Engineering whiteboard interviews: yay or nay? | Hacker News Reader

Engineering whiteboard interviews: yay or nay?

keyvalues.com

196 pointslynnetye359 comments

  VP of Engineering at Eaze, previously CTO at Getable and engineer at Yammer.

  No, not at all. Whiteboard interviews test one thing
  well: How well does a candidate code on a whiteboard.
  Engineers on my team never have to code on a whiteboard
  (whiteboards are really bad at running code), why would I
  make candidates do something that I don't ask the
  engineers already on my team to do?

This comes close, but I think the real issue that most of these reasons aren't hitting on is that software development is typically done asynchronously. Whether it's a whiteboard or a computer connected to a projector or TV or a live peer coding session it doesn't matter: if done in the interview, it puts the candidate on the spot in a way that they rarely (if ever) will be on the job. Developers typically are able to take time by themselves or with trusted/known colleagues to solve a problem. There are very few opportunities to get over performance anxiety related to coding in front of others that we don't know.

> Whiteboard interviews test one thing well: How well does a candidate code on a whiteboard.

What if you consider part of someone's job to be communicating concepts to people, possibly with the help of visual aids and diagrams?

I hate this concept that if it's not typing code into an editor then it's not "real work".

I absolutely communicate with my coworkers using a whiteboard and pseudocode. I reject the idea that being put on the spot is necessarily "artificial". To the contrary, I think that the number of engineering jobs where you can assume you'll never be put on the spot or have to communicate complex ideas verbally or visually is relatively low.

Now if this particular skill isn't interesting to an employer and they'd rather spend the time talking about some other candidate capability, that's a whole different story.

Ugh, communicating with a whiteboard has no overlap with developing a solution to a completely unexpected quiz (if you didn't cheat) on a whiteboard in front of someone scrutinizing your every move with your future career on the line.

Whiteboard interviews rarely test the ability to communicate on a whiteboard because the interviewer knows the answer they are looking for and the presenter does not.

If you defend whiteboard leetcode using this excuse you are deluding yourself and you are part of the problem. The only thing the modern FANG interview hires for is people that can solve algorithms problems in a silo under pressure. No checks for software engineering skills, collaboration skills, testing skills, reviewing skills, and on and on...

The whiteboard interview is why so many engineers unexpectedly suck.

> Whiteboard interviews rarely test the ability to communicate on a whiteboard because the interviewer knows the answer they are looking for and the presenter does not.

They you are asking the wrong questions. I use whiteboard for developers, and for junior developers I give them a simple task such as reverse an array, turn a string into a palindrome (and explain what it is if they don't know). If they want to be a developer they must be able to solve those simple tasks. I don't really care how, as long as they think loud. Then I use it as a starting point for enhancement discussions, perhaps stack/heap questions, recursion, assignments ect. During all this I coach them, teach unknown concepts in simple terms - this is what they can expect when they ask their mentor a question so in that sense they get a feel for us too.

Some argue that it's still not good, and that we're filtering out people that work best alone. That is true, but that is by design. We work as a team, having 10 individual developers not working together unless forced is an architectural nightmare.

Agree with most of what you said except the conclusion.

The reason many engineers unexpectedly suck is because it’s a high paying job where performance is difficult to measure let alone predict.

White boarding interviews are a symptom of the problem, not the cause of it.

I always thought it would be a good idea for the interviewee to also come up with a question and then have the interviewer try and solve it with their guidance. I imagine it may be more helpful than the other way around.

If the interviewer leaves frustrated because they couldn't solve anything and it was a mess, or the question was formulated terribly, or the interviewee was under prepared, etc etc, then it provides a lot of insight into the capabilities of the interviewee. It almost provides more insight into the qualities required of a developer than just putting them on the spot.

>I reject the idea that being put on the spot is necessarily "artificial". To the contrary, I think that the number of engineering jobs where you can assume you'll never be put on the spot or have to communicate complex ideas verbally or visually is relatively low.

I think it's safe to say that being put on the spot in front of people you don't know who are judging you is quite different from being put on the spot in a team meeting with someone from product who will watch your demo of a potential new project/idea/concept.

We do whiteboard problems, BUT we also emphasize beforehand that the interviewers in the room are there to work through the problem with them. This is meant as more of a conversation, we're less concerned about getting to the "right, fastest, or most optimized" answer. Not a hard problem, no mindgames, no syntax rules, and no complicated pre-known algorithm work other than an if and a loop. Let's just talk about some pseudocode.

We've never rejected somebody for having a "bad" answer, because they were able to at least talk about what they're doing. We have had people flat out refuse to even try, and that seems like a pretty big red flag.

Every time an interviewer has told me something like this, they then nitpick syntax and appear to be primarily concerned with "does my whiteboard code compile" sorts of problems. And getting stuck / asking for help feels like I get docked for getting stuck. Same with less optimized. So it's hard to trust such an explanation - clearly my interview would be better if I came up with the perfectly optimized solution, or gave a lecture on the options rather than trying to discuss options.

So it seems to me entirely reasonable to still be nervous.

Mind you, I don't have a better answer, but I don't think just explaining that suboptimal is ok and it's a discussion really helps.

That sounds like a much better way of handling a whiteboard problem.

In my most recent on-site interviews, my interview would be in a room with full floor to ceiling glass windows, the next interviewer would come in, ask one or two personal questions, and then either cut me off after a couple minutes or just simply say "we're going to move on to the whiteboard now, do this". Meanwhile people are walking by staring into the room, looking at the whiteboard, etc.

Which is just too rigid, IMO.

When I interview a candidate, I want to know that they'll be able to work together with myself and my current teammates as a team. So I would rather do something similar to what you described.

You are asking the candidate to compartmentalize their thoughts to solve a toy problem during a period of time when they are attempting to put various nascent clues about your organization together in order to form a better understanding and to identify how they can provide value / fit into the flow.

Fizz-buzz tests have their place, but I think an onsite interview is past that point. In general I have found that white board exercises don't allow me to either understand or convey how I can apply my expertise to provide value for the organization.

As a candidate, to me, that sounds a lot like, "we're going to do the same things we would normally do in a technical interview, but we're going to evaluate you based on random, arbitrary, subjective criteria that you will never know about rather than your actual level of skill."

I'll bite: yes, if there's rapport and the candidate doesn't feel caught off-guard or needlessly pressured. If it's whiteboarding like you'd whiteboard while on the job with your coworkers, that's a good idea that can tell you a lot about a candidate. Unfortunately, in my experience, that doesn't seem to happen very often. A lot of people seem to treat it as an adversarial process or expect, for lack of a better description, a "performance" over contrived problems.

I've had my share of the second kind, which were without exception some of the most unpleasant interviews I've ever gone on.

> We do whiteboard problems, BUT we also emphasize beforehand that the interviewers in the room are there to work through the problem with them.

Except at the end of the day this is an interview, you are judging their answers and evaluating their performance. I have been part of white boarding sessions like this and it does not reduce anxiety and still feels completely removed from what I actually do on a daily basis.

> We do whiteboard problems, BUT we also emphasize beforehand that the interviewers in the room are there to work through the problem with them. This is meant as more of a conversation, we're less concerned about getting to the "right, fastest, or most optimized" answer. Not a hard problem, no mindgames, no syntax rules, and no complicated pre-known algorithm work other than an if and a loop. Let's just talk about some pseudocode.

If only all interviewers were like you.

Well, when you first start your job, you won't really know the people you are working with so I don't see how it is that much different.

Incredibly different.

I highly doubt a normal engineer is going to be giving a demo to their team with high stakes in the first three or so months. I understand if they're possibly a senior engineer, architect, or lead of some sort because they were probably hired with a plan to spearhead a specific project.

I'd say in the average case a new engineer will have enough time to at least break the ice with their new team before being put into a high pressure situation.

I communicate my ideas using a whiteboard too. But I never produce those ideas using a whiteboard in front of my coworkers under pressure of being fired.

Well this is hardly the fault of white board interviews. That’s just interviews. Any interview (regardless of the style) is going to be in front of “coworkers” and “under pressure of being fired.”

That’s just interviews.

Nope. There was a great submission a few years ago about a guy who applied to two categories of jobs: management and software development.

In the SD roles, everything was oriented toward rejecting him. Any possible "red flag".

In the management roles, the interview was focused on finding areas where he could contribute to the company.

Whiteboard interviews are a problem in and of themselves, and the propagate terrible attitudes into the rest of the interviewing process.

The "produce the ideas during the interview" part is the fault of whiteboard interviews. There's nothing that says an interview should cover brand-new material as opposed to material the candidate is supposed to be familiar with.

If communicating ideas is part of your job, I'd bet any amount that you decide what ideas you want to communicate well before you do the actual presentation (or other physical delivery of the ideas).

My experience agrees with this. I am a different, significantly less comfortable person to be around when strangers are asking me so many questions about myself. I think it’s just about the worst way to get to know me.

When I get hired this way, I always feel a little bit of guilt because I know they actually hired somebody else.

What can be done? I’m not sure, but I can say one thing: my level of discomfort is multiplied by the number of strangers in the room. Does this ever get considered with interviews?

I am far more comfortable coding in front of 1 or 2 people who each give me a little bit of background about their coding experience; just so I know. If they are experienced engineers, it’s probably not going to change what I say aloud, but it just makes me more comfortable. I guess more overlap of technologies in our background does help. It’s nice to not feel pressure of worrying if my solutions reflect general programming conventions enough to be language agnostic. I have never really used Java, for example.

Then test that. Having someone code on a whiteboard, even if they are explaining what they are doing while they go along, is simply not the same thing as being good at using on the fly visual aides while communicating. In any case, that isn’t coding either.

You have 40 minutes to make a slide deck that explains bloomfilters and 5 minutes to present in front of the director of engineering.

That is the same as saying who can code the fastest rather than who does the best job. But at the PhD level there will be a presentation given regardless.

It's bizarre to me that you think that whiteboard coding tests "communication", in any way. It's not like a presentation, or anything. It's one-sided combat where someone with a secret tries to get someone who doesn't know the secret to regurgitate the secret, on the spot, while pretending that s/he didn't memorize the secret in advance while cramming a great big "Cracking the Programmer Secret" book to prepare for the entire silly exercise in Kabuki theater.

If you want to test communication skills amongst programmers, I dunno...ask them to write something in coherent English. Or here's a crazy thought: ask them to document some code. I guarantee that 80% of working "rockstar coders" will fail (but not before whining, crankily, about how unfair it is that you would actually make them do such a useless thing, since, y'know...never actually documents anything IRL, dude).

Programmers like whiteboarding because it lets them believe that interviews can be reduced to purely objective functions, and because GooAmaSoftBook does it. They're too scared to deviate from the pack, lest people think they aren't as "elite" as everyone else.

> It's bizarre to me that you think that whiteboard coding tests "communication", in any way. It's not like a presentation, or anything.

I feel like we're probably at an impasse if I can't convince you that a whiteboard is a decent medium to communicate ideas around datastructures and algorithms, but I appreciate your point of view.

I can only say my experience, which I hope you will take into account as one anecdote. I don't read books about "cracking the coding interview" or do leetcode or hackerrank. I left school 14 years ago or so my stash of cs trivia/secrets/gotcha isn't particularly full. I've done whiteboard interviews where I come up with at best a naive solution.

And yet I've received offers for fairly senior engineering positions at Amazon and Twitter and (hopefully tomorrow) from Google. Most of the whiteboard interview isn't even around the code, although that's a small part. Most of it is analyzing the problem, discussing constraints, discussing tradeoffs, walking through data structure manipulations, drawing arrows and boxes, that kind of stuff. Some code, maybe 30% of the interview. I just keep having this experience where nobody wants to play the gotcha game, they want to know how well I can communicate while solving a problem and they think they get that information out of the interview (I agree with them).

That experience makes it hard for me to understand a viewpoint that believes that whiteboard interviews are about memorizing secrets in advance.

> I guarantee that 80% of working "rockstar coders" will fail (but not before whining, crankily, about how unfair it is that you would actually make them do such a useless thing, since, y'know...never actually documents anything IRL, dude).

Are you sure you're not just replacing one stereotype with another here? Do you think it's fair to say that 80% of people who do well at whiteboard coding value documenting their software so little that they'd be fail at it while disparaging the concept?

There are plenty of things one can gauge candidates on via the whiteboard. System design, application design, psuedocoding implementation (or real coding, if necessary - it should be natural flowing from psuedocode to implementation though), problem solving with visuals to denote intuition...the possibilities are vast.

