software engineer for 25 years.
I realized later that it was because I was nervous about how I could present myself in this once chance, and that made all the rest of it difficult.
Otherwise, in what possible high-pressure situation will you ever be expected to solve an obscure data structures problem on a whiteboard. With no syntax checking or debugger. In front of three scrutinizing neck-beards. Answer: Never.
THIS. ALL OF THIS.
I recently had a technical interview where I was asked to write some Java code that printed out every 7th number. I could not, for the life of me, figure out how to do it. And I've been a software engineer for 25 years.
I realized later that it was because I was nervous about how I could present myself in this once chance, and that made all the rest of it difficult.
If I ever get to lead a software team I will make part of our product open source and then ask potential new hires to complete a real task on that codebase. I think that would be the most realistic display of their coding ability and quality and at the same time they could get an idea of how the company manages and reviews code.
121 comments
[ 3.2 ms ] story [ 202 ms ] threadThe hard part about the job is
1 Getting the actual real requirements nailed down 2 Designing the system to run in the real world accounting for all those edge cases and falilure modes.
If on the other hand you are saying, "should I be evaluating product specification and requirements gathering when hiring business analysts, or product managers" then I would say of course! What else would be evaluating them on?
Similarly, there are environments where static requirements are cornerstone for success. Certain methodologies deal with this environment better than others and certain developers will be more successful than others in these environments.
Wherein the programmer plays the role of social worker. Seriously though, the skill sets are very similar. The client is typically the ultimate source of the requirements, but it's never a simple case of asking the client "what should I build? More often than not, the client doesn't know what s/he wants, wants something that will actually hurt him/her, is actually seven different people who want opposite things, wants the roses simultaneously painted white and red, etc. Which is why I find requirements collection the most challenging part of the job.
So in each case we come up with proxies for this measure. Having done the interview side of the fence a lot, my strong intuition is that nearly everything people try to measure in developer acquisition strategies either don't correlate or negatively correlate to hiring good developers.
At this point, the only advice I give people about this is figure out a repeatable, measurable process for your organization and measure every single filter you use during hiring. Test the measures on just hired applicants, applicants a few months out, and applicants a few years out. It is time consuming and fraught with problems but there is no magic bullet out of this morass.
A.) Are great in their dayjobs and don't wish to spend their free time writing Open Source. Maybe they have families. Maybe they have non-coding hobbies.
B.) Work on private repos all the time.
Whiteboarding lets me prove raw problem solving IQ without being negatively judged for lack of twitter followers or technical blog posts.
Portfolios allow devs to prove raw problem solving IQ without being negatively judged for lack of documentation or a debugger.
There are developers who aren't good at whiteboarding interviews because they don't think that way and work better when they're at a laptop and can refer to other source code and documentation and can run their code continuously as they write it.
With those devs, they may not look impressive in a whiteboard interview, but you can look at their GitHub portfolio and draw conclusions about their programming ability.
An ideal interview should be holistic and include both off-the-cuff code and polished projects.
A good judge at a whiteboard won't mark you down for uncertainty about interfaces, which partially ameliorates this, for those cases where you have a good judge...
I'm also tempted to say that needing a debugger to lay out a mostly correct solution to a small problem is weird - can you elaborate on how that might be applicable?
In any event, I agree with your ultimate conclusion - holistic is probably a better way to go.
For my current job I was provided with the specifications for a simple app to display images from a flickr rss feed and asked to code however I felt was appropriate. It was interesting and fun, and vastly less stressful than any whiteboard test.
It was also a great indicator of what working at the company was actually going to be like.
This is how I bombed a DevOps role interview at Twilio, but was picked up as a VP of Engineering elsewhere. +1.
Agree with OP. Coding under pressure is a horrible way to measure good coders, because it's an artificial construct.
For people looking for alternatives to whiteboarding or phone screens where the candidate writes code in the blind, check out https://coderpad.io
It lets you write and execute code with your interviewee in real-time, and provides a much more native programming experience easily and over the browser. Some rather large companies have started using us exclusively in the in-person technical interview, even buying Chromebooks especially for the application.
Disclaimer: I am the guy who makes CoderPad and am obviously biased.
