You are 100% correct that most research into interviewing has show it to be far less predictive than companies believe. However, I'll add that no one is doing the false negative studies that would really be required to show this. For example, Google has shown that among people who pass their interview GPA is not correlated with success. This does not at all mean that GPA is not correlated with programming ability. In fact, it almost certainly is. People with low GPA who pass google interviews have a lot of other things going for them. To really analyze this, you'd have to give a job to a random sample of applicants. No one has done this. My goal is for Triplebye to get large enough that I can.

Not for coding specifically, but there have been some studies correlating GPA with career success, and the correlation has been found to be very low. But the reason there are very few studies that look at this is that GPA A) isn't epistemologically meaningful B) isn't a measure in the mathematical sense.

"A grade can be regarded only as an inadequate report of an inaccurate judgment by a biased and variable judge of the extent to which a student has attained an undefined level of mastery of an unknown proportion of an indefinite amount of material." -Paul Dressel

Why do you think the programmer from your article didn't get a job when you "sat back to watch him get a job"? You didn't tune your test to what the company was looking for.

Any YC founders who are reading this: Refuse to talk to Triplebyte until they set up a work hire test for your company. Force them to do it, and force them to work with you to tune the test to the work you need. You will get a 100% hit rate. And you'll notice something else: Your retention rate will go way, way up. Want to not deal with firing people? Get them to show you they can do the work. Nothing else matters.

The test needs to be set up so that the candidate can demonstrate they can do exactly the same things all other employees do on a day-to-day basis, or it's not a work-hire test.

If only it were that easy. I am not at all opposed to work-trial tests. They are almost certainly more accurate. But a high percentage of programmer will never apply to a company that requires one. Read any HN comment thread about the topic. It's very controversial. Lots of programmers are against it. Work-trials require MORE time upfront from the candidate, and of course still often lead to rejections (by design a hiring filter rejects most people). A failed work trial burns an entire week of vacation. An unemployed programmer (who is being paid the work-trial) may like it, but someone working another job may not.

sillysaurus3: "Spend a couple hours remotely fixing some bugs and adding a feature on a fake iOS app"

ammon: "A failed work trial burns an entire week of vacation."

You're not talking about the same thing!

sillysaurus3 is talking about doing a small amount of work that is similar to what the company does, under conditions similar to those you would have as an employee. You seem to be talking about short trial period contract-to-hire. Yes, the thing you're talking about sucks and is a huge turnoff to candidates, but (mostly speaking for myself here) what sillysaurus3 is suggesting is massively appealing.

But if your biggest objection is that it will take additional time, that depends entirely on the nature of the trial. I was given a work trial at my first job and it took about 40 minutes. It didn't go as smoothly as I'd hoped, but I would much much rather have done that than spent 40 minutes answering brain teasers in front of a white board.

If I'm already taking a day off, I don't care if the interview lasts three hours or five. I'd rather work on a collaborative task with my potential teammates to see if it's going to be worth radically altering my life to take a new job. I understand that there are many programmers who are opposed to work trials, and they can say "I'm opposed to work trials and would rather demonstrate my abilities in another way."

It seems like your research so far has been very comprehensive, why dismiss an effective technique because some people are opposed to it?

Ammon mentioned Weebly, above. They are one company that I've seen require an on-site week as a contractor. I know of a Weebly candidate who lost her current job because the employer considered the leave to be job abandonment. Thankfully, she got the job at Weebly. Perhaps Weebly's weeklong trial is effective for them at weeding out bad hires. But I suspect most good programmers would never consider giving up a week of their lives to an extended job interview.

And I wasn't suggesting job abandonment. Asking for a week of unpaid vacation is a different conversation than asking for a week of paid vacation.

Only if they don't need you and think you are not being productive. So be needed and be productive.

> This was a tough decision and one that we were on the fence about. We really appreciate you taking the time to work on the take home project. We're aware this requires a substantial time commitment and we are really grateful that you invested the time in completing it. We thought you wrote a great, very full featured [trial project]. It was especially impressive how much you dug into the academics behind [the project].