There seems to be a bunch of ingenuous assertions here that is characteristic of someone who doesn't think one should have to prove oneself. I'm at a FAANG and we don't even ask anything crazy hard for algorithms/data structures like Google or Facebook does, but our interviews still manage to be intense for technical and soft skills - they will challenge even strong candidates and one cannot pass the interview without displaying strong technical acumen or communication skills.

There is so much more to technical assessment than just coding. A whiteboard is just a standard medium for communicating ideas in a collaborative manner that still works the best for any sort of deep collaborative design when compared to any other medium so far.

or do a code review

The point is that the interview style in question overemphasizes on-the-spot responses. How is prowess in a game of "gotcha" an important indicator of engineering skill?

I tend to agree with (what I think is) the sentiment of some of the respondents in the article: if your whiteboard interview is a game of gotcha, then the problem is your interviewer, not the whiteboard.

I think the complaint that programmers pretty much never have to think about code under pressure is a fair criticism. The whiteboard really puts you on the spot. But if it's your company and your hiring managers are playing a game of gotcha on the whiteboard, you need better interviewers.

State dependent learning is a bit of a factor too. I can count on one hand the number of times I’ve presented anything on a whiteboard at work, yet just about every company asks these questions during interviews. A screenshare would be a better representation of on-the-spot problem solving, if that’s what you’re testing, because I’m at least in an environment I typically work in.

Whiteboard coding interviews are one thing that I have to go out of my way to practice for and get better at over time when interviewing (ie over the course of a few failed interviews). I don’t feel more skilled or smarter by the end of the set of interviews, but I inevitably do better at these kinds of problems. Should the interview be testing my competency at day-to-day work skills or how well I’ve practiced my interview skills, because whiteboard coding problems only achieve the latter.

> How is prowess in a game of "gotcha" an important indicator of engineering skill?

It's not; I couldn't agree more.

In real life when using engineering "skill", not only am I not anxious/stressed by trying to write on a whiteboard, but I'm using my laptop, with my editor, as well as access to man pages, docs, and google to look things up.

In addition to the overemphasis of performance responses, I think "how much do you know from memory with 10 seconds of thinking" vs "how much do you know with 10 seconds of google" is useless. The only fairly complex datastructures I can perfectly describe from memory are the ones which I used most recently...

"Gotcha" is about being adversarial. An adversarial attitude and a high-pressure situation are different concepts entirely.

> What if you consider part of someone's job to be communicating concepts to people, possibly with the help of visual aids and diagrams?

Perhaps it would be more effective to have the candidate whiteboard a concept that they are already familiar with, be it a high-level engineering principle or a system/solution they have built in the past. Attempting to solve a problem you have just been presented with AND communicating the solution effectively is a big ask.

That's an excellent idea. I'm in no way saying the existing method is perfect. Just that some of the things it tests around communication and being put on the spot and analyzing a problem in a way that is understandable to the rest of the room is actually a really good engineering skill. There can be other great ways to measure those skills.

I know my opinion is unpopular, but I sometimes do think I've figured out something other people miss. I think when it comes to whiteboard inteviews, candidates are often playing the wrong game. They think it's about gotcha questions and they think they fail because they didn't leetcode hard enough. I don't think that's true, they are just trying to game the thing that's easy to measure.

In my experience with the "terrible" FANG companies, it's not about gotcha questions. I get the offers even though I don't often find a non-naive solution. The people I'm in the room with really do want to see my thought process and they really do want me to communicate the tradeoffs with them. People don't fail the gotcha questions because they don't know trivia or because they forgot a detail from their CS classes. They fail the whiteboard interview because they see an unfamiliar question and say: "I don't know that trivia" instead of drawing out possible solutions and having a conversation with their interviewers.

That's an awesome idea -- as someone who rabidly hates whiteboard interviews, to the extent that I'm looking at the responses in this article as a note of who to consider applying with next, I would love the challenge of "explain a complex concept you're already familiar with". I've never gotten that in an interview before.

I try to start there, but most candidates act shockingly uninterested in going into details of their past work. I've directly asked about what the interesting parts of a project were to them and gotten things like "I had to learn ES6 and I hadn't used much JS recently" without much elaboration able to be teased out.

If you can't give me good examples from your past, I'm going to throw my own questions at you. I tend not to do coding on the whiteboard, though. Lots more boxes and lines and schema and system interaction-y stuff.

even then, this is not what's being tested with the whiteboard.

thinking and presenting are two separate skills, and, in my opinion, can rarely be executed at the same time.

So if you're looking for someone who can execute both at the same time, whiteboard coding is ok?

If your looking for someone who can think, code and present at the same time, yes.

Coding on a whiteboard while being questioned by two or more interviewers isn’t really a true representation of a work situation. It’s always more stressful and some people thrive on it, some people don’t. Im not sure what the better alternative is to demonstrate verba/technical communication skills however. The problem is, it’s hard to figure out if someone is a good fit with just a couple of hours of vigorous questioning. Maybe that’s the problem. I’d say I’ve worked with plenty of colleagues where it took me weeks or months to fully appreciate how extremely valuable they are to the team.

I agree with the sentiment but find most of my whiteboard time concerned with a much higher level type of problem solving than writing lines of code or absolutely correct algorithms. I also mostly do it with colleagues I know on shared problems we are working cooperatively to solve. Where I’m using one with strangers I’m much more often working as an expert in an explanatory manner. It feels quite remote to my experience of solving an unknown problem in an interview situation.

Yes!

What you're testing is the ability to discuss a problem with your peers. That is the critical thing.

Do the whiteboard but don't make the coding the standard. Don't get too hung up on the details of the problem.

Make the problem easy but fuzzily specified.

VPE at Eaze here -

Clear and respectful communication is one of the most important skills for an engineer to have.

If you read the second half of my answer in the article I go into the communication side of things. In my experience there are better ways to evaluate a candidate's emotional intelligence and social skills than asking them to code in front of someone.

I'd say you should stick to your guns - I like your approach. Interviews are inherently adversarial: "are you good enough for us to hire you?", "are you better than the other 5-10-1000 candidates". As a result there's a built-in bias which makes whiteboarding a bad fit.

I've written code on white boards at work, and I've discussed about code on white board with colleagues or managers. It's not the same thing as during interviews. The schedule is never as tight, scrutiny is way laxer and the overall mood is completely different when I'm working with colleagues or even bosses.

I kind of understand why the BigCo's do it (they need a super rigid filter since they have so many candidates), but in their case they could select for people with the best pink leotard on interview day and they'd still get decent programmers, because their companies are in such high demand and they generally already control their markets (software, natural monopolies, etc.) and can afford to pay a ton so the competition would be cutthroat anyway.

>> What if you consider part of someone's job to be communicating concepts to people, possibly with the help of visual aids and diagrams?

In that case I would recommend reading the candidate's resume beforehand, look for prior experience publishing, presenting, or teaching and ask them about it.

Never seen anyone write code on a whiteboard at work.

Do write lots of sequence diagrams, block diagrams, ui sketches, arch diagrams and very occasionally someone might draw a data structure.

This. Plus network diagrams, message passing sequences, and first drafts of checklists.

I agree, and as an employer I find a happy medium is to test on a whiteboard, but be completely okay writing filler, a comment saying "code for this goes here", etc. So long as they're building the correct general approach, that's all I need to see.

Whiteboard interviews where the interviewer says "it's actually `.toLowerCase()`, not `.lowerCase()`, you fool!", or "oops! You forgot to `position: relative;` the containing element that you `position: absolute;`d below!". you're not testing anything useful. Maybe if we were still coding everything on punchcards and using massive user manuals, but docking applicants for things their editor would highlight or one ctrl-r on the page would find is ridiculous.

What if you consider part of someone's job to be communicating concepts to people, possibly with the help of visual aids and diagrams?

This is incredibly common, and you tell the candidate in advance that they will be expected to give an n-minute presentation on a subject of their choosing (need not even be tech) so they can prepare.

Tell me this: when did a mob last ambush you at your desk and demand that you invert a red-black tree on a whiteboard that they happen to have with them right now? I'm guessing that never happens where you work. If it does then fair enough :-)

I think now a days, selection through the whiteboard process simply comes first to how much you grind leet code and the like.

Ironically this leads to min- maxxers who ignore a good chunk of their courses simply to master leetcode. My smartest peers are ones who work on interesting projects and research, thus never finding time to grind interview questions.

Then there are also companies that expect your white board code to compile and have perfect syntax, which imo is very unreasonable.

I don’t think there’s much criticism of those types of whiteboard interviews. I’ve had programmer interviews that involved sketching out a system architecture or database scheme on a whiteboard, which I think is perfectly reasonable.

It's one thing to do it with colleagues for whom you have rapport and the risk of saying something wrong is low. It's a different thing when doing it around total strangers and a new job is on the line.

How about testing architecture (and implicitly communication) in front of a whiteboard and programming behind a keyboard?

VPE at Eaze here (the original author of the quote above) -

You're 100% correct. We give our candidates the choice between working through a problem at the office (on the spot as you say) or doing a take home project asynchronously. Not everyone has the time to do homework, but not everyone wants to do an in person coding interview either. We try to stay flexible.

Thank you for the response. I appreciated your take on the situation in the original article and I am glad to hear that you accommodate developers like me that don’t perform well for strangers on the spot. This type of thing really pushes a company up the ranks if I’m interviewing with them.

That's great to hear. I love getting feedback about this stuff, positive or negative! We're constantly evolving our process.

I think this is a great approach. It keeps the company flexible to hire a broad array of talent under differing circumstances while also letting the interviewee play to their strengths.

How do you compare the resulting apples & oranges? I can do a lot more at home (especially if I cheat on time, which everyone who really wants your job and can afford to do will do)

We're less concerned with the resulting code than we are about your ability to explain the code that you wrote and your ability to prove to us that you have good judgement when making technical decisions.

Also, we're growing fast enough that we're rarely comparing candidates. If we can hire both the apple and the orange (assuming both candidates are evaluated to be good for the role), we will.

I'm actively looking/interviewing and really like the "cut of your jib". (both your answers here, plenty of aspects of Eaze as a company, the prospect of moving back to SF...)

For both the sr backend and devops positions, how important to the roles is familiarity/experience with your technical stack? I might have some of the skill set/experience you're looking for -- sr backend engineer, sr systems engineer, devops are all positions I've held -- but honestly I don't know JS, .NET, or Chef. Bluntly, I don't want to waste your (or my) time. ;)

<tiny>Also, I wish HN supported private replies</>

Minor kvetch: quoting the way you did it makes long lines very difficult to read. I usually quote like this:

> I am a quote

This makes the quotation doubly-distinct, and still readable on mobile devices.

===

To the actual point I wanted to make: plenty of engineers at companies code on a whiteboard. They schedule a meeting, grab a room, and talk things out, while making notes, diagrams and sometimes even actually writing out some code on a whiteboard.

I hope nobody asks people to write compilable code on a whiteboard, but asking a person to talk through a problem and jot down some pseudocode doesn't seem like an awful thing to me. Communication is a pretty critical part of our job.

“Jotting down some code on a whiteboard” with a colleague is a much lower stress scenario than deriving and writing down, under a time limit, some DP algorithm you might not have used since college, and proving its correctness and run time to some stranger who may be barely out of college themselves.

I don't think people should play CS-quizshow during whiteboard interviews. Being asked to implement a breadth-first search, or whatever, is kind of annoying - especially since there's effectively one correct answer. At best, you're testing my memory and whether I read the right interview book.

I had to do that at my last interview. I got through it ok-ish. The more interesting question was a more general system design question - how would I architect a system to do X, under constraints Y and Z. What would I do if a new constraint came up? Ok, now how would I make it more resilient? There, the whiteboards is just a tool I can use, not the primary focus.

Surely not BFS. That's something you can figure out from first principles and knowledge of some data structures like stacks and queues, but I'd expect knowledge of those data structures anyway.

Interviews are necessarily more stressful than typical work. That's not just a property of whiteboard coding.

More stressful, sure. For many people, whiteboard coding is disproportionately more stressful than typical work, and that’s the problem.

"Interviews are necessarily more stressful than typical work. That's not just a property of whiteboard coding."

My sister, a pediatric ER nurse would disagree with you. The interviews are largely behavioral and are a breeze. No dummy is wheeled in with head trauma, random "new" diseases aren't invented and asked to be treated, etc. The simple fact of the matter is if hospitals hired in the same way, they would have no staff.

As an embedded systems engineer, I disagree. I have never been asked to design a circuit, or even pseudocode firmware, on a whiteboard. In the electronics manufacturing industry, interviews consist of talking, mostly about projects I've worked on before and projects they've done before. Occasionally someone will give me the same written test they give technicians, but those tend to be pretty easy.