The proper reaction to such a request would be: "So, you guys do all your work here on whiteboards? That's seems unusual, I used laptops or workstations at all my previous jobs."
I've done this -- had people talk to me about their anxiety or problems with it, and put them at ease. For some, I have recommended that they just go practice of an hour or so with a friend -- it's really not that hard to code "whiteboard" level code if you code every day and practice it a little. It's such a common tech interview style, that if you are looking for a job, it's worth working a little at it.
Frankly, in a code-editor, my expectations are much higher -- I don't even require you get any framework class or method name right (or even perfect syntax) -- but the compiler will. It gets in the way of the essence of the question -- which is more about collaboration.
This is for a screen to make sure you are a programmer. If you can't code up a four line function without an editor, there are going to be a lot of jobs you might like that you won't get. Ditto with calling strangers stupid.
I don't give a flying fuck about minor syntax errors, forgotten API, or anything a compiler can catch. If you forget an API, I either give it to you or ask you to make up something reasonable. (Chances are, I can't remember it either). But if you can't describe an algorithm to me in front of a whiteboard, I don't want you on my team. And if you can't translate it to pseudocode that resembles the language you're going to be working in, I also don't want you on my team.
I use a whiteboard at work to pseudo code a solution, and the writing looks like chicken scratch but even my co-workers get what I'm expressing to them. Then I open up a console and start typing out my idea to see if I was right. That's how I normally solve some throw away piece of code. I don't go into a manager's office and call some other random people and in detail explain to them my idea to a solution.
I also got the feeling they never saw my resume because I was asked if I had a github profile, and it's on my resume.
That said, I would say Github contributions and maybe paired with StackOverflow activity may be a good substitute for a technical interview when available.
So, what shall we do? How can we very strongly incentivize every qualified person to upload their profile, in a way that lets us curate people's actual abilities?
The suggestions this article makes, to demand githubs or complete projects done for demonstration purposes, have both been rejected. Sure, it may be a good strong signal. But then so is the signal of someone creating a complete application for your company (complete with branding) to actually A/B test on the front page, and then if it does well, for the canddiate to show up for a 3 month unpaid trial during which he or she must contribute as a full member of the team and can be dismissed at any time for no reason. I guarantee, anyone who passes that test would be a great candidate. The issue is that there are perhaps three people on the planet who would even consider applying to a company on those terms - and they're the original founders' siblings.
The problem is that it is not a reasonable burden on job applicants, and neither is a complete github profile, and neither is a complete programming project. It's just too much.
What is needed is a smaller burden that is a good, strong signal.
We think we've found one, but are not sure. (You can see it on our web site as soon as we announce.)
In the mean time, if anyone here has any breakthrough ideas in this space, we would be very interested.
Even if you don't have the right answer, you at least found the right question.
He invited me down for the technical interview. I came down a week later. His wife had just delivered a baby, so he was out. The guys in the room were a business level ex-marine who was there for god knows why, and a technical dude who clearly was pissed he had to sink to the level of doing the interview in the first place.
For the first 45 minutes, I aced everything he threw at me, and I noticed that he wouldn't drill in on anything that I clearly knew very well. He kept jumping from topic to topic, and eventually asked to do an extremely tricky SQL query, but write the code to do it in awk on the whiteboard. The job posting said nothing about awk, and I told him I didn't know awk that well. He then sensed i wasn't a command line master (the job posting said nothing about needing to be a fucking sysadmin or know advanced CL stuff) and hammered me on things that I had already said I didn't know.
2 things became clear:
He wanted to show off to his boss and make himself look like a badass while simultaneously making me look incompetent.
There was no fucking way I was going to accept a job working here if it was offered.
I didn't so much as get a phone call or an email to thank me for driving 6 hours round trip. Nothing. Which screams out to me that this company sucked.
Companies routinely pay for airfare and lodging, it would seem to only make sense if they also paid for long car trips, no?
See http://www.irs.gov/2014-Standard-Mileage-Rates-for-Business,... for more details.
After I left, I also did not get so much as a phone call or e-mail. Neither thank you nor followup. When I attempted to re-establish contact, it was like shouting into a black hole.
