Don’t you think it’s silly that Google’s interviews are so ineffective that candidates need to learn how to “game” them? Asking as someone who’s been on both sides of Google’s interview process.
That's an interesting viewpoint, so I'd be interested if you could do a critique of the following claim[1]:
> Hackerrank-style interviews suck and they aren't representative (not even remotely) of real working conditions. But they are good at something: evaluating the tenacity and the drive of a candidate. You need to work hard to ace these interviews, and companies are looking for candidates who are persistent and able to work hard—even on things they don't choose to work on.
To be completely honest, it's actually an intriguing idea. I wonder, assuming some basic level of competency, if you gave a candidate some completely orthogonal task unrelated to the career and they excelled at it (assuming no cheating), would they be a better hire than another candidate?
Maybe! Quite often success at work is determined by getting crap done that doesn't require the skills you trained for initially.
This would also imply that there's probably fruitful hiring ground in picking from anyone inside your organization (where you have the most direct knowledge) who does their job well, no matter what it is.
I'm given to understand companies used to do this.
Let’s take it to another level: the candidate realises that if they kill a current employee with the skills exactly matching theirs, would you hire them (assuming you can’t prove they killed the employee but it’s blatantly obvious) or would you think that’s unacceptable?
I remember a book by Jim Cramer in which he told a story about how he was offered a job "if you go in that office and fire that guy, right now". He did, and was.
Hiring people on the basis of arbitrary criteria opens you up to indirect discrimination claims. If some protected class X just so happens to perform below-average on that arbitrary criteria, then members of that group can claim indirect discrimination. If the criteria have some rational relationship to the actual job, that's a defence that may succeed. If the criteria are arbitrary and have no rational relationship to the actual job, then that defence isn't going to work. (The fact that there was no discriminatory intention behind the criteria, even if true, may be difficult to prove – arbitrary criteria that discriminate against people in practice, and which have no obvious justification, are likely to be presumed to have a discriminatory purpose, and how can you rebut that presumption?)
Interesting choice of words, but unfortunately there isn't really anything you can do about discrimination in a tech job interview in California, especially age discrimination and the authorities don't send secret shoppers into the tech interview process so it continues. I suspect you are over educated (which is not a bad problem to have, except it will cloud your judgment until you dumb down a bit).
skissane isn't stating that there's no discrimination, just that legally the practice of (algorithm-based) tech interviews is unlikely to be found discriminatory in a US court of law. Yes, there definitely is discrimination to be found, although typically it's not as overt as IBM's age discrimination.
Talking about age discrimination in particular, it's difficult since often older hires may not have the shiny new skill on their resumes or may have something indicating that this person may not be keeping up with tech (the bias). One of the weirdest things about joining a startup is realizing that 98% of engineers are in their 30s or 20s, while not having any engineers in their 50s, 60s or 70s.
> By that logic, surely every interview practice is justified?
In a sense, they are. As long as the interview practice is well understood by the applicants, it only leads to filtering the applicants that cannot pursue the requirements for that practice. Granted, the FAANG interview process and filtering only works because the SWE jobs are desirable enough.
I understand that the required practices and filtering also lead to a certain type of applicant. This is the main fault of the system (the lack of diversity potential).
Without the tenacity to build a wooden boat I would never hire a programmer. How would they demonstrate that tenacity? Through a series of open ended questions in an interview.
You can't hire, or even interview, everyone. If you pick some random skill that a reasonable subset of programmers have, then it's entirely plausible that being good at it will correlate with programming skill and supply a population large enough to hire from. But that won't provide you any shelter if it has disparate impact on protected classes.
That may be true but will also effectively exclude strong experienced candidates with options. It worked well for Google in post-dotcom crash frenzy of early 2000s but may not work very well today ;)
> That may be true but will also effectively exclude strong experienced candidates with options.
What options will you have as an experienced candidate which are competitive with FAANG?
I'm an experienced engineer, and my leading FAANG offers were over 50% higher than my top non-FAANG offer.
In fact, my non-FAANG offers as an experienced engineer weren't much better than the offers junior candidates get. So unfortunately, it seems if you're an experienced candidate who wants to be appreciated and rewarded for your experience, FAANGs are basically your only option. Non-FAANGs will pay you maybe 30% more than a junior, and for that they'd expect you to lead a team and work twice as hard.