> some DP algorithm you might not have used since college, and proving its correctness and run time to some stranger who may be barely out of college themselves.

I'm not opposed to white-boarding, but to reinforce this point I'll note that I would have done much better on some of these problems when I was just out of college than I would today, because they were all fresh in my mind.

I assure you I wasn't a better programmer or employee then.

"Whiteboard interview" typically does not mean "sketch a general design of a system", or the other things whiteboards get actually used for on the job. It means "here's an algorithm challenge, solve it by writing the code on the whiteboard, and we'll flunk you if you make any mistakes".

Not always a whiteboard, either; one of the best engineers I ever knew flunked a Google interview because of a situation where he had to write a bash script and read it over the phone to the interviewer, who apparently didn't transcribe it correctly on the other end. But "they only hire the best", you know, and their practices are rock-solid and proven.

Oof, I've had that experience during a technical phone screen with a different "hires only the best" company. I was asked to write (over the phone) a trivial statistical algorithm and started to describe the algorithm: "Function F returns a double and has two parameters, pointer to the start of the double array P and integer N for length of array." Apparently on the other end of the line was a human compiler that kept rejecting my input and preferred "double F open parens double star P comma int n close parens"!

This is hilarious! Such a strong indicator that you, as a candidate, should reject the employer.

Yes, it was the final straw and led to me rejecting that industry as a whole :)

What are you doing now?

Makin' games, burnin' savings.

https://www.youtube.com/watch?v=V7MXfg2wt28&list=PLqJcvcLs4C...

Hah! I love the AXBY code-entering minigame :P This is pretty neat so far.

I, too, have a side project I'd love to quit my job to work on, but I think it's less monetizable than an actual game.

Thanks! I really enjoy coming up with the themed minigames. The next one planned is a Kanban board minigame ... think of a cross between Tetris and the boardgame Patchwork.

Ha! I did that for a year...and thinking about doing it again. I mean, selling my home and moving to say, costa rica, i could stretch my dollar, dev my heart out, and watch all the sunsets my heart desires

I've worked at quite a lot of companies. Many of us will, occasionally, grab a room / whiteboard to go over something but 99% of the time it's a sort of flow or UX type thing.

I've _never_ seen someone, or have asked to participate in, some sort of whiteboard coding thing. Even with pseudo code. That just doesn't seem helpful to me, at all.

Agreed; when there is code talking, we share screen on a projector/flatscreen and do it that way. How is writing code on a whiteboard good for anyone? Cumbersome and not reusable.

I don't doubt lots of companies do this, but I skeptical about the coding on a whiteboard. They talk in person, sketch out diagrams, highlight important aspects - why? because these things are all easier on a fluid analog medium like a whiteboard or piece of paper.

Actual Code? Not a chance. Portability, Durability, Efficiency - It's the worst of all worlds.

Thank you for the quote pointed. I will remember that for next time! As to you’re point, I think others have covered it well, but I will add my agreement that communicating in an environment where you are comfortable and familiar with the topics ahead of time and won’t have your career prospects put in the line is much easier. Some people like myself simply can’t prepare for an infinite number of possible whiteboard scenarios and look competent.

I've aced every whiteboarding interview when the problem was the same or similar to some problem I solved before, and I failed pretty much every interview when the problem was new to me (except when the problem was trivial, ala fizzbuzz). I think that proves how whiteboarding just tests whiteboarding skills and nothing else.

A team I was on at Google (XLA, the compiler for TPUs) used whiteboards almost daily to express algorithms to each other. Definitely helpful for us, but I can see that just being a single team's dynamic.

No comment on it being an effective interview process.

Really? I have multiple in person discussions per week about how we are going to tackle some problem where everyone chips in and those are just the weekly problems we are working towards. There are also the much larger design docs and while some of the feedback on those happens asynchronously, a lot also happens in real time during meeting reviews. Oh, and both of those problems are decently well covered by your standard interviews (the former generally being data structure/algorithm design, the latter generally being system design).

It's funny, when I saw the title I immediately thought "Hell yeah, whiteboard interviews are absolutely necessary". As I read the article, I realized it meant actually making people code on whiteboards, which is absurd. I've always used whiteboards for high level design and architecture questions, and there mostly to aid the candidate in organizing their thoughts. If I want to see someone's code, I'll ask them to provide a code sample.

I guess it depends on the job, but for outage response, being able to understand a problem, and update or write new code in a reasonable amount of time is kind of a useful skill. Discussing it on a board with others can be useful, if the code isn't complex. Of course, that also depends on the environment; if you're in an environment can't actually push any code, even emergency fixes, quickly, then it doesn't help a lot.

Do you design clever search algorithms during outages?

If the existing search algorithm is the cause of the outage, maybe. If I a clever search algorithm will help me filter the data so I can reduce the scope of the outage, yes.

Are there no existing tools to search through data?

When's the next 50% off sale?

There's some logic to that. It's the one of the same reasons we have high stakes exams in school: to show what the individual can do under pressure. In an interview setting you can't simulate what it's like to be at the end of a high-intensity sprint, but you can ask the candidate to answer a tough question on a whiteboard and (ideally) get a notion of how they work when the going gets tough.

As a former air traffic controller and now data scientist, nothing in this career field is ever done under any real pressure.

tremendous respect to air traffic controllers. i cannot imagine the level of professionalism and responsibility these people must have! no amount of monetary compensation is enough, in my opinion.

Thanks. While everything you said is true, its also a hell of rush. Hardest part of changing careers ... no rush with a daily dose of dopamine.

But what about that one time on Swordfish!? /s

The pressure of a whiteboard interview has nothing in common with the pressure of a short deadline. It's like testing people for Tour de France by putting them on a unicycle.

Now there's a prologue I'd actually watch.

And what is it measuring exactly? In real life you don’t get the score. You either did it or not. Nobody cares about how you did it, only answer matters.

And interviews are not pressuring, they are simple. No hard desisions, no anxiety whether it is going to work. You in and 5 hours later you out. Lame.

unless the job requires whiteboard coding, you are not testing for the job.

an advice to candidate: when you've been asked to do whiteboard coding - run away. this is a negative signal.

All of these interviewing "tools" (or tricks) attempt to be time/cost efficient proxies for doing the damn job, and they all suck.

* Whiteboard coding is awesome for software development shops that don't actually own computers or use punch cards to load programs.

* Shared coding environments with a time limit works well when you want to double screen for someone who can also diffuse suspicious packages that arrive at the office

* Riddles and puzzles are good when your workplace has a chicken coop, an office fox and you can't figure out how everyone can go for lunch without leaving the fox alone.

* Behavioral interviews are appreciated by candidates who took Psychology 100 back in school for an easy A.

* Take-home tests go over well with the huge population of talented software developers who can't find a job and have loads of time they want to spend decoding the operational cost of a bubble sort

The only thing I've seen that's at all realistic and effective is a very short, __paid__ project. I did a two-day one for Indeed (ironically one of the worst promoters of all this bullshit) and a 4-hour one for my current employer. The payment doesn't even have to be market, it's more important as a signal to candidates that the company values your time.

I ask for 3 things from perspective employers:

1. value my time like you value your own (both the number and composition of your interview steps)

2. keep me updated as to where we're at in the process and when the next stage/decision will be made

3. Move forward in the process in a timely manner. It should not take more than 2 weeks from when you initially contact me to the process concludes.

I've never gotten more than two of the above from a single organization, but I will someday, and I'm betting that a company that treats potential employees that well will treat actual employees better as well.

Riddles are the worst, I was recently asked the "You have a 5 litre and a 3 litre container, how do you get 4 litres" question (never heard it before).

It's so simple to figure out if you're on your own and just play around with the idea. When asked on the spot I just kept thinking "hey these people want an answer NOW, don't make them wait" while feeling their stare, and couldn't figure it out without help.

I guess if you are old enough you'll know that one from the movie. Doesn't mean you remember how they solved it (because what did I care at the time?). Nothing to do with programming unless someone asks you to write a short program to figure out the most efficient (least steps) on how to do it. Behind a computer.

Yup. Bruce Willis and Samuel Jackson were able to figure that one out. It can't be that hard. They were under a lot of pressure as well.

I have found the best course is to treat the interview as a situation where a colleague has come to me with a problem that we are hashing out together. I feel a lot less pressure with that mindset, though I do get thrown off by interviewers who take an adversarial approach. But I find this an acceptable loss.

{

t0:{[3,5],[0,0]},

t1:{[3,5],[0,5]},

t2:{[3,5],[3,2]},

t3:{[3,5],[0,2]},

t4:{[3,5],[2,0]},

t5:{[3,5],[2,5]},

t6:{[3,5],[3,4]},

}

Thinking about making this an algorithm, are we forced to use recursion?

I agree with much of what you say, heck, nearly all of it -- but it's not reasonable to expect, or even ask for a _very_ short paid project. I see no reason that a possible employer should pay actual money for some useless task.

If it's very short, then almost by definition it's useless.

That being said I've had great results from doing, say, a tiny consulting project, and ending up with full time employment. The best jobs of my life have followed the pattern of "am either offered, or I manage to ask/negotiate for a small consulting project" -> "before the project actually ends, am brought on full time".

> I see no reason that a possible employer should pay actual money for some useless task.

My time is valuable. I see no reason to work on a useless task for free.

So is their time. Takes candidate time to do, takes employer time to test/evaluate. Both parties are risking wasted time. You're wasting time driving to interviews and answering calls etc etc. Its the cost of doing business.

How did you follow "take-home tests suck" with "two day paid projects are great"? Sure you're at least compensated for your time but most employment agreements almost certainly block this (unless you're a contractor), and that's assuming I'm willing to use 2 vacation days to maybe get an offer.

Following your style of writing:

* Paid projects are appreciated by candidates who are currently unemployed.

The issue with a paid project is you can't legally do it with people on work visas, which is a large part of your interview pool.

They are all imperfect proxies due to various constraints.

What? Just let the programmer to it in his homeland. Easy. Just as outsourcing.

I wish this project approuch was more common. It would probably be cheaper than wasting time on multiple days interviewing reality shows where people are voted out one by one like some companies do it.

It doesn't work when they are already working in the foreign country. They aren't going to travel 12 hours to do a homework assignment for you legally.

A regular viewer of my daily Twitch programming stream (link is in my profile) asked me to solve a problem he was asked to whiteboard. He wasn't asked to draw diagrams or write pseudo-code so they could learn how well he communicated. He was asked to write the code and it was heavily implied that it must compile and run -- even though whiteboards don't do that.

Here's my attempt: https://youtu.be/6LHqrxrC6Uo

I solved the problem, with an IDE and a compiler. I communicate during the entire exercise. It takes me over 2 hours. Granted, I debug it, so that takes extra time. (Can't debug on a whiteboard!) I had to take one break. On top of this, I've solved this kind of problem before and it still took me more than the 45 minutes you're likely to get in a whiteboard interview.

And I try to cover all of the "hot topic" points I know the interviewer is thinking about: size .vs. time tradeoffs, iterative .vs. recursive, etc.

If I had to solve this problem on a whiteboard, in 45-60 minutes, with the bar being "it must run"....I would fail.

I guess that makes me a poor programmer. Instead, I'm a grumpy old man who is really tired of all this fraternity hazing bullshit.

I feel like you're vastly overcomplicating the whole thing. Granted, I don't know if there were any parameters specified other than "Find the largest connected segment in a 2D grid".

I had to solve almost the exact same problem very recently. I find the best way to start is to just implement the easiest solution you can think of, such as the array approach you mentioned in the beginning. I did that in 5 - 10 minutes. We spent the rest of the interview optimizing and fleshing out the problem (e.g. let's not use an array, but there is only one "color").

Overthinking the problem straight from the beginning and potentially being unable to provide a working solution at all is probably the worst thing you can do. I actually had that happen in a previous stage of the aforemention interview, and had to "reset" by going back and implementing the simple solution before optimizing it.

Kudos on being able to "reset"! That's probably the hardest part, fighting against the sunk cost of a snowballing probably wrong answer when you have such limited time.

I got an impression that just specifying the input data took almost 1.5 hours, because of C++ and a decision to have it as a graph, not as a matrix. If an interview is in Python, one would write something like ``grid = [[1,0,0], [1,2,0], [0,2,2]]``, and be done with that part.

Probably you got exhausted after doing all these preparations and debugging issues not related directly to the algorithm. I'm not sure the final solution is correct, it is not tested on a graph with multiple connected segments of the same color. It looks more like a convoluted way to count colors globally, discarding standalone colors, if I understood your solution properly.