The feedback I received on site from the interviewers was neutral to positive, with one interviewer claiming that I was the only applicant to come up with the correct response to their abstract brainteaser.
If I take two whole days off from my existing job to come to you and indulge you in your cute little skill tests intended to prove my bona fides, you ought to have the decency to follow up.
So you're not the only person to be on the receiving end of this unacceptable behavior from interviewing companies.
In my case, I was pissed, but mainly due to opportunity cost rather than the trip itself.
By the way, you should email their founder and let them know. I doubt he/she is aware of this, and would be equally furious to know that their workers are burning bridges like this, because that's exactly what they did.
In my case, the company did actually burn a bridge. My much larger company was looking for a strategic partner with high levels of data science expertise on sensor based data. This particular company came into consideration, and I made sure to remove them from selection. Hell hath no fury like a geek scorned.
But from what I know, running into such people is pretty common in interviews.
My goal is to answer "Are you honest about what you know, excited and curious about what you don't, willing and able to learn, and driven to excel at the role that we're discussing?" Sometimes answering this requires code, but generally not a whole lot, and mostly I just use it to tease out where someone actually is vs. where they say they are. I don't want to hire a candidate who is best suited to make my problems go away tomorrow. I want, a year or three down the line, to say I built the best team possible.
Maybe I can say this because I don't think the "technical" part of what we're doing is the major challenge, though I think that most engineering teams are actually in the same boat as us. Google has far different problems than we do, so their hiring process is rightfully different than ours. And maybe I'm just getting it all wrong. We'll see what happens.
I haven't had many, but every whiteboard interview I've had has been miserable. At best I've just about managed to convey what I mean, at worst I don't even understand the question.
One particularly memorable exercise had the interviewer write out HTML on the board and ask me to write out the CSS to turn it into a dropdown menu. I didn't even know where to start, because I couldn't construct a mental model of what was going on where. Because outside of the white board interview I never have to.
A much better technical interview that I do not with death upon is one that put me in front a computer which was rigged up to a projector that the interviewer could see. We talked through my solution to an exercise as I constructed it, and I was free to switch between the browser and editor as I wanted. You know, like I might in a normal, actual working day.
An even even better interview was one that left me in a room for 45 minutes with a task. After that time I sat down with the interviewer and talked through what I did.
In my experience, being bad in either one of these interview types is not correlative to being bad at being a developer.
That's fine for startup-ish hires, but in most of those cases, isn't the candidate being sought out/referred/etc? Almost all of my work is on non-public SCM systems, and frankly, my job doesn't leave a ton of time to post to *overflow sites, and when I go home, I'm home, and spend it with my hobbies and family.
I also consider myself a really great developer. Especially in the more corporate .net world. When I review candidates, I find that a simple (at a PC, with resharper installed) coding test, with a prebuilt solution, needing only an implementation of a method done, is a great filter for competency. And after all, rough competency is the most I hope to get out of a coding test. The far more important part is always the stuff like how well I think they'll fit in a team, how their approach to solving problems in general is, etc.
Syntax memorization is no longer a measure of a good developer. All it shows is basic competence in a language.
Of course the flip side is, you can probably get by with minimal outside work. Do some code katas, have an up to date best practices simple app in github, etc. I think as long as you have something in github that looks good it will do the job, even if you're not working extra hours daily on open source projects.
Having worked under such terms, I've never felt I could participate in open source without jeopardising the project.
They don't want "You were doing exactly the same thing over there, now you'll be doing exactly the same thing over here", because they don't have the money in the budget for a real applicant.
You have two cohorts which comprise a majority of your applicants:
1. Unemployed, X% qualified. Has time.
2. Employed, 100-X% qualified. Does not have time.
X is small. If both candidates are qualified, tests for X also find how qualified they are so they can be compared.
His methods are objectively better, provided that both candidates have the same amount of free time. If you compare two unemployed engineers, you'll pick the right one. If you compare two employed engineers, you'll pick the right one.
If you compare an unemployed engineer vs an employed one, however, you'll probably pick the unemployed engineer, even if the employed one is better. Because he doesn't have as much time for your homework or to maintain a Github profile.
Thus, if you're comparing 30 engineers, half of whom are employed, you'll pick the most qualified unemployed candidate, rather than the objectively most qualified candidate.