I moved on triumphantly from my previous programming job where I built a company's product from Greenfield to million a month with awesome performance and features.
Other start-ups in my area are offering me 5-10k more than I was making at my last job to do the same thing again. My counter offer from a large company was 50k higher for a laid back environment. Guess which opportunity I took.
Hilarious to me. One lead engineer is worth three junior engineer at least. I see startups doing the same thing over and over again. Hiring Junior developers fresh out of boot camp who are completely worthless, and refusing to spend money on high quality candidates who are interested.
> One lead engineer is worth three junior engineer at least.
Far more, I'd say. Fresh graduates and juniors are often held to almost no production bars at all. In many startups, you could be a junior for a year, accomplish very little, yet be in no danger of failing any performance metrics.
Before I left the startup world behind, I observed that seniors were often held responsible to 95% of the production that was happening at their startup.
You'd expect them to be paid x10-20 times as much as juniors, but instead it's typically no more than x1.3-1.4.
So as a senior engineer with a track record of shipped projects, I get to work harder than all juniors, be held to x20 their performance metrics, and for that I get 30% better pay?
No wonder every single senior engineer I know has a "FAANG or bust" attitude nowadays.
There’s a number of companies (mostly public, ofc) in the bay that pay competitively. In fact google is known for lowballing unless you collect offers from fb etc
Google is known for "low-balling" if you compare to Facebook. My Google offer was higher than both my Amazon and Microsoft offer by a significant margin, and these offers in turn were higher than almost any other offer I had in the Bay.
The only companies that got anywhere near Google on pay were some very exceptional unicorns.
Google will definitely “lowball” you if you don’t have competing offers. However once you do have them, google is very quick to match them. This is just my anecdotal experience though.
Not in my experience, nor that of my friends who got offers.
My typical friend who got a Google offer, already had a bunch of other offers by the time their Google process finally completed. Then the Google offer came in and it was typically higher than most or all of these outstanding offers. Often substantially so.
However, if you have an outlier offer that is based on some theoretical equity valuation ("an extra $1m in stock options based on our last fundraising round"), Google will not match it.
Frankly, they don't need to when their initial offer is already higher than most other options, and far safer.
Also from what I've seen, Google won't just match other offers, but they do have some specific companies whose offers they will match, like FB. FB is also one of the few companies that may offer you higher comp than Google.
My experience is that FB will offer a marginally better offer (30k or so more than G) for comparable levels, and those other public companies give a vaguely competitive offer that still falls short of G by about the same amount. Though I'd love to see actual data.
Is it necessarily gaming? It seems to me that this a focused opportunity for self-aware and ambitious folks to carefully study what's actually really important to organizations.
Honestly, I think 80% of what keeps candidates from succeeding is an unwillingness to conform their character and personality to the needs of the institution that is recruiting them.
That said, I don't want to be a corporate drone either. There are more important things in my life. Like family, health, my hobbies, etc.
Yes, I know that if I lived the organizational values of my company I'd get ahead. But honest, I'd rather go play catch with my kids than obsess about customers[1]
I think people consider it gaming because hackerrank is not accurately representative of day to day as a FAANG engineer.
That being said I don’t think there’s a really great way to assess that in a 5 hour onsite interview. So hackerrank ends up being the best they’ve got.
Predicting the success of an employee and how well they will perform on a brand new team/stack over the course of potentially several years is very challenging and likely impossible especially when the people interviewing you aren’t even on the team you’ll work on. Hence this is why leetcode interviews are the best they’ve got.
No doubt. That’s my point — if you have a process that’s easy to effectively game with coaching, what’s the point?
A company could probably save lots of time and money by cutting some of the fluff in the process. From the pov of a candidate with a good job already testing the waters, these companies come off as trying to hard to add objectivity in situations where it doesn’t exist.
Saying that tiny puzzles are the best you can do is a very large leap from saying that you can't predict an employee's performance.
Someone's job performance over time depends on lots of different factors, many out of the control of the company. Understanding someone's potential and risk can be the result of conversations about what they have already done and possibly looking at what they've done. No irrelevant puzzles needed.
I don’t think it’s just a few weeks of gaming. Maybe if you do 60 hours a week.
We’ve found that an average undergraduate or masters in CS or STEM field who just knows how to code basic Python will require ~250 hours of total work to become interview ready for FAANG. That’s a lot of time which is why I wouldn’t consider it gaming a system.