Thats the fun part 'to me' about mastery of cpp.

Sure you COULD use a graph, but using a different code abstraction to represent the idea of a graph is much more pragmattic and maintainable, of course that varies with application.

Im reentering interviewing phase right now, and its funny revisiting these problems with a new lense. The mind is very plastic and its exciting, so far as things to come.

Total nerd out moment :v

``vector<vector<int>> grid{{1,1,0},{1,2,0},{0,2,2}};`` in C++

I think it's trivial to code an optimal solution (and discuss tradeoffs) in 45 minutes on a whiteboard. It's just BFS or DFS.

Of course most people prefer to interview in a language which is less verbose than C++, which saves time.

Perhaps it is trivial, but what's the point? If you can reason what algo to use and reason how to determine which one will be the most efficient of the two without testing them both (or explain the abstraction which allows you to test them both without significant overhead in code), it should be enough.

In theory there should be no difference between theory and practice, but in practice there is.

I skipped around on the video. You'd literally have 1/4 the code if you just coded a graph correctly.

You also keep a "counted" and "visited" variable separately for no reason. There is no reason to visit a node and not count it, and you can't count a node you didn't visit. Whatever optimization you think you are getting surely isn't worth it.

It makes you someone that can't listen. The most important part of communication is listening. When the interviewer asks you to whiteboard a problem, they aren't asking you to "solve it". Obviously, you only have a whiteboard and you only have 15 minutes max. You might throw out some comment about size vs. time (eg) but make the decision and move on, quickly. The interviewer might decide to latch on but if not, you have to cover this in 15 minutes so you limit the depth so as to do that. Stub out all the deep functionality and then dig into each part of it as time permits and as the interviewer leads you in those directions.

As an experienced and good programmer candidate, the most important part of an interview is for you to evaluate the company. So the whiteboard is a great way to see if they understand what can be accomplished in 15 minutes on the whiteboard.

If they don't get it, you can end the interview short. Time saved today as well as the next several years working for a shite company.

I love the whiteboard interview.

At what point did I ever imply I don't listen?

I explicitly described the conditions of the original interview. The interviewer asked the candidate to "solve it" and it had to run. What part of that wasn't clear?

I also find it interesting that you countered with a completely different kind of whiteboard scenario. I agree that the scenario you describe is fine. I would enjoy those kinds of interviews all day and twice on Sunday. I'm not railing against an honest discussion with a potential employer. But that's not what happens most of the time, in my experience.

Also, the tone of your reply is that you reject the employer if they don't meet your definition of reasonable. That's great, but not everyone has that luxury.

The scenario nybblesio describes ("asked to write the code" with expectation it would be able to run) is not the one you described ("they aren't asking you to "solve it").

I think an under-tested aspect of software engineering is the ability to research solutions on the internet.

Just give people a search bar and ask them to solve a problem outside their comfort zone. See what terms they use and how quickly they can arrive at a useful link. No pseudo-code or architectural diagrams needed. Just a link with a plausible solution. Because that's how most learning on the job actually happens.

I agree. I interviewed people who optimized for white boarding interviewed and, as such, could knock up CS algorithms which you can trivially import from libraries all I wanted. But when asking them a problem they did not encounter before and Google, they got very stuck. While, every time, there was a way of writing the query to Google that had the answer in the top 3 (usually #1). People waste crazy time on reinventing the wheel.

Anecdotal: we had some problem to solve which required qua a lot of code to solve and only old sites with broken links came up in Google for the guy solving it so he decided to implement it himself. When I asked why that task was taking so long he showed me what he was implementing. The code he was writing had an implementation on an old site which had a broken link to the code; when I copied the name of the zip file that was linked in github; there was the code. It had no description or comments; someone just checked it in. It was not perfect code, but it shaved off a few days of writing it from scratch anyway.

I care about being able to code, which can be assessed from past code and a pair programming session and then efficient problem solving using your brain and internet. The latter is somehow less common.

So my favorite way now is; I ask them to 'solve' a problem they have never encountered and that cannot physically be done during the interview time (I am interviewing for software development and in our field you get asked for things that cannot be done in the time you are asked to do it all the time, so I need to see how you handle that); give them time to Google possible ways to do it and present a little plan. Then they need to pick, out of the plan, something they think they can do inside the interview time (see how confident they are and if they have any sense about low hanging fruit or estimations; I might suggest something else if the task they pick is only installing libraries for an hour ;) and we'll sit together behind a monitor and do it together. I'll usually tell them to stop after 15-30 minutes as I have seen enough at that time.

I did an interview problem like this a while back (was about combining CIDR blocks) and I found an answer on the Internet. I coded up a similar solution (I used a slightly different data structure with a slight optimization just to show I didn't copy the algorithm exactly). The place actually complemented me on my Google foo as they claimed they asked the question because they knew (thought?) people couldn't look up the answer. Of course, they dinged me by saying my code had too many comments (nevermind I was detailing WHY and HOW my algorithm was different from the Google results). I didn't get any love for the man(1) page I made either. Oh well, can't please everyone. :-)

If I were interviewing that would absolutely be an important part of the interview. It never ceases to amaze me how bad some people are at searching.

Hell, I've coded in languages that I didn't even know (didn't really even know what the language was -- I think it was like VB or something) by copying terms from some sample, reference or other part of the code and putting in the term or idea I was looking for.

Additionally, just try to look up items on a language that is in (constant?) flux like Swift. Swift 1.x? 2.x? 3.x? 4.x? What's the API today -- naming convention changed, parameters changed, etc. All I can say is that thank goodness Google allows for date ranges with searches. :-)

Principal Engineer from huge gaming company here.

White board coding interview are testing only one thing: how well candidate is prepared for it. And here what is wrong with it: The assumption that candidate should spend his/her valuable time preparing for someone's assessment is arrogant. I do understand why Google and Facebook do it (the do other arrogant things), they assume you want to work for them so much you will study to make the cut. So, they track bunch of metrics, which makes THEIR life easier. They have baselines, calibrations, and other things you never have in the real life. Hilarious part is that they are both risk averse (better to not hire a good candidate than hire a bad one) and using brood-force approach (they will interview you endless number of time).

And if you are not prepared - you will most probably fail, unless you are really good at it or lucky. So, why should I prepare for Google's or someone's else interview? I have a interesting and very busy job, I just don't have time for this. I have commitments to my team to work on real products, not on fake problems from Cracking Google Coding Interview.

Just think how absurd it is?

So, starting from 6 month ago I: 1) Tell recruiters upfront I will not be spending a minute preparing for their interview process 2) I decline coding on the whiteboard. High level designs are fine, writing code - no, no exceptions 3) And the most important, I stopped asking candidates to code on the white board.

The most important skill of the software engineer is (IMO): ability to keep complex problems in context for long time (not 5 boxes and months), ability to manipulate them and ability to convince others that you idea is worth pursuing.

> So, why should I prepare for Google's or someone's else interview?

Because you want the job? Everyone is busy, that's a terrible excuse not to brush up on your interview skills.

By the tone of your post, my sense is that prior to your recent epiphany you weren't very kind to candidates who you interviewed.

Do I? There are a lot of great companies, why should I waste time on Google? Working there is not as fancy as you think.

>> Everyone is busy, that's a terrible excuse not to brush up on your interview skills. I believe that their approach is arrogant, they just don't value candidates time.

I have a confession to make, I spent some time at Google. And after I left I was interviewing people using their approach and I regret it now.

P/S/ When you interview for a company, you actually taking more risk that the company (talking about folks who are headhunted into interviews). If it doesn't work out, you will be out searching for a job, you career growth may be slowed down. And what is the company risk?

> Do I? There are a lot of great companies, why should I waste time on Google? Working there is not as fancy as you think.

I'm definitely confused. You don't want to waste time preparing for an interview, but you will waste time interviewing at a company for a job you don't want?

Yes, there are lots of companies hiring. But in my experience, in preparing for one software interview, you're preparing for others as well.

I interviewed at Dropbox, Amazon, Valve, and Indeed, within a span of a few weeks, and the prep work (Cracking the Coding Interview) was applicable to all.

>>The assumption that candidate should spend his/her valuable time preparing for someone's assessment is arrogant.

Plus, if some one has to 'prepare' for an interview that is supposed to test people's capabilities to do 'everyday' jobs, then you are hiring the wrongest possible people out there.

How do you interview candidates for your company?

I would talk about candidates experience, try to find some project he/she are proud of and dive deeper. I might ask to write some code, but it is not going to be CS exam and it will be in real IDE. I might ask a design question. At the end of the day I need to answer 1 main question: "are you passionate about what you do or it is just 9-5 work"? Cause if you are you will be able to learn new things, and if you don't - knowledge of merge sort not going to save you.

Brute-force, not brood-force. Do you work for Blizzard? ;)

Heh, right :) Nope, they too far from SF

To me, it's not about the code on a whiteboard vs a computer and correctness.

It's about the thought process when breaking down a problem and the ability to organize and communicate their thoughts. It's also about exploring a problem space. How carefully does an engineer consider edge cases? What kinds of things are important to them?

A whiteboard to me is a much simpler and accessible medium through which to explore a problem vs setting up an environment and having to type code and dealing with all the minutiae that comes with actual development.

Of course, it isn't the ultimate method of evaluating a candidate but it's certainly valuable.

If the goal is to test the thought process why not give interviee a self-directed project then ask them to present a 10 minute talk about it? You get some code written to verify the engineer can build something, plus you get confirmation that they can organize their thoughts and effectively communicate them to another engineer.

Isn't that a more realistic scenario? How often do you give an (employed) engineer a task then ask them to immediately explain what they'll do to implement it?

Simple, I want them to work on the same topic as other engineers I've interviewed, so I can benchmark them against each other, and ensure the topic is close to what we work on, and not just what the candidate happens to know a lot about.

It is more realistic, but you can't really expect someone to be novel, and if they're not being novel, you should expect someone else to solve the problem. So this method has problems, and it kind of demands payment for results.

I'm confused by what you mean by self-directed. Would you give the person a specific problem to solve? Or would they be coming up with both the problem and the solution?

How many times did you solve a real problem with capturing all the edge cases in less than 45 minutes?

I think it's important to measure what's actually relevant to the job. If the job entails large periods of coding on the whiteboard, then yes, it's a perfect metric.

If, however, the job requires careful consideration of dependencies, time domain, and, yes, edge cases - then you are NOT testing for that.

Lock the candidate in a room and give a few hours to let her to produce the result would be a better measurement.

When I do whiteboard interviews I want a little bit of code because I'm stunned by the number of people who can't write a loop. But mostly I want to see the design process. I ask for a simplified version of a real feature that has an algorithmic core. This lets us code something but lets us discuss all sorts of edge cases and real world complexities that would come up in a true deployment.

Lots of interviewers suck. That isn't a property of whiteboard interviews.

>I'm stunned by the number of people who can't write a loop

I won't say this is you, but it's funny the number of people who will interview and say something like this and then still give a non-trivial problem/question to whiteboard.

Reminds me of a guy who interviewed me several months ago at a place in Mountain View (not Google, but another).

He started by asking me to create an object with a couple attributes. Then he slowly began to add to the problem by asking for X, then Y, then Z.

Little by little, the code became more complex because he wanted iteration, etc.

He would say "I just want to know that you're able to code."

Later on, I found out he didn't like what I wrote, meanwhile all the basic elements of coding he wanted I was able to add without any problems and I met his specs.

If there's a "communication issue" in whiteboarding, perhaps the interviewees shouldn't shoulder all of the blame. Perhaps the interviewers need to temper their expectations and learn to communicate them better, because obviously there is some disconnect in a situation like this, and it's not the first time I've seen this before.

so let's discuss the edge cases. let's operate with abstract concepts rather than actually write code.

Having just been through this process (landed a new job after a 2-month search), the worst part about whiteboard coding is how humiliated you feel if you don't get it right or don't know where to start. Is it really necessary to do that to candidates?

I've been doing this for 15+ years professionally, have a master's in CS and I struggled to answer many questions. Just stood there blinking like a goldfish.

If companies really wanted to simulate programming work, you'd give me 10-30 minutes to Google the problem because THAT'S HOW IT WORKS IN REAL LIFE. Which is how it should be; rarely are you solving a new problem, usually other people have faced it before, it's probably far more efficient to start with a known working solution and iterate from there.

Look, of course there things you should know. Data structures, algorithm basics, some complexity theory, etc. But asking me to output numbers in a matrix in a spiral fashion does not translate into real-world ability.