> However we made the decision because we felt that while going through the project together during the interview, we didn't see the fluency of programming when adding to it that we had hoped for. While we specifically designed the take home project track to help overcome the difficulties of coding under time pressure with someone watching, we do still need to see a certain level of programming during the interview. This didn't seem to be the case here, where making changes to the project seemed to be slower and more difficult than we'd have liked.

tptacek had some advice for hiring-through-a-work-test that I (from my armchair) agree with, which is that it's important to give everyone the same work test so your applicants are comparable. There's a little bit of tension (I don't think it's unresolvable) between that and "The test needs to be set up so that the candidate can demonstrate they can do exactly the same things all other employees do on a day-to-day basis, or it's not a work-hire test".

A friend of mine's company usually sends out a little canvas-based browser paint app with a bunch of features, and asks them to implement a few things, like flood fill. It's really enjoyable.

Unfortunately, they then flew me out to their office, whiteboard hazed me, and sent me home. Called me a few weeks later to tell me I didn't get the job, but that they sent me a hoodie in the mail (felt like a consolation prize).

It's a pretty comfy hoodie. They spent that VC money well.

I identify with the "product engineer" concept they outlined, I want to talk UX and user testing in an interview not prove I can understand someone else's code base. Honestly many companies will get far more value of a programmer who has user sympathy and writes user-friendly applications than one who writes a fully tested, bug free and technically sexy app which users don't want to use.

I'm not the best programmer, but I'm comfortable I'd slot in the top quintile, probably top ten percent, of people that walk through a startup's door. And I'd never even return your call. No work-sample test I've ever seen would take less than four hours. That's a $550 opportunity cost for me at my standard rates. What the hell makes you think you deserve that for free? You aren't Google and you aren't Facebook. You need me more than I need you. And almost everybody else who's like you--and given the way you are acting, probably you too--will forget about me or blow me off at some point in the process, wasting my time even further.

On the other hand, an in-person, discussion-based interview isn't work and isn't priced as work; not only does it tells me about the company and whether or not I actually want to work there, but I enjoy meeting new people in a professional setting (there are multiple companies where I've turned them down, but have become friends with people I've met through the process!).

Look at my Github and decide if I can hack if you want, that's why it's there, but I don't work on effing spec.

I'm very old and out of the loop, I have no idea what modern work-sample tests look like. I remember when I did one many years ago it was to the effect of "Write a script that checks to see if a URL returns a valid response code, and trigger an event if it does not". My prospective employer liked my coding style and I got hired. What was so wrong about that?

As I recall, they intended to modify the script and use it in production whether they hired me or not. I also don't care about that. It took less than an hour, and it seemed like a valid form of evaluation. Maybe it's a pride thing, but I had no problem with it and I think a lot of other people would have been ok with that practice as well.

If a company wanted to pay you for that work-sample test, that's different. But they don't, of course, because that costs them money instead of costing you. There's no need to let profit-seeking entities use you for nothing. You just don't need to be their monkey.

You made this exact case yourself in a related comment, in which you called meeting people business development. I agree there's a non-zero distinction here between work-hire-test and consulting-about-consulting, but claiming that the issue is that "you shouldn't work for free" is misleading.

Also, you suggest that if they pay, that costs them instead of costing you. It already costs them more to interview than it costs you. But sure, if making sure they don't get any benefit out of their relationship with you unless you get benefit too, tell them you'll only do a work-hire if they make a donation to the EFF in the value of your hourly rate.

Of course. That's the sit-down-and-chat "interview". But lawyers don't draft up a contract for you to demonstrate that they can write up a contract to your satisfaction. If you want a lawyer's expertise applied concretely to something of your direction, you pay. And while lawyers do have bar exams, they don't have a very detailed demonstration of their work up on Lawhub for your perusal!

> I agree there's a non-zero distinction here between work-hire-test and consulting-about-consulting, but claiming that the issue is that "you shouldn't work for free" is misleading.

I think it's misleading only if you consider the beneficiary of that work to be the same in both cases. I don't. The only beneficiary of every work-sample test I've ever been given was the company--it goes into some black hole and whether it was even good or not, to say nothing of any actionable feedback, has literally-literally never been forthcoming. On the other hand, I have very rarely not benefited in some way from sitting down and chatting with a technical leader at a company, whether it was directly about their problems (or their solutions!) or about tech in general. Both in a consulting-about-consulting capacity and an interview one.