"Average" is the starting point. After learning data structures algorithms and how to communicate, they are no longer an average candidate (by FAANG standards)
It's unrealistic to expect someone to spend 30 full time days (6 unpaid weeks) on that kind of material in one go. I've done it myself and gave up multiple times in the process. More than 2-3 days full-time and I burn out immediately. Even doing it just on nights and weekends is incredibly tiring (especially after a normal days work). It's incredibly tiring to do leetcode or similar for 8+ hours/day every day. Especially in the beginning when you start out - because you will feel completely hopeless with the process.
Just FYI - this interview process isn't a one and done thing. You have to refresh yourself at least every year with ~80 hours of prep. Otherwise, you're going to be really bad at it when you search again.
Definitely not one and done! But I doubt you will be searching every year? Maybe once every 5 years hopefully on average. And in the sister comment I mentioned taking unpaid leave to do it, which will be paid by the next company paying you more salary. If I lived in the US where this would make a difference I'd be doing it now. However where I live there is only Google (not FAAN etc.) and they don't pay as much as in the US, so probably not worth it.
I agree, that's actually how I prepared. Unfortunately, most don't have the luxury to dedicate 30 full-time days. Many that we work with our university students with courses or full-time employees. This is why the 250 hours if a pretty big barrier to entry.
Those 6 weeks where you're studying full-time unpaid seems like a lot of time to me. That's 250 hours that you can't spend on what is actually important in your life (maybe friends, family, relationships).
Ok, lets say you earn $80k a year. Google will pay you $120k. So you stand to make $40k in 1 year extra if you get the job. And if you don't get the job the skills transfer to other interviews anyway, so you'll get some kind of job at, lets say $100k.
You ask for 6 weeks unpaid leave. Could be one block, or 1 day a week for 30 weeks. This will cost you $10k gross by my calculations.
Even if you get the $100k job, you are $10k up after 1 year, with no friends and family time sacrificed at all.
As a bonus you were effectively paid for 6 weeks to learn CS skills rather than bash out features and not learn any new computer science skills.
Thanks for the breakdown. I appreciate the chance to see your perspective. When you attach those numbers to it, taking that 6 week unpaid leave makes sense financially. Your comment makes me consider a 6 week sabbatical sometime in the future.
I'm always impressed when I hear stories of quiet, but strong dedication to reach critical long-term goals.
I have a friend that is in his late twenties and is now a partner for a boutique investment shop. About four years ago he made a commitment to pass the CFA 1 in nine months. Studied a little each day after work, and then studied more on the weekends. 180 hours total. Passed the CFA 1 without any fan fare. Hardly told a soul. Did the same for the CFA 2 the following year. Didn't tell any colleagues. No fan fare. Boss was pleasantly surprised when he listed the new qualification on the firm website.
He does the exact same thing the following year for the CFA 3. His wife tells someone at the firm, and when he walks into work they throw a surprise celebration for this very impressive accomplishment (many people spend a decade reaching CFA 3).
A couple months later the managing director moves to have the firm make him a partner. Easy decision. You had a junior analyst quietly dedicate themselves to substantially improving their skills.
Now, while his salary is certainly quite impressive as a partner and for someone that young, the true freedom he has achieved is being given responsibility of a new order of magnitude to establish things he wants to establish in that world (internal values maybe, many new types of clients, new employees he can then bring up through the firm, etc.).
Point is: good on you. You're a lesson for most of us.
I used to work as a SWE at Google, and we sometimes "joked" that the process ("here's a list of computer science topics, we expect you to know all of them, take as long as you want to prepare") was designed to identify insecure overachievers.
> Don’t you think it’s silly that Google’s interviews are so ineffective that candidates need to learn how to “game” them?
Don't you mean "effective"? I assume that given the ratio of applicants to positions, the primary purpose of the interview process is to eliminate as many unacceptable candidates as possible, not to identify as many acceptable candidates as possible.
The FAANG interview is meant to reduce the number of false positives, however, the cost is having a lot of false negatives. This isn't really a problem for them since they have so many applicants.
If by “effective” we mean “deny most everyone” then yeah it’s very much so. If we mean “hire people right for the job” then no one actually knows this because a test would require control group. I would argue that if some leetcoding and other rote can make a significant difference it’s probably not effective.