I find the more important developer skills are people skills and project management skills. How well do you work with others? How do you resolve conflicts? What's your methodology for grouping work-- Jira tickets, sprints, etc?

But almost none of the interviewers seemed to care. They wanted to know what exotic technology skills I had, not that I could work with project managers and marketers effectively.

Take-home projects are more time-consuming but the least unfair, whereas whiteboard coding is the shortest but most unfair.

I blame Google for this curse. I hope the next time I go looking for a job this process has changed.

Same. Being so far out of school (10 years in the industry) I really struggle a ton with these interviews. It's really only 10% relevant to my day to day work. I think to get through a Google-level onsite I would probably have to practice for a year, couple of hours a day. I just got a new job at a company that doesn't do any algorithm testing and that's how it's always been. I guess I'm maybe in the "false negative" threshold that all the hot companies are fine with.

Asking someone with your amount of experience a coding exercise is silly. You wouldn't have lasted 15+ years if you didn't know how to code. Spend more time evaluating the "softer skills," since more experienced engineers have a bigger impact on culture and process.

Right?

Ask me about past projects. Want GitHub examples? Fine. But whiteboard coding some non-real-world problem?

I get that companies need some sort of filter, but whiteboard coding is not the answer, IMHO.

Length of tenure is definitely not a good enough gauge.

I've worked with folks with 15+ years that could technically code but they took eons to deliver and their solutions were unmaintainable. Some of them with advanced degrees too.

Small sample size, but the few folks I'm thinking of were such bad hires that could have been avoided if we had them write some code or put the proper emphasis on the coding exercise we had them do.

But being a slow worker is going to be revealed during a whiteboard coding test?

All small programs are maintainable, no interview process will tell you if a developer can deliver large, complex programs that are clean, modularized and easy to maintain because there's no time to do something of the requisite size.

It's true there are experienced devs who suck, but that doesn't mean whiteboard coding is the process that filters them out.

If you must do whiteboard interviews, here's a technique I've used with great effect:

Have someone else pick a coding problem (that you as the interviewer don't know beforehand), and work on it from scratch together with the candidate.

Being upfront with the candidate that you don't know the answer yourself puts the candidate at ease, and you'll be able to better judge the candidate's soft skills (communication, critical thinking, empathy in case I don't understand the problem).

I like this idea. Personally, I struggle with knowing that the interviewer knows the answer. My wife's in medicine and she and her colleagues call this process "Guess what I'm thinking?" I prefer to have dialog and come to a conclusion together, and I feel it gets what you want out of an interview: how does the candidate think, how deep is their understanding of the technology, how do they communicate, and would you enjoy working with them.

I find it weird that the responses seem to draw a distinction between whiteboards and coderpad. The assumption seems to be that the problem making the whole thing stressful and unnatural was the communication medium and not the fact that it's a super time compressed artificial problem with many unstated considerations and a tremendously outsized influence on the rest of your life, no pressure, just talk through everything exactly like you never do.

Most of the problem seems to me fundamentally unaddressable without verifiable portfolios, but those have their own problems. Even in industries where portfolios are standard, plagiarism is a thing - how do you get the verifiable part? I want to hire someone great, not just someone who was on the same team as someone great.

Even just coding on a computer with the screen cast up on a big monitor or projector would be a vast improvement over literally writing on a physical white board.

I wonder how many of these people are actually active coders which gives them authority on opinionating on how to interview developers. VPs (especially Chief XYZs) are typically good sales and marketing guys with engineering background that often can only be described as hopelessly out of date (or more politely "has-beens"). They need to keep their heads out of real development stuff.

I'm also frustrated with how people have developed disdain for anything whiteboarding. Asking leetcode puzzles in interviews is bad. Asking questions from CTCI is worse and amateurish. But... asking to whiteboard a problem that you had to solve on your own job in limited time is good! Interviewing is hard because crafting such questions in a way that it abstracts away minutia of transient technologies but retains essential problem solving is hard. This is exactly why, a right policy for a company is to establish being an interviewer as a privilege that needs to be earned because not even all brilliant programmers are good at it and interviewers must take time to craft good question and show care. When they don't, they turn to leetcode, pick random question and give whiteboarding a bad name.

And no, whiteboarding is not bad. What is bad is asking to do X using MongoDB and RectJS because technologies like that are transient. Specific technical tasks that employees might perform in a given point in time are transient. Ultimately what matters is ability to learn, process and use that to solve a given problem. Ofcourse, unless you are company who does nothing but data driven web forms. In that case, yes, go ahead and ask candidate to do just that.

The term "whiteboard interview" seems to mostly refer to "solving algorithmic/coding/puzzle problems on a whiteboard", which I agree is almost always a waste of time for a software engineering position.

As the article mentions though, the term is a bit vague and is overly broad; for example if you're doing a system design question, I'd argue that the whiteboard is the best choice for an interview, since that's how you're most likely to collaboratively sketch the design of a system in the real world.

I think the problem can be restated at a more general level; does the interview test the sort of skills that the job requires? Most people here are correctly pointing out that the skill/knowledge required to write out pseudocode for quicksort is almost entirely uncorrelated to the skills involved in most software jobs.

Whiteboards, in practice, are never used for writing strict code. They're used for high-level diagramming and high-level pseudocode.

The only time I've used whiteboards in my job is either when I need to hash out a solution or approach with my coworkers, because I'm confused or don't understand the problem, or when I need to understand high-level architecture for a system. It's a tool for clearing one's thoughts and for getting a team "on the same page".

If you're hiring someone to code-monkey a bunch of JIRA bugs, then whiteboarding might not be a relevant test for them. If you are hiring someone to build out new features efficiently or redesign architecture, then whiteboarding may be useful for having them describe at a high-level how they would solve problems. In both cases, writing syntactically flawless solutions on a whiteboard is a waste of time for both the interviewer and interviewee.

At a big 4 company recently, they specifically told me "avoid writing pseudocode"

Most selective companies require you to have near correct code in the most efficient algorithm, clean, correct, concise, with all edge cases in about 30-45 minutes depending on the question. They also seem to love dynamic programming algorithms and competitive programming tricks (the less common stuff you glossed over in an advanced algorithms class).

I say this as someone who worked at one and interviewed at others. You really do need to treat it like an exam.

Interviews of the sort you describe are like judging a fish by its ability to climb a tree. Indeed, whiteboards in interviews are used nothing like they are in real work. I'd hope more companies try and make interviewing closer to an actual work environment where the candidate can comfortably use their computer for solving the interview problem, using a whiteboard solely for communicating concepts and other surface level issues.

> "Whiteboarding" has become too large of an umbrella term, one that groups together everything that's wrong with the interview process.

That general premise is right, we've lumped too much into the term "whiteboarding", and it's worth unpacking into the different interviewing practices we may like or dislike. Some things people don't like have nothing to do with an actual whiteboard (e.g. testing esoteric CS knowledge or deliberately creating a stressful environments), and some things that involve a whiteboard IMO are ok (e.g. sketching out a system diagram).

Yes. The whiteboarding that gets hate is effectively the one where you would be asked to write a depth first traversal or similar, in some verbose language, correctly. On a whiteboard.

That is STILL a thing (judging by HN) and it's useless. I don't think anyone would ever think it's a bad idea to have a whiteboard on hand if a candidate is asked to describe how they would design some nontrivial system. Being asked to describe something without a whiteboard in that case, is even harder. The objection you often get for the latter kind is that it tends to benefit the extroverted kind that likes to babble and sketch, whereas a candidate who would prefer ten minutes alone with a pen and paper might not do so well. I think are some merits to that complaint, but I also know from experience that being able to babble and sketch is really important.

My most recent interview involved a coding assignment, which I topically don't do, but this one was a very interesting challenge that would make a good blog post afterwards so I did it. What was surprising is the interview not only included going over my code from the assignment, but also whiteboard problems about recursion. I've still never used recursion in my day job. Can interviewers really not gauge technical ability from my GitHub as well as chatting about problems and solutions?

Interviewer: I see you have built an app just like ours five years ago. That's great. OK now let's get to the important stuff. How many cakes can you steal from the queen if your duflle bag hold 5 pounds and there are 10 different kinds of cakes?

True story.

You have a software engineering job and you never used recursion of any kind? This strikes me as super odd.

I can only think of a couple occasions in my professional career where recursion made more sense than iteration. Both involved traversing trees with unknown depths (parsing XML and scanning a file system).

That hasn't stopped it from showing up in every programming interview I've ever done.

I haven't really used it for anything other than super simple things with a very tightly bound max input. Stack overflow and honestly many recursive algorithms are quite hard to parse in your head especially once you add some edge cases and some other entropy.

> Stack overflow

Well, yeah, other than tail recursion with tail call optimization, stack consumption, if not stack overflow, is always an issue.

> and honestly many recursive algorithms are quite hard to parse in your head especially once you add some edge cases and some other entropy.

I find lot of things are easier to conceptualize recursively than iteratively, though parsing really depends a lot of the language.

I do embedded work. In this field some compilers don't even allow recursion.

They might generally use a library that handles a lot of common usages for recursion, like traversing a tree for whatever reason.

A question I've had in an interview was what is the difference between recursion and iteration. Fairly easy to explain, including examples of usage.

I too never used it. Recursion is usually an inferior solution.

It's almost always a norm that you never use recursion in production code. Always convert it to iteration. Bounded iteration.

Depends on the language too, I suppose. I'd be more shocked if they said they had never used recursion, but were using haskell at work.

Whiteboard is GREAT for communicating design or technical ideas (e.g, what is consistent hashing, can you show how you'd merge N sorted lists schematically?). And these skills are super important in both big companies and startups. And we should definitely test candidates in these areas using whiteboards.

Making engineers write code on whiteboard is just comical. If we weren't so used to it, we'd probably have laughed at the idea. In 2018, any company still using whiteboards to write anything more than pseudocode is just playing it too safe or being really lazy in changing their process. Just give them a laptop and have them use coderpad/skype interviews or any of the plethora of collaborative code editors out there. Or even, dare I say, an IDE. You can still have the whiteboard in the room to discuss the idea. But write the code on a laptop with internet access.

I think most people that are opposed to coding interviews feel that way because they don't feel confident in their ability to do those interviews well, whereas they do feel confident in their ability to "get work done".

Here's the thing though. Let's say you're the one giving the interview to two people who claim to be able to get shit done. One is able to code up a solution quickly and the other struggles, producing less, or producing something that's flawed. Sure, that second person might actually be productive under the right circumstances, but that doesn't matter because the first person already proved they can code.

That's the whole point of interviews, getting high signal in a short period of time. If you think coding interviews don't provide high signal than propose something better that costs the same amount of time.

Your resume, Github account, demonstration of previous projects, ability to discuss problems you've solved, references from previous coworkers & instructors...

The whiteboard is not providing "high signal". Everything else is. The whiteboard is providing low signal, and taking quite a bit longer to do it.

This doesn't take into consideration how much studying and preparation the person who produced a solution quickly did compared to the person who took longer.

If that person studied much more, then they weren't really "quicker", they were just better prepared.

Perhaps that's still a reason to hire them, but not everyone has time to study lots of programming puzzles before an interview, and this is then less a measurement of speed, and more of preparedness.

It also doesn't take into consideration whether the whiteboard problem being presented is at all relevant to the job being offered.

I generally ask candidates what they did last week at their existing job, and then ask questions about how they solved specific aspects of that particular job.

If the person says "we built real-time tracking for our in-house vehicle fleet", I will just probe further and further about what role they played and how they went about the specific implementation details.

If there is any kind of secrecy surrounding their work, I will simply offer them insight into what they'll be tasked with during their first week and ask questions in a similar manner.

I'm not sure what advantage any other approaches have - I am hiring to fill a particular position. It's quite likely that the candidate will be in a similar position, so either I ask them about their current work and how they went about it, how they intend to accomplish the work assigned to them when they work with me.

Serious question, if whiteboard interview aren't the best or helpful at all, what are the successful/useful alternatives for engineering interviews?

The best I have come up with if you are requiring them to write actual code is to give them a problem to solve for with code prior to interviewing. Have them bring that code along and then have them explain the code and why they did things the way they did. It is not 100% verifiable that they wrote the code, but being able to talk to why things were done the way they were tends to be good enough. I have also found that understanding how people approach life and problems in general are more telling of how good they will be than actually evaluating their coding skills. Monitoring coding exercises will weed out many good candidates as well as the poor ones in my experience.

Exactly.