> Also, you suggest that if they pay, that costs them instead of costing you. It already costs them more to interview than it costs you.

In an absolute sense, it definitely costs them more. In a relative one, it emphatically does not: they've got plenty of bodies that can parallelize. I can be interviewing with them, or I can be working, or I can be mopping my bathroom floor. I can't be doing all of those things at once.

*The primary beneficiary being you, if you're a top-10% programmer: It gives you a better chance to accurately show off your ability, which is probably worth thousands of dollars in salary.

But the alternative--it's not the whiteboard, it's the meeting people. Right now, I consult, but I do take interviews for "permanent positions". But we all know they aren't actually permanent positions. The last thirty years of corporatism have made this loud and clear: we are all expendable tools. and I go into them with that in mind: a W-2 doesn't mean I'm anything other than working for a client under a tax-advantageous status. Interviews are business development; I am effectively offering my exclusive services on retainer to a single company. So meeting people to discuss the gig (whether or not the whiteboard comes into play, the only interview I've been on in the last two years where I was both interested in the company and wrote code on the board was Google) benefits me, too: maybe it's not a fit, but maybe I meet somebody I'll remember as "hey, I'd like to work with them again." Or "hey, I don't want to work here, but I know somebody who will, so I can do them a solid and they can help me in the future."

So, yeah, it is work. It's just business development, which is a different, and valuable, kind of work. And it applies to everybody, not just consultants; it's the game we're all playing even if one doesn't realize they're playing it.

(Disclaimer: I strongly prefer to quit my current job before looking for new jobs. If that weren't the case, I'm sure I would be more sensitive to anything that takes time, but I think I would still feel this way.)

Look at the opportunity costs: what if you could spend the time they want you to give them, for free, building something not only personally creditable but maybe even generally useful on Github? If you're good enough to show off, as you suggest, then your time is valuable enough that it should be respected. (A work-sample test that's a useful, valuable problem and can be open-sourced? I'd be down for that. But that would be haaaaard.)

If companies were cold emailing me on LinkedIn asking me to do their two hour project, I would not do it, but neither would I bother going to interview with them. Given that I'm already cherry-picking which companies I care about, I don't mind investing some time to make them care about me.

EDIT: I agree that it would be nicer to spend time doing something really useful. I hope that if I handed them a FOSS thingy that was representative of my ability, that these same companies would respect that in lieu of their work sample -- although standardization is valuable for evaluation purposes, so I can understand if they wouldn't. (I just did the projects they suggested because I thought they sounded fun.)

(Right now, I'm helping out a few days a week over at a fairly large Boston startup, and while they have a general work-sample test that they roll out, I never took it. I probably wouldn't have followed up past the initial phone call if a monkey dance was necessary to get in the door.)

It was fun, and more profitable than a free boring discussion-based interview.

Do we care about equality, or not? I tried not mentionining this aspect, hoping people would realize on their own. But a remote work hire test is also mostly anonymous. It doesn't matter whether you're black, white, male, or female. All that matters is whether you can do the work.

On the flipside, what you're saying is that you genuinely want to spend a vacation day meeting a new company instead of with your family or working on your own projects.

And it's like, if you think you're a good dev, why wouldn't you leap at the opportunity to show it off? I get that it's a little annoying to spend a few hours on it, but the standard interview is literally random noise. Why subject your future to a random process?

I don't know. I respect your view. I'm going to bow out now. Have a good week.

And, FWIW, when salaried, I not once have taken a vacation day to interview. I've said "hey, I'll be in late," and because this industry is so deranged as to think that 50+-hour weeks are normal, no manager has had the temerity to get mad at me for taking a morning off because invariably it will be cashed in when I have to pull a sixteen-hour day to deal with a problem. (The days around Heartbleed earned me some goodwill at that gig, if you follow!)

The typical cost of hiring a programer is way higher than that, more like $30k I'd guess so they could always try paying you $550 to do something semi real? I'm not experienced but I've got a friend who hires remote workers and he says the only way to tell if they are good is to actually hire them to do a small job.

Not really. The best programmers have other stuff to do besides your make-work silliness.