That's why, even if the test is slightly worse, with an interview room only test you'll consistently pick the better candidate: a large, large segment of talented candidates are employed.
I would love a statistical analysis of Simpson's Paradox and technical interview methods if anyone's ever done one.
I talk to the applicant and try to work something out. I interview weekends and evenings if that's what they want. First interviews are always on the phone. Eventually, they have to come in, but by then, we are pretty sure it isn't a waste of time.
Especially if their day job sucks, which is one reason people look for new jobs.
Not too long ago I was at a job where management demanded 11½ days. Guess how well the team members could handle an interview at 8pm.
When I interview candidates, I look for a good foundation for what they will be working with like do they know the difference between class and struct and the implications of using them (day 1 kind of stuff).
Then I ask them about how they would go about solving a problem they were unfamiliar with. Google-foo is a skill that must be learned and honed. I don't care if you can spit off all sorts of acronyms, I want to know if you are capable of using common sense to solve a problem.
Last but not least, how do they stay up-to-date and relevant. Not looking for the 8 to 5 developers and not interested in those cutting-edge guys either.
Must have good understanding of the basics, principals and skills in problem solving. Telling a block of code that you have a masters in CS will NOT make the bug go away.
Source: Been interviewing supposedly senior level candidates for the last three weeks.
However, it's true that coding exercises that can be completed in an interview are arbitrary and abstract (by necessity, since you have no context.) In my opinion one should use live coding exercises to test fluency -- can you code at the level you are claiming? Debugging something might be a better test. Then pair that up with a short (short!) take home exam to test problem-solving.
The one thing that really has to go is whiteboard coding. Ask the candidate to bring a laptop or give them a used one.
OP also thinks that high-pressure situations don't occur with programming. But if you work on the web, sooner or later you will be debugging a tricky issue that's costing that company $X per hour while your all your managers up to the CEO hover behind your desk. Being able to communicate with non-technical people and inspire their confidence is also very important. Often more important than coding ability.
If only this were true. There is a whole class of people out there who are stellar face to face and not great for your specific technical/business situation.
But past a certain level, it's your job to explain your own weaknesses too. I just had a technical interview the other week and they relied way too much on definitions and whether I'd heard of this and that. I happen to do great at those but I want to be sure I can be productive (the position uses Elixir, which I'm totally unfamiliar with) so I'm suggesting they give me a take-home exam.
I have had my fair share of false positives. And probably I have been one or two times the false positive. And this is something that everyone can relate I suppose.
Sometimes you just need luck - like the gotcha question being something you have debugged soon in a language you are not that competent at.
A smart candidate will bring a notebook computer, loaded with suitable IDEs etc ready to go, without being asked to. Impresses people when you tell them to continue the interview while you write the requested code (multitasking, competent enough to spare nontrivial brain cycles answering unrelated questions).
Will some of those with opinions of this suggestion describe those opinions, so that I can better evaluate whether or not it's actually worth considering? Thanks!
However, I don't think bringing your notebook to show off some of your work is a bad idea.
Downvotes? Meh. Taking a notebook and writing the solution with sensible tools - while fielding unrelated questions from 4 interviewers - got me an offer.
Why the downvotes?
The problem is that problem solving with fancy algorithm's can even be part of the job, but are normally just a small fraction. I've never seem dumb people going incredibly well in technical interviews, that's true, but I've seem a lot of bad programmers giving great answers to those abstract problems. And sometimes, they're knowledge wasn't actually a fit.
I thought most people came to the conclusion that this is a horrible metric by which to gauge competence?
Then again...
Now us Python developers, Ruby developers, Java developers, PHP developers, and (God help you) Perl developers, let us put aside our meaningless and largely annoying quibbles to unite under a single banner. For once, let's say who cares whether Django or Rails is the superior framework and just join our voices to call for an end to the modern technical interview.
The final paragraph shows what demographic the OP belongs to, so it's hardly a surprise they think that way.
Once you're dealing with someone who has verifiable experience, then repo activity should be weighed much less (unless it's something really awesome, then chances are you're dealing with an awesome developer).