During my recent job search I was asked many questions I didn't know the secret algorithm to, but if those companies had given me the problem ahead of time and allowed me to research it -- just like what happens in real life -- I could have coded solutions to all of them.

And if you're worried about people just plagiarizing existing solutions, have them talk through it, as you say. Should be fairly clear if they understand each line or not.

I really wish we had multiple formats that the candidate could choose from:

1) Whiteboard

2) Review and technical discussion of existing code or projects

3) Paid take-home project

4) Pair programming on live code

5) Technical discussion only for rare cases for those Top Secret/NDA candidates that really can't discuss their past projects but are also time constrained

Some people do actually like to whiteboard. Keeping it around is kind of like legacy support for people who practiced hard to excel in those interviews.

This is pretty much the union of technical interviews that are well-known. If you switch out a technical interview format and can't get meaningful insights into a candidate, you're testing for the wrong things.

I haven't had the opportunity yet but I really want to take a piece of code from our system, de-optimize it a bit and add some bugs, then give users a unit test around it.

Most of what we do as engineers is trying to read code to figure out what it's doing, how to fix it and how to enhance it. Going through real code with a candidate and assessing their ability to understand brand new code (and what questions they ask) seems like it would give good insight to how they would perform on the job.

In one of the best interviews I had, the interviewer had printed off an actual (undocumented) class in one of their projects (I got the job, so I confirmed it later), then asked me to tell him what it did, and if there was anything I would do to improve it. I think he might have added a couple of logic or syntax issues as well, but I can't quite remember anymore.

I've been on dozens of interviews, and that's the only time I've ever encountered that, but I think it does a really good job representing the job, as that's invariably what I'm going to have to do when you hire me, is sit down and familiarize myself with the code base.

He also asked me a handful of questions, said "Okay I know you know enough to do the job, let's see how much you really know," asked me a bunch of really low level stuff, and proceeded to teach me concepts when I told him I didn't know certain things. I walked out of that interview knowing more than when I walked in, something else that has never happened in all the other interviews I've had (except maybe I learned a new term that I never heard before that I was apparently supposed to regurgitate back because that's what was written down as the answer on HR's answer sheet).

Coding challenges either take home or live session.

Most whiteboard questions are always some esoteric algorithm any engineer worth their salt would in real life would look up before ever consider using to verify their own understanding before using it.

Putting someone on the spot saying whiteboard x is not ok.

However if the test is to see if they can take part in a meeting, then by all means do a white board, but ask them draw a workflow diagram or some other 40,000 foot type of diagram. Do not ask them to put code on a whiteboard.

It's not just that real working engineers would go look it up. It's that the culture around these "prove you can code" problems got into an arms race, where companies kept trying to come up with ever more difficult problems and stringent requirements on solutions. It finally reached the point where, at some companies which use them, the bar for "can you code at all" is actually "given a problem you've never seen before, can you match or beat the absolute best solution the world's best CS researchers have ever come up with, in twenty minutes, on this whiteboard". If you aren't able to, they label you a bozo who can't even write a for loop, and pronounce you unqualified to do any type of programming, and pat themselves on the back for having kept "fake coders" out of their company.

(and if you wonder how anyone ever gets hired there, the answer is: by cheating. If you're going to interview at one of those places, odds are their interview problem and an accepted solution will be online somewhere, so you just go look up and memorize)

Yes, my favorite was a while back, someone's list of algorithmic interview questions got posted on here, and a couple of his answers were wrong! And he STILL insisted they were good interview questions!

Our field is full of some of the smartest morons on the planet.

You absolutely nailed it.

We could be like real Engineers (civil, certain classes of mechanical and electrical) and require a license. One 8 hour exam covering fundamentals and application of them along with simple project management and soft skills.

Depends on "why" you are recruiting and what the end result of employing someone is.

So much of the technical hiring process tends to be about trying to find people that match what you expect people to be. Which might be fine, you need to fill a particular position which requires a specific set of skills. You need to validate those skills. You do that by doing something semi realistic while trying to minimize the pressure of an interview situation.

But, then there is the other kind of approach where you are looking to see what the candidate can offer, trying to find people who know different things to you and do things differently than you, in this scenario you are more inquisitive and let the candidate direct the interview process and get them to show you things that sound interesting (including coding).

The first approach is all about finding specific skills. The second approach is all about being clear about the values you are looking for.

The second approach requires more skill and more time while wearing the hat of a technical recruiter.

You may need a little bit of both approaches.

So the big thing is to be clear about who you want to work with at the end of the process.

Just talk. Have a conversation. There is no point in trying to put someone outside of his comfort zone. You should take him to the very center ot it, and then see how well he can managd. Any question that the interviewer already knows the answer to is not a good question.

My favorite interviews have been those that had a small project as a screener and then proceeded to have you modify and/or explain decisions regarding the project at the in-person stage.

Letting candidates to do actual coding on a computer.

Yeah, but that's also stupid. All to often it comes down to "Make standup a completly working client server system for this arbitrary problem from scratch in 30 minutes, starting... NOW!"

You end up spending all your time dicking around with missed semicolons and looking up library routines and shit like that. Sure, it's all stuff you have to do as part of a job, but it's also all trivial stuff.

The last time I had to do this, I spent time googling how to open up a port in python, and then read from stdin. It's just a waste of time.

All this does is test the wrong things.

Find a reasonably self-contained problem that doesn't involve too many interactions with the "environment". For example, ask the candidate to refactor an existing piece of code.

Whiteboard coding measures penmanship while nervous.

I have some firmly held opinions [1] on this topic. Interviews are too much of an "art" for how important it is that their results be repeatable. Keep them on topic as much as possible. Making a coder perform circus acts on a whiteboard they will never be asked to do otherwise is simply measuring for skills you don't need.

1. https://blog.benroux.me/whiteboard-coding-measures-penmanshi...

They're fine but they need to be understood properly. Whiteboard interviews don't really capture how well someone codes. They are better at capturing thought process, high level understanding, system architecting ability, and a candidate's communication skills.

If you want to see how well someone codes, look at their code. Or ask them to write something for you. Don't do a whiteboard interview and think it is the end all be all. Take it as a data point.

> They are better at capturing thought process, high level understanding, system architecting ability, and a candidate's communication skills.

If that were the case, ask me about a past problem and my solution and to defend it, not the proper way to invert a b-tree.

For anyone that wants to work for a company that DOES NOT do white-boarding during interviews, here's a solid list: https://github.com/poteto/hiring-without-whiteboards

Nay.

We have a round of interviews and then send the candidate home with a choice of 5 coding tasks. Intention is to spend no more than 2 hours. Then they return and hold a code review in front of the team and explain what they've written. That way, they code at their own pace and in a comfortable space where they can focus, much like the way they would actually work. Then they've got it all right in front of them for discussion.

The review then weeds out people who don't really understand what they've written, possibly because they borrowed it from somewhere or asked a shill to write it for them.

We've had a range of submissions, from the casual-careless-broken code to extremely complete to elegance like we've never seen before. It works well.

Random fun idea: pair whiteboarding, have one of the reviewers start improvising a solution (or put on a show of not having done it dozens of times before) and see how the candidate will contribute.

If the problem is not one of the standard ones (that will be perfectly memorized by desperate candidates, while rightfully confident ones may struggle), it will demonstrate understanding, not the irrelevant ability to sequentially write out code in an unfamiliar medium. After all, you probably don't want to hire those who have already been rejected so often that they can code on a whiteboard as if they did it all day.

Yes? It designs to filter people based on certain criteria. Nowadays, you just can't get a job by merely talking, the market is too hot, too many people want to get in.

I don't think design interview alone is that useful either. Someone without actual experience designing stuff could still fake it by reading other people's past solutions and answers, there are templates on the internet.

Nothing is inherently bad just because it fails you. Everyone saying Google's process is broken, blahblah, doesn't change the fact Google employees are hot AF on the job market.

>Nowadays, you just can't get a job by merely talking

You can if you're willing to work for a smaller, non-famous company.

I'm an almost-graduated computer engineering student. I had a relatively pleasant interview last week with a whiteboard component which I thought was fair. The interviewer asked for a Fibonacci function, an easy problem that still allowed me to show that I understood some more advanced concepts. I gave a basic recursive solution, was asked to critique it, explained how it was inefficient in time and space, and would blow up the stack (without tail call optimization). Then I thought I'd mention dynamic programming as a possible solution to the time complexity, but one that would still cost a ton of stack space. Finally, I gave the iterative solution, and explained how it was more efficient, and how you could use it to pre-compute a lookup table.

The interviewer propmpted me along the way, it felt like a friendly conversation. The fact that it was an easy problem prevented me from stressing out, but I still felt like I was able to demonstrate some academic ideas.

I was always pretty decent at standardized tests as a kid, and this argument over whiteboarding reminds me a bit of that. I don't think it's useful for gauging general software engineering skills, but it might have some value for evaluating communication. It also feels like the sort of thing that software engineers with a strong academic background tend to enjoy more than those with more practical/industry experience, but that's just anecdotal.

Honestly I’m not afraid of whiteboards as earlier in the career. The interviever is whether some fresh graduate smartass trained in a narrow domain and it’s easy to route the discussion out of their comfort zone, or someone with particular problem - well, I’m not able to pull out immediately answers to the world problems and if this is the case I’ll just admit it. In the worse case I’ll not get hired - I hadn’t got hired too many times in my life to care.

As with everything, it depends.

To me, whiteboard interviewing makes sense when hiring for a role that leans toward a comp-sci background. People who do well in these roles typically have a brain that works in a certain way.

The roles that lean away from a comp-sci background are usually the type of developers who don't do well with whiteboard interviewing techniques.

What too often happens is that a hiring engineer doesn't interview within the context of the role they are hiring for.

> People who do well in these roles typically have a brain that works in a certain way.

You know this how? Intuition?

People have all different kinds of backgrounds and brains.

I’d be curious, are interviewers running whiteboard interviews willing to do one, conducted by a colleague, in front of their team?

Does this practice exist anywhere?

It would be interesting to do this maybe once a year? And put $100-500 toward getting a pass or non-pass(money gets donated to interviewer-selected charity), just to throw in some money in the game?

And I mean an interview based on non-trivial CS problems, not brainstorming / code ideation which we do often.

I'd propose a supplement for interviewing. If your company has at least one open-source repository:

1. Have a senior engineer chat with the candidate for an hour about a few bugs/features in the repository, and how the candidate would solve them.

2. If the candidate is promising, give the candidate a week or so to fix the bugs or implement the features.

I think this would be a more relevant test than just giving them context-less problems to solve on a whiteboard.

Make it a paid project, and it at least partially addresses the sibling comment's (valid, IMO) complaint that it filters out candidates that are otherwise in high demand.

It would also provide a significant signal regarding valuing the candidate's time and effort [1], which would comply with the suggestion of another comment elsewhere in the thread.

[1] As well as the additional signal of putting money, rather than just "free" candidate labor into open source.

Companies already pay thousands of dollars (per candidate) to fly out, accommodate and feed candidates for on-sites, so this doesn't seem unreasonable. Similarly, phone-screens and other filters would be done before a paid task is given.

This basically filters out those who already have a job or getting multiple offers. Most candidates don't want to spend more than a day on an interview.

Many people believe that it is completely unreasonable to ask an interviewee to write code on a whiteboard at all. I disagree with such an absolute position. I do think it can be a reasonable thing to do, but only if you are asking the interviewee to code something really simple.

When I say simple, I mean really simple, like FizzBuzz simple.

There are interesting questions you can ask that are really simple which are not FizzBuzz. I've used the Fisher-Yates shuffle algorithm and more recently Selection Sort. I should stress in both of these cases that I carefully explained the algorithm before asking the interviewee to code it up. Even though these algorithms are quite simple, I do not expect the interviewee to have memorized them.

I do expect someone who is a professional software engineer to understand the algorithms after a decent explanation and then write code which will implement them.

I also don't necessarily expect the interviewee to get the code exactly right. I just expect them to demonstrate that they could write correct code given enough time and access to an actual computer to test the code on.

If you're looking for them to do something _really simple_, like FizzBuzz, then you're doing it wrong already. This is something that could have taken place over a technical phone screening.

Once they hit the on site interview it should be more about how they can work together with a team, how they would design systems, what they would do if they needed help, etc.

Coding up FizzBuzz on-site is a waste of everyone's time.

We're a small startup, and we don't have a really structured process. Also I have been the only person asking a detailed coding question (in consultation with my boss). Everybody else on the team has been responsible for those other questions.

I should also add that I personally prefer to do coding questions in person on a whiteboard, rather than doing them during a phone screen with some sort of collaboration software.