"Get them to show you they can do the work. Nothing else matters."

This is even more silliness. Suppose you have a diverse team, and the a candidate comes up, and passes the test, but is Donald Trump. Are you then going to say that only the work matters?

All 3 ended up being very valuable and 2 of them stayed on well beyond their internship.

Oh, and these projects were done from home before we ever brought them in for an interview and we hired them before ever meeting them.

The effort required to administer an on-site work-hire test is non-zero, therefore I cannot administer such a test to every applicant.

I therefore need a way to determine who to bring on-site, in order to administer such a test. I cannot phone screen every single applicant due to the cost involved.

That process could also be a work-hire test of some sort (e.g. a remote coding project), but regardless of what has been said on this thread, many good applicants will drop off at this step. I know this empirically, because I've experienced it, many, many times.

Additionally, the people who tend to drop off because of this extra effort tend to be senior applicants, who often have multiple offers from multiple companies due to the competiveness of the current hiring market. It also disproportionately drives away passive candidates, who are often the best candidates, because the best candidates are often not looking for work (since they are good, people who have worked with them previously want to work with them again, and they get poached).

So I need some method of sourcing and filtering candidates down that is non-intrusive to both our development team AND the applicant. This is the reality. This simply has to happen in a startup's hiring pipeline.

Currently, most companies do this by looking at resumes. That is obviously sub-optimal.

Any suggestions for alternatives?

EDIT: spelling.

Try segmenting potential hires into different categories and using different strategies for each group. You should also clearly communicate the process, so they don't get the wrong idea.

>Additionally, the people who tend to drop off because of this extra effort tend to be senior applicants

Is this before or after the phone screen? If they think the work-hire test is going to be in addition to a phone screen and an on-site whiteboard session, they'll drop. If they're senior devs, you should be able to tell from their resumes. Do a phone screen, and if they pass, explain that they can either do a work-hire test or a whiteboarding session. Explain that the work-hire test should be doable in 4 hours, so they don't envision a week long project. Also explain that it's a substitute for the whiteboard portion of the on-site. People who tend to hate whiteboard coding will choose the work-hire test and vice versa.

If you're trying to find diamonds in the rough among junior talent, sending out a first-pass work-hire test might work. You'd still get better results if you explained that it was short and a substitute for the onsite whiteboard session.

Set up a system that can spin up a droplet for a remote candidate. Any time a candidate expresses interest in your company, spin up an instance and email them a link to it.

What does the link do? That depends on your company. Are you making an iOS app? Then the link takes them to where they can download source code for a fake, hypothetical iOS app. It says "X, Y, and Z bugs exist. Find them and fix them. Then add a feature: here is a clear description of what to add."

When the candidate is done doing this, they zip up their code and send it back to you.

If it sounds way more effective to look at that than to look at resumes, it is. If it sounds like it will repel candidates, well... Two things. First, if you're chasing a specific developer, then that isn't really the normal hiring process. You want them already. This pipeline is for everyone else. It makes no sense to subject them to a work hire test when you're actively seeking them out.

Here's the other point. The type of candidates you will find with this method will shock you. They will be so skilled that it won't matter whether they're called a senior or fresh out of college. You'll know immediately that you want them.

Everything I've described up to this point is a remote process. There is no on-site work hire test. By the time they come on site, you're mainly checking they can show up, and telling them about your company. You're no longer trying to filter them based on ability; they already demonstrated it.

Let's say your company's website is the primary focus, not an iOS app. Ok. The link will take the candidate to a hypothetical, fake website built with a similar framework. Again, it will have multiple bugs and a missing feature. Tell them what the bugs are, and tell them what the feature needs to do. Then have them send you their code when they're done.

I feel like at this point no one will even try to do this. You can think of so many reasons not to try: it takes too much work, it will scare too many people off, it will... Etc.

These reasons turn out to be largely fake or mistaken. Try it. Invest the resources to build this pipeline, tell HN when it's ready, and you win.

If this sounds prohibitive or unlikely, remember how counter-intuitive the most effective techniques in life are. Penicillin was discovered by accident. It sounds pretty unlikely that it would work. Same deal here.

I've explained this as clearly as I can. It's up to everyone else to either try it or to watch others win after they try it. Because the filter I've explained is the only way to let talent find you.