If you're critical of someone lacking github repos but they have experience, then I think it becomes a symptom of "everyone trying to stand out"-itis. It's probably best to have different solutions for different skill/experience levels.
In my technical interviews, I look to talk through specific technical problems, and they're usually real-world bugs or design issues I've worked through in my job. There's not usually a lot of code involved, and the code isn't that tricky. It doesn't usually even rely on too much specific domain knowledge. It's about being able to take a broad technical question, formalize it a little into something concrete enough to be discussed, identifying constraints on the solution, working towards a solution, reasoning about performance, and discussing tradeoffs of this approach vs. others. In short, it's about technical reasoning. I don't care if we get to the right answer by the end if we've had a fruitful technical discussion.
Code samples are a good idea, but they're as deeply problematic as interviews. Most of the code from most of the best engineers I know is closed-source, not on github. 48-hour programming projects can tell you what someone can cobble together, but they don't say anything about the person's attention to longer-term concerns around design or code quality. Moreover, being able to code up a technical solution to a 48-hour problem is like basic literacy: I expect at least that, but I really want to find people who can make forward progress on the uncertainties associated with a 3-month project, at least.
Our team "technical interviews" by having a technology discussion with the applicant. One of the first lines of question is figuring out what they are most familiar with so we can discuss that particular thing, area, library, or whatever. If a person can't discuss what they are most familiar with in the high-pressure interview, I'm not sure they can discuss something they just learned about in a team design meeting either. It's also a great way for the candidate to figure out if he wants to work with us - something that is just as important as the reverse.
Quit making technical interviews a quiz show. Quit checking off boxes on your form. Quit with BAD technical interviews. But don't remove them entirely - that's just as dumb.
*ps: github as a metric is also a bad metric.
It is not the technical interview that is a problem; it is the interviewer. I sit quite often on both sides of the table and I have noticed that some interviewers are eager to show off their skills more than trying to learn about the candidate. Or they are auditioning bros to go clubbing or to invite to barbecues . I have not ever cleared a tech interview when I was interviewed by guys in twenties or early thirties. I have a 100% success rate when I am interviewed by people,all races and gender , over 40.
When I am conducting interviews, I place a high degree of importance on the candidates aptitude and approach to problem solving and I usually build up a pretty good team with candidates who the bros wouldn't hire.
Some people suggest looking at candidate's github repo; this would not work for enterprise software developers and most large companies have restrictions on what code a developer can claim as his/her own and publish.
edit : grammar
In interviewing a diversity of experience is especially crucial. Further, being experienced at interviewing is super important, and that sort of experience is even harder to gain than other software experience.
I guess that's great for people that have a sweet Github profile...
If it sounds like they are currently not full-time employed as a programmer (and not programming a lot), I explain that they will probably not do well on technical tests (mine or anyone else's) if they don't practice. I recommend they just find some sample questions online and practice them -- let me know when they think they are ready.
I explain it's just like playing an instrument -- you wouldn't audition without practicing beforehand, right? I also explain that it's very hard to tell the difference between someone who is rusty and someone who is not skilled -- I want them to be at their best.
To everyone, currently programming or not, I explain what the interview will be and everything about it that I can except for the questions. I am hoping they will self-select out if they know they can't program (or talk to me about it).
I also explain that I am not looking for people who know all of the answers, but I am trying to calibrate their resume and see what it's like to collaborate. All tech screens are conducted in the language of the applicant's choice.
The meta-question I am asking them: If I tell you a bunch of requirements and some guidelines for success, will you do the work necessary to succeed.
THIS. ALL OF THIS.
I recently had a technical interview where I was asked to write some Java code that printed out every 7th number. I could not, for the life of me, figure out how to do it. And I've been a software engineer for 25 years.
I realized later that it was because I was nervous about how I could present myself in this once chance, and that made all the rest of it difficult.
THIS. ALL OF THIS.
I recently had a technical interview where I was asked to write some Java code that printed out every 7th number. I could not, for the life of me, figure out how to do it. And I've been a software engineer for 25 years.
I realized later that it was because I was nervous about how I could present myself in this once chance, and that made all the rest of it difficult.
Take a problem that I recently solved or trying to solve and turn it into an interview question.
Email them a coding problem that's very easy to understand.