My technique of explaining the algorithm before asking the interviewee to implement it works way better with both people in front of a whiteboard.

I'm not saying everybody should do it this way. I'm just saying that this is my current approach and it seems like it's working OK.

I mean if it works for you then great but I've seen that back fire far too many times where I had to interview an incoming candidate who didn't do any coding over collaboration software and then couldn't do simple FizzBuzz. Basically wastes their time traveling and interviewing and my time when a quick check just to make sure they can do a few, quick basic things over the phone can save so much time.

While I dislike the way most large tech companies handle their interview process this one I don't mind as much.

How exactly could you evaluate FizzBuzz over the phone? Its designed as a weed out question early on, if they can't do that (and yes some have failed) then it makes a decision pretty easy

Most places I've interviewed at (or was the interviewer) talked to the candidate over the phone but had something like Google Docs open so you could watch some live coding. Usually just go through some very basic exercises to make sure the person at least knows how to code and ask a few, general programming questions. Stuff like that.

Is there any engineer who you'd regret not hiring, who is unable to code FizzBuzz in real time in front of an audience? I'm skeptical.

You certainly have a point. But those whiteboard interviews people complain about are usually a bit harder than just solving FizzBuzz.

I've seen a respected engineer that I worked with for years, be told by an outsider that he was incompetent because he couldn't fizzbuzz an interview. Yet, they had no problem acquiring the project he designed and wrote.

So yeah.

Why couldn't he do fizzbuzz?

"Why couldn't he do fizzbuzz with someone else watching his every move while he wrote on a blackboard with no interpreter to test his solution or catch his mistakes?" seems like the better question.

How do engineers approach solving a problem?

Whiteboard coding does a good job at assessing that, regardless of the candidates ability to actually solve the problem (for all the reasons why whiteboard coding sucks).

Do you analyze the problem?

Write test cases first?

Write a mini spec?

Gather Requirements?

Ask Questions?

Jump into code?

Google?

Verbalize the problems you encounter?

In that respect, engineering coding interviews work well. But too often, they are essentially a binary Yes/No: Candidate solved the problem the way I want them to solve it.

Whiteboard as a communication tool in interviews is great. Sometime it's easier to write things down, draw diagrams, and list technical points than just talking about it. Writing it on the board provides an outline to talk about stuffs.

Whiteboard as writing programming pseudo code or steps are fine. Just don't be a human compiler and demand perfect syntax.

It seems like there are two things at play - whiteboarding interviews vs. using whiteboards in your day-to-day job.

If the test is to see how well a person works in a whiteboarding interview to solve a problem, then it should be a simulation of collaboration, while trying to avoid the Clever Hans effect.

Otherwise your DP whiteboarding problems are worthless.

In general, an interview is to understand what a candidate brings to the table. Therefore, one of the best ways to assess "glass half full" is to talk about their past work. The interviewer is a subject matter expert who can definitely ask pertinent questions about the candidate's work. In many cases, more so for scientific positions - the candidate's work is publicly available.

Currently, many interviewers go for - as I like to call it -"trivia questions". Personally, it just doesn't seem appropriate because "Everything may be obvious, once you know the answer". In some such interviews, I've asked my own set of trivia questions to the interviewers. The interviewers weren't pleased as they couldn't answer some questions for which I knew the answers beforehand.

Unfortunately, some companies' HR departments actively encourage "trivia questions", because they ask interviewers to ask every candidate the same questions so that a "fair" comparison between candidates.

I think you've touched on a crucial point here about making the comparison "fair". In my view, I find the "trivia questions" approach to be an abdication of responsibility for the difficult, nebulous task of assessing a candidate's abilities. It gives you an objective measure by which to compare candidates... even though the measure itself is usually arbitrary and irrelevant.

The only "trivia" questions I like are ones that don't necessarily have a single "correct" answer (i.e. not "guess what I'm thinking") and can actually lead to further discussion.

For example, one I like to ask (in person, with no computer access), ostensibly to gauge familiarity/intimacy with the Unix CLI, is:

"For each letter of the alphabet, see if you can name a Unix command (anything one can type at a shell prompt that won't return a "command not found error") that starts with that letter."

I preface the question with something suggesting that it's supposed to be fun, and I encourage skipping letters and coming back to them to avoid pressure (nor do I write down the answers). Most people are surprised at how many they remember without much thought.

Although not particularly deep, further discussions can arise from differences between shells with respect to built-in commands, as well as differences between OS flavors. Also, I've discussed how a somewhat new language like Java has made one of the otherwise difficult letters much easier.

Regardless of the trivial nature of the question/exercise, it provides a spectrum of jumping-off points for deeper knowledge discussions, all of which are signaled/chosen by the candidate.

A couple of comments have mentioned LeetCode. While reading those comments I had an odd idea on a different way to use LeetCode when interviewing. The usual way would be to get problems from them and spring those on the candidate to solve on a whiteboard or on a computer. Numerous people have raised objections to that kind of interview.

Instead, how about sitting down in front of a computer with the candidate, going to an interesting problem on the LeetCode site, and reading the problem statement with the candidate. Make sure you both agree on what the problem is asking for.

Then, instead of having the candidate solve the problem, go to the discussion page for the problem, and start reading the posts with the candidate. These will include posts from people who are asking for help understanding the question, and from people who are showing off their solutions.

Pick a few where the poster is asking for help understanding, and ask the candidate how they would clarify things for the confused poster.

Pick a few where the poster is posting their solution, and ask the candidate to do in informal oral code review along with you. Let the candidate take the lead, just providing nudges if necessary. For ones where the review finds serious bugs, talk about how to make test cases to demonstrate the bugs.

From what I've seen on LeetCode problem discussions there will be a lot of posted solutions that have edge cases that fail, or fail to meet time or space requirements that were given, or leave out major cases, or make implicit assumptions about the input that if violated invalidate their sultion. There should be no problem finding plenty of posts with plenty of things to bring up in review to discuss with the candidate.

This seems to me that it would not be overly stressful on the candidate (way less stressful than writing code under a short deadline), it is exercising skills that working programmers actually use, and it would let you do some work with the candidate to get some idea of how they get along with others when working.

I recently had a coding interview and it was a disaster. The annoying thing is it was actually a super simple question and I had done a lot of preparation, and had solved a similar issue in my actual job just by chance a couple weeks before. So it started out great.

But instead of acing it an issue cropped up in the coding pad which I had to try and debug, which threw me off guard. The interviewers had muted their mic and were chatting away and my nerves just crumbled and I ended up profusely sweating into my eyes, I literally had sweat going into my eyeballs. I just turned into a bit of a mess, which is strange for me as I'm normally quite confident.

I think if the interviewers were a bit more involved it may have gone better. But I ended up just not finishing in time feeling like I had wasted everyone's time.

Yes to whiteboards if they're used correctly.

If the candidate can't explain or solve something simply on a whiteboard, they probably don't understand the problem or domain.

That's not to say that the interviewer might not be a great interviewer in the first place though.

Quote: "Feynman was a truly great teacher. He prided himself on being able to devise ways to explain even the most profound ideas to beginning students. Once, I said to him, “Dick, explain to me, so that I can understand it, why spin one-half particles obey Fermi-Dirac statistics.” Sizing up his audience perfectly, Feynman said, “I’ll prepare a freshman lecture on it.” But he came back a few days later to say, “I couldn’t do it. I couldn’t reduce it to the freshman level. That means we don’t really understand it.”"

Yeah, no. At least, not to test the technical proficiency of a potential team member.

Instead, have a conversation about how to solve a well defined problem. Use the whiteboard to record potential solutions in high level terms, or if the candidate chooses, to note key concepts or concerns.

Your interview environment should be a "staging" environment for your day to day operation. The interview should gauge how the candidate might affect your development process.

Once the interview ends, snap a picture of it. When you're debating candidates and try to remember your impression of them and how they contributed. Consider the sum of the input they verbalized and the input they wrote. That should give you a good idea of what they'll offer - well, along with your impression of them to that point.

In my experience conducting engineering interviews (about 300ish so far) I much, much prefer pair programming interviews in front of a computer with an IDE and access to the internet and standard resources. This most closely resembles how we work together and engineers and gives candidates the best chance to show off their skills. Whiteboarding code is something people almost never do in their professional careers, and as such, you focus on all the wrong things (is that the name of that library method? is that where that semicolon goes? is that best practice?). It's neither fair nor representative.

Of course I have no problem testing architecture and communication ability in front of a whiteboard, as this is much closer to how these things are done in the workplace.

I think in images, and always visualize the problem, so I would love doing it on a white-board, I actually find it easier then writing the code itself, and I find it very hard to describe the problem in words. While others think in code, and yet other people think in words. So in order to not dismiss a good candidate, explore how they think about problems. Can they describe it with words ? Can they describe it with code ? Or maybe they can only describe it in pictures !? There are also other ways of thinking. Some people think in shapes, then solves the problem like a puzzle making the shapes fit together. These guys/girls might be on savant level and would have a very hard time passing any traditional interview.

So I've been frustrated by this type of interview for years (on both sides, but the last 10 years on the hiring side). I don't think it's great finding people that are good at writing software (as it's a LOT more than a single algorithm and more than just coding).

I wrote a book it - along with what I've been using the last few years, that actually seems to work - or work at least a bit better. In the end, I don't think you really know unless you work with someone.

https://www.amazon.com/dp/B07FJ6N8P1

It's my first attempt at a book (small as it may be!) - I'd really appreciate some feedback (and it's free right now)

-Greg

We offer a whiteboard for the onsite if it will help a candidate communicate comfortably. Narrative is totally fine, gesticulations, metaphors, interpretive dance, whatever means of communication works to get the point across. We send out a small HW assignment before an on-site, after a phone screen. Not all candidates enjoy it, some expect an algorithm driven whiteboard experience. Our process delivers for us, and we tend to naturally attract a good culture fit as a byproduct.

In practice we do end up at a board day to day explaining ideas and so on, and everyone seems more than comfortable (with the exception of handwriting insecurity).

I always do horribly on these. I also don't have a degree in computer science so although I can setup a binary search tree, etc. well enough if work demands it I don't have the muscle memory of working with them intensively over and over again to spin out solutions to interview questions.

Once in the workplace however I usually get promoted up pretty quickly. I'm pretty handy with generics and meta programming along with architecture and design patterns, meaning I can turn out more complex systems in much less time than most of my coworkers. But that crap never comes up in the google/microsoft interviews of the world.

IMHO, people with/without CS degrees should be evaluated by different standards/methods.

If you do have a CS degree, then I expect you to be able to show decent competency with that material because you were sitting in all those classes for all those years. If you didn't learn anything, it makes me question whether you apply yourself to what you do or just coast through stuff and are going to coast through at this job too. Since we have the shared experience of a CS degree, if I understand what you got out of it, I think it tells me something about your attitude and approach to your work.

If you don't have a CS degree, then I am going to focus more on skills required for the job role itself. If being strong on algorithms is really necessary for the job, then I'll need to cover that. For any job, though, I'll need to convince myself you have a certain base level of competency so that, for example, you know that a linear search of a billion items is bad.

I've always been in the camp of not doing coding interviews...whiteboard or otherwise. Overall, that's worked well when hiring engineers with 3-4 years of experience or more.

Less experienced engineers are very difficult to assess without some kind of coding exercise. All the hires I've made that I'd characterize as mistakes are of junior engineers that I didn't do a coding exercise with and interviewed well, but didn't bring that interview polish to the job. Since instituting whiteboard exercises for junior positions, our success rate has been higher (no pun intended).

If you're going to interview someone that is going to write code that a computer runs, then make them write code and run it on a computer IN THE FUCKING INTERVIEW.

Silicon Valley is retarded.

Whiteboard interviews would be fine if they weren't taken so damn literally. Writing code on the board is dumb.

Spitballing solutions to a problem collaboratively and with access to reference resources (a nice little programmer's bookshelf in the room) followed by actually coding a little bit up seems that it would be way less stressful and way more likely to give you a sense as to what someone is capable of and what they're like to work with.

If you can solve arbitrarily hard problems while on the spot you must be a stellar candidate from a problem solving perspective. If you can effectively articulate your thought process as well you probably have great communication skills and can handle technical discussion and group problem solving for harder problems. It's gonna weed out a ton of people that can do the job but the guys that rock it will probably be valuable assets.

I've been on both sides of this. Whiteboard interviews are good for a couple of things: Can the candidate explain a problem on a whiteboard? Is his hand-writing legible?

It's not sufficient to decide if someone can code. You want some actual code written by that person.