The type of people you'll discover will range from passive people who found the process amusing, to well-off senior developers who are demonstrating why you should pay them X equity or Y salary, to high school dropouts who turn out to be one of the most valuable people that join your team.

I'm not even going to touch the topic of what tech companies currently do. It doesn't matter. I've described what works, and if whoever reads this suppresses their instincts and builds this, they will discover it's practically the key to winning.

I don't hire programmers, but have been a hiring manager, running fairly technical business teams now for 10+ years, and have found that initiative is often the deciding factor in any given employee's success.

There are other ways to select for it, and clues you start to pick up on over the years, but once you've figured out how to gauge relative levels of initiative across different people, I've found it makes hiring/resourcing questions quite a bit easier.

We've evolved the same process in my own startup - in the last 2 months we've hired 5 remote programmers in Vietnam and India by giving them 2 rounds of remote work-sample tasks (a simple web app), then ending off with a text or voice interview. We ended up retaining 2 out of 5 of them (both Vietnamese, incidentally) who demonstrated a higher level of ability after they started.

The retention rate is low, so we've been tightening up the process. What we've found is that we probably needed to raise the bar a little by asking candidates to design and architect a feature (DB-backend-frontend) during the chat interview. That would have given us a better insight into their code structuring and teamwork/communication skills.

We've just brought on a 3rd hire (Vietnamese again) who passed our improved interview, and I have a good feeling that the process works very well to weed out candidates who don't have the needed level of ability and scrupulousness.

The best of our hires didn't attend college or have much relevant experience, but was clearly very capable and possessed great initiative. We're really glad to have managed to hire him. Another guy is finishing his 3rd year of college.

In summary, I think remote work-sample tests are fantastic, and I hope you're right that they're the key to winning :)

Also a lot of experienced programmers are generalists as well. You might have been C++ system guy in your past life, worked on JavaScript couple of years back and now working on Hadoop backend. You may not remember all the nitty gritty details of past platforms and tools but you can ramp up on demand.

So this 2 hour work-hire test that narrowly focuses on specific tools and platform won't work for vast majority of programmers who haven't branded themselves in to one narrow area. Also remember that a lot of programmers at places like Google, FB and Microsoft work in highly proprietory environment with their own internal build system, platforms and tools. They are likely not familiar with everything that's latest and in fashion. However they are very smart people, can change gear easily and, most importantly, ramp up on demand very quickly.

Yet another big issue is that work-hire test usually fail to measure hardcore computer science problem solving abilities. Sure, you are good at fixing some iOS bugs but can you figure out how to do driving directions on maps or may be scale email system to 100 million users? Now I do agree that lot of jobs don't require hard core CS but many do. And for a startup, it's usually impossible to tell if your future 6 month down the line will hing on having someone with hard core CS skills.

As far as hardcore computer science abilities, a test still won't help, because what you are really testing is how far people are from college: Generalist cs is touched in college, but there are entire sections of computer science that someone with 10 years of experience might not have had to touch. I've had to do pretty specific CS stuff for work, but to get any of that done, I started with weeks of research, and then add my own spin to our specific situation top. You can't test the ability to do that by just checking if someone has memorized Raft, or black red trees, or any other random problem. I don't have a lot of CS algorithms memorized: that's why there's the internet, and why I have Knuth in my shelf.

You want to check for hard cs, ask the candidate to choose his favorite CS topic, and ask for am impromptu presentation on that. anything else is just going to fail at finding candidates that have different backgrounds than you, and that's PRECISELY the people you want to hire.

Isn't that pretty common among incubators and VCs? It's not the only factor, but for first-time founders, being from Stanford is a large plus when it comes to getting in the door. If you already have a track record then of course they just go by that. But when it's your first company, I don't think the "startup ecosystem" is any less reliant on these kinds of signalling credentials to filter the large volume of people who want funding.

So.

I'm assuming if it was a fake problem, it wouldn't be something the other company could actually use.

One thing to realize is that your employer can't really retaliate. The hypothetical iOS app I mentioned is set up for one purpose: to let candidates show they can do work the company is looking for. Not only will the code be thrown away, but it wouldn't make any sense to use it.