And you don't need everyone on your team being able to explain things well on a whiteboard. Just try to figure out how they are thinking and how they are approaching a problem.

I've done whiteboard only interviews, as well as shared doc/coderpad only, and I find both limiting. When thinking about the problem, it's great to have a whiteboard to sketch examples etc, but writing code is much easier with a text editor.

Recently I interviewed at Google, and they had no problem fulfilling my request to do both: sketch the solution approach on whiteboard and then write the code with a laptop.

great article! it would have been nice to have a comparison of yays vs nays, like for example 67% of engineers say nay, whereas 50% of companies on key values do have some sort of whiteboarding in their process. with the caveat of course that the kind of companies that talk to you (keyvalues) are probably more progressive anyway so there is selection bias.

but it very much does seem like whiteboarding is on its way out.

> What’s great about a white boarding session is that it is a blank piece of canvas, where the interviewee can effectively think out loud, collaborate, draw diagrams and also brainstorm with the interviewer.

Is the typical engineer conducting an interview good at contextual inquiry? At holding space for people to think out loud? Lets not kid ourselves here. The only people we’re harming are each other.

>Software engineers hate whiteboard interviews. How do I know? We continually tell complete strangers just how much we hate them.

And yet, whiteboard interviews are probably 3rd on the list of flamewar topics behind vim, and whitespacing. So there is clearly some people who don't hate them.

I count myself in the camp of people who don't relish the world before fizzbuzz.

If someone gets hired, and on the job in the next year, they do not implement/discuss the interview questions in any non-interview situations, then drop those interview questions. When a candidate fails to answer a question/task, ask the co-worker who asked the question/task to complete it (and be just as strict with them).

Programming without being able to insert lines is not indicative of real programming. To me that's the bottom line.

whiteboarding is essential when hiring for roles where development environment is not productive. there are still huge systems which take 5 minutes to link on fastest machine, or embedded, real-time and distributed systems where it takes so much time to debug that one's daily productivity would be close to zero unless he writes code which would be mostly correct before first run.

developing under such constraints is radically different from environments with high iteration velocity - i.e. browser or scripting.

that said, in many modern environments whiteboarding is completely irrelevant, as modern application level programmer productivity is often ability to do plumbing-and wiring- work between many components and quickly assess why something is going wrong, rather then creating a highly performant or memory-efficient algorithm implementation.

I'll stop short of saying whiteboard interviews, with or without coding, are suboptimal. I will say that any interview question which attempt to see how I think is suboptimal. How I think when on the spot does not reflect how I think on the job.

Speaking for myself, I like pair programming interviews.

I like pair programming interviews too, the only thing is that there's still some murky-area between interviews of how much should you lean on your pair and how much should you prove independent problem solving.

The meta of how on whiteboard interviews.

https://medium.freecodecamp.org/how-to-organize-your-thought...

Just embrace it. It will be okay.

whiteboard tests are always going to produce fuzzy red and white flags. those will come handy later when it gets to assessing the applicant based on sympathy and arbitrary objective arguments are required to back whatever the interviewer wants to conclude.

I've done close to 500 intermediate and senior level engineer interviews for one of the top 3 tech companies. For a time we offered candidates a laptop to code on during their interviews. To my surprise over 70% preferred the whiteboard. Go figure.

I’ve interviewed a lot of folks, and bitter experience had taught me to always ask at least a couple of whiteboard problems. I’ve often been completely shocked that someone I thought seemed promising was unable to think through basic problems.

No in general.

If yes, the problem should be simple and relevant, and the session highly interactive. The assessment should be about (a) How interested the candidate is in the problem (b) How receptive they are to feedback.

I feel like this topic comes up on HN every six months and we all lament the broken state of technical interviewing, and the ardent supporters of all the different styles come in and push their position based on success perceived from anecdotal evidence. I have a hypothesis that hiring a good candidate is mostly luck.

When we view the hiring process in the context of what it actually is, a sales relationship, we realize that the customer (i.e. the employer), is ideally looking to buy services and time from the business (i.e. the employee/contractor). Viewed in this context, we quickly realize that many different things become part of the equation. Think about hiring a contractor for your home. If you happen to understand the job they do, you might be able to assess a the potential contractors previous work, quote and reputation effectively. Most people can't because they generally don't understand what it might take to do a job and have no experience in executing it even if they do. So you either heavily rely on recommendations from others, or go for a lower or higher priced contractor based on your perceived value (cheaper job, w/e the results vs more expensive job and risk mitigation) and hope for the best.

Considering most software development doesn't actually happen at "tech" companies, the reality is most companies are like most people and likely can't properly assess talent no matter what. They essentially get lucky if they get a good candidate or unlucky if they don't. That would lead us to think that maybe the "tech" companies are able to do a better job of it. In essence they do, but not because of their methods. I'd say it's a form of selection bias.

Google became successful and happened to be run by generally smart people technically capable people. As such they were able to attract generally smart technically capable people to work for Google. I'd venture a guess that 90% of people who make it to the live interview stage at Google would be qualified to work there, but since Google has it's pick of the litter they need to create a filtering system somehow. They happened to make it the "whiteboard" interview and similar high achieving Peer companies all tend to do the same thing. It became a thing, and as such all non-high achieving companies followed suit so they can behave like the high-achieving companies. The filter a company like Google applies is only necessary because they have too many candidates. Most companies aren't in a position to apply that filter if they really want talent. So really, they need to get lucky enough to get those who don't interview well with the "tech" company method of interviewing and hope they get some of those prime candidates.

I'm pretty sure the tech talent pool, as with all things in life, follows a Gaussian distribution. If you interview enough people you're bound to find some decent ones, even if they don't actually interview all that well.

I think Google is seeing a positive signal on their hiring process because a) they get a high percentage of the best programmers applying there at some point in their lives, and b) they run de-facto IQ tests on the onsites to pick the cream of the crop from that pool. It might not be the "best" raw programmers (that's impossible to define anyway) but being able to perform well on the algorithm tests means the people in general are highly intelligent.

Exactly. In general, if you do well it means you're probably pretty smart (whether you've memorized or figured it out on the spot). If you don't do well, it just means it's not your type of test most likely. I'd say this creates an opportunity for non-Google like companies to moneyball their teams and get some high quality talent at a discount because they suck at passing the Google filter. If these companies were smart, and considering how much luck is involved in the process anyway, they should look for a different set of criteria to evaluate a candidate on.

I may be an outliner but I prefer them. Mixing presenting with randomly writing things is better than talking for an hour. Makes me less nervous doing two things at once.

Yea* or nay, not "yay" or nay. Damn millennials...

This is one of those things that even if it can be done correctly in theory, it’s often so bad in practice that it’s probably best to just not do it at all.

IMO whiteboard interviews are disadvantageous to shy people. However, that aside, generally in meetings whiteboard could have some benefits if not abused

The Chief engineer at my current place of employment said, and I quote verbatim: "I get a sort of sick pleasure from watching them squirm."

As a senior-level self-taught, I redacted this post because of my views on whiteboard coding. Downvotes are not cool.

Whiteboarding or not, what are some good resources to prepare for a coding interview that worked for you?

Waterboarding and whiteboarding are phonetically similar as well.

I actually really like whiteboarding problems and it is a common and effective real-world problem solving skill.

I still only do OK at whiteboarding in interviews.

whiteboard is more useful for design than code, i give them a computer to code on

There are three goals in interviewing: for the interviewer to gather enough evidence to prove that the candidate is going to succeed in the role, evidence that they will grow in the long term, and to sell the candidate on the position.

A whiteboard interview isn't a good growth indicator, but it's great at leveling a candidate. Moreover, it signals to a candidate that their peers will be at a similar level.

So I'd vote yay. If you pay attention, you can determine someone's proficiency at coding on the whiteboard in about 15 minutes. This is particularly important when you have people who are changing careers or junior or otherwise don't fit the mold.

I'm honestly shocked at the arguments against whiteboarding as many of them seem entirely reactionary.

"But I'm never going to work that way."

Didn't we all do the better part of two decades of formal schooling? We test people like that for a reason: you can carefully focus on the specific markers of mastery of a subject.

I understand why people make this argument: it's a reaction to some of the weird puzzles companies used to ask. But it's never substantiated that this will help the interviewer and candidate make a better decision, so it has the same problem the puzzles did.

"But at work I have Google and StackOverflow."

This is a good point, the questions you ask in a whiteboard question should be based on the fundamentals rather than trivia. But it's a terrible point because it accepts that the interview is just a ritual you go through and this argument is that we should make it fair.

Whether a candidate knows "the answer" is not the point, it's more that they're able to talk about what's going on, have good instincts, etc. If you're an experienced engineer, you've seen lots of people write code, you know what looks like "junior" or "senior" and "guy who thinks anyone can just pick up coding."

"But if I work on a computer I can compile and see the edge cases."

Are you interviewing the candidate or the compiler? Again, this is the ritualistic view of interviewing.

"But at work I have plenty of time to work on things."

Everyone has this problem, so everyone is less able to perform, and the interviewer can simply adjust expectations.

"All of these interviewing "tools" (or tricks) attempt to be time/cost efficient proxies for doing the damn job..."

That's very much the ritualistic view. Other industries have had notions of apprenticeships and so forth for thousands of years, and our problem is that we view recruiting, professional development, etc. as things someone else does, rather than something we take ownership of. Interviewing is a critical business function, it can produce value and thus is not a waste of time. If you treat it as a useless ritual, though, it will be.

Nay. I do not understand the obsession with whiteboards. I've had coworkers that went gaga when they discovered that whiteboard paint was a thing, and they could cover their office in whiteboard surface. Doodles and doodles and doodles all over the place, but not very much working code ever makes its way into production from those offices. Lots and lots of movement and noise and grandiose planning, signifying nothing, too often.

Lock me in a dark closet under the stairs with just the glow from a couple LCDs and throw pizza and requirements documents through the slot, like you're feeding the Rancor. /s

Nay, unless the candidate wants the whiteboard. Even Google doesn't mandate whiteboard anymore: you can write your code on the provided Chromebook. Some people choose whiteboard, though, for reasons unknown. I think whiteboard (or a sheet of paper) is reasonable for what it gets used for IRL: for sketching out what you're going to do, by drawing or writing very high level description, but it sucks for everything else.

I did a interview with Google last summer. They didn't use a whiteboard, but instead had me code in a shared GoogleDoc. Pretty similar issue, limited time, somebody staring over your shoulder, not your usual development environment. Even more frustrating when you _know_ they have a web-based IDE for internal software development but make you write code in a GoogleDoc.

Was that an on-site interview or a technical interview over the phone? I did an interview at Google about 7 years ago and that sounds like the Technical Interview.

When they brought me on-site after that, I interviewed face to face with 5 or 6 people in a row (except for a lunch break), and every one of them asked me a two or three questions and then switched to tricky whiteboard problems.

I remember one was 'write a function that draws a 2D skyline from a bunch of 3D buildings', which I got totally stuck on because there's so many possible configurations of heights and widths and how they overlap that complicates it, so I didn't really know how to properly solve it.

I didn't get the job. I thought the experience was interesting at the time, and basically treated it as a free tour of Google and a chance to eat In and Out burger again, but I don't want to go through that grueling exercise again (or spend a ridiculous amount of time prepping for it... as it was I spent a solid two weeks studying for it the first time around), so I've said no to Google recruiters every time they've gotten in touch with me since.

I might have given it one more shot, though, if I were willing to move to Silicon Valley at this point in my life, but I just bought a pretty decent house and I'm not really wanting to give that up to take out a mortgage five times what I have here just to live in a smaller house in the Valley and be able to say I worked at Google.

With doc you don’t have to stare over the candidate’s shoulder. The interviewer can just share the doc with him/herself and watch as you type. That’s what I used to do. The main goal is actually to make sure the candidate stays on the right track. Some folks are prone to veering way out into the woods otherwise. Unsupervised, they would just run out of time.

I did one about ten years ago, two of the guys asked me to do the program on legal pads, two on a whiteboard, it was kind of funny because the Google HR person gave out a pdf explaining they were going to use a whiteboard and to practice on a whiteboard before coming on site.

I love these kinds of post. It's like saying:

"Highways: Yay or nay?"

"Food: Yay or nay?"

"Air: Yay or nay?"

Hate on whiteboard interviews all you want, but that attitude is going to put you a massive disadvantage in the current rigmarole of software engineering interviews. It's not a thing 99% have any interviewees have control over. Nor is there a way to tell the interviewer: Hey, I prefer not to do whiteboard interviews. It doesn't work like that.

Reply on news.ycombinator.com