Google's “Director of Engineering” Hiring Test

Disclaimer: I also work for Google, opinions are my own, etc etc.

> "i always take "transcripts" of interviews (or anything else) with a grain of salt"

I mean sure, a single instance of this might be overblown, exaggerated, or false in some way.

But there is an avalanche of reports like this, to the point where it's become widespread industry insider knowledge.

I enjoy working here, but the interviewing practices are such that I actively warn friends applying/being referred to temper their expectations of a repeatable/reliable process.

Most colleagues I've spoken to about this, including myself, have strong doubts we would have made the cut if we interviewed again - even though all are strong engineers with great perf records.

At what point do we start taking reports like these seriously? We don't have to accept every detail of the reporting as gospel, but there's clearly something here.

The problem with Google's interview methods is that they all select for a very specific type of programmer: heavily math oriented, deep knowledge of obscure Computer Science theory, but not one test on knowledge of languages, architecture, design, or actual real-world problem solving. I walked into an interview with one guy and he literally did not even say hello: he just jumped straight into some problem I had to solve on the whiteboard.

The problem with that approach is you end up with a very homogenous team of really smart, logical people, but without the balance of more creative, empathic types. Ideally, a well-functioning team will have both, and will have people from many different backgrounds and educations, because that's when you get true collaboration and innovation: by mixing unrelated disciplines.

(Same disclaimer)

If you interview frequently, at least for SWE, this is certainly not how we go about things. ghire guidelines for SWE don't allow for questions like this, or behavior like this.

Is it possible this was an SRE interview? I guess, but it really sounds ungoogly and these questions sound like they don't give great signal. I'd be ashamed if this is how we hire SREs.

Is there really an "avalanche" of reports like this? Most negative reports I hear have to do with our SWE questions which tend to be difficult.

> At what point do we start taking reports like these seriously

My guess is when the number of applications per position actually drops far enough that the false negative rate starts to hurt.

Until then, an interview processed optimized for avoiding false positives at all costs will persist. Totally makes sense for a company worth hundreds of billions though, can you imagine if they had a few more bad hires sneak in? Oh my god, it would destroy everything.

You should be ashamed then, because these are definitely the questions used on SRE phone screens.

That depends on if the interview process actually does protect from bad hires.

This explains why Google is incapable of attacking any product that requires an understanding of humanity to be successful rather than just raw data. (e.g., google+, youtube comments, etc)

> "The problem with Google's interview methods is that they all select for a very specific type of programmer: heavily math oriented, deep knowledge of obscure Computer Science theory"

I'd be marginally okay with it if the interviews actually selected for this sort of engineer! I've seen multiple people who fit this description to a T who flunked the process, hard.

If the goal here is "pick the hyper-mathy, deep-CS types out of the crowd" I'd argue the process isn't even very good at that.

(Standard disclaimer, speaking for myself)

1. Your interviewer didn't give you a good interview or follow guidelines. In interview training they tell you the first thing you must do to start an interview is to ask if the candidate would like to get some water / use the restroom, then break the ice before starting any questions (applicable also during phone screens).

2. Proper interviews actually are supposed to lean heavily toward real-world problem solving approach rather than arcane knowledge. For example, when I interview I look for rational decisions at every turn (not a random example but considering boundary cases, adding a new example to help you visualize the solution should give information gain rather than be something random). My questions are not math oriented, nor do they require deep knowledge of obscure theory. Based on what questions my coworkers ask, I know at least for my team this is not a correct characterization.

What we do test for: understanding of fundamental data structures and algorithms, ability to thrive in uncertainty (ask clarifying questions! state your assumptions!), ability to break a problem down and solve it from first principles.

Good interview questions are required to have multiple solutions.

And then you have the generalization at the end about creativity and diversity; in my limited experience we seem to get pretty decent diversity and even if there is some homogeneity (we need more women and minorities) it's certainly not the kind you described. No, it's not a bunch of mathy theory wizards writing code at Google, it's way more diverse than that. Not perfect, but not awful like you're describing.

Thanks, I am realizing now that it was SRE. All I can say is I'm definitely a fan of how we interview SWEs and I'm sorta bummed this is how SRE interviews go. TIL.

In my (albeit anecdotal) experience, incompetence is the norm rather than an outlier in BigCo SV land.

Frankly, if you are more qualified for a position, chances are you will be rejected because your interviewers will fear for their own job security.

I've always found that type of logic strange, though. Wouldn't you want someone who was better than you currently are on your team? Wouldn't you be able to learn from them?

> but not one test on knowledge of languages, architecture, design, or actual real-world problem solving

Maybe they fail to do so but I do believe the goal is to test real world problem solving. However, I think they stray from specific language or domain knowledge because they want you to be able to work in different roles, since you don't have to interview again to switch teams.

From what I've read, the idea is to hire people who would be smart enough to learn any specific domain knowledge necessary, because the expectation is engineers might have to tackle problems they wouldn't have seen elsewhere. I don't really know whether thats true anymore as my impression is now Google just has a bunch of overqualified people though..

> The problem with that approach is you end up with a very homogenous team of really smart, logical people, but without the balance of more creative, empathic types. Ideally, a well-functioning team will have both, and will have people from many different backgrounds and educations, because that's when you get true collaboration and innovation: by mixing unrelated disciplines.

Can't disagree with you there, but its a weird assumption to say that people who are logical are not creative or empathetic. I do think that they hire for "Googliness" whatever that means, which may lead to a monoculture though.

In any case, I guess you can call me a Google fanboy. I don't agree with everything they do but I feel like bashing Google's (or most other company's) interview process is the cool thing to do here, but most people don't seem to have tried to understand why it is the way it is, and thus don't offer any true alternatives that meet the same goals nor do they reject the goals in the first place.

The questions are fine for SRE. The problem is the behaviour and the expectations of the interviewer.

The people doing these interviews are non-technical people who read off of a cheatsheet. The cheatsheet covers alternative answers, but a situation like the OP describes can never end well.

I got rejected just this month and I can certify that there was no crazy bullshit in the process. I mean, I feel like you made a mistake rejecting me, but I also can imagine a valid process behind the scenes which would reject me based on my "ok but not great" performance. I do hear anecdotes from people I trust which sound crappy (being asked very specific technical questions on subjects that candidate doesn't have experience on and not being flexible about it, being rude etc).

It tells more of the interviewer than the interviewee. They don't ask the questions you mentioned because they don't know or they don't know how to best judge the answers. They asked the questions they know well, that show you what the breadth of their knowledge is. It's like the saying A-players hire A-players while B-players have a hard time judging A-players.

these are just lack of good product management, nothing to do with engineers. I do suspect Google's PM culture is... lacking.

It happens all the time Google acquires a new company, those employees aren't going through these crazy interview processes.

Google turned me down a couple times before I got hired. The first two were definite mistakes (false negatives) where I should have been hired. Just treat it like a process to be optimized. You can reapply every 18 months.

I got pinged by Google about a year ago (after being narrowly rejected 9 years ago) asking if I'd be interested in re-applying. I said, "Sure, why not?".

I was immediately asked which department I wanted to join and why. I said, "Err, not sure, how about SRE?". To be told, "Oh, well that's not my area, let me ask them."

Shortly after that I got a curt message saying "Thank you for applying to Google. We have no vacancies that would suit you right now, thanks for applying, goodbye."

Somewhat bemused by the whole process (you contacted me, dude!), I went about my day.

actually, acquisitions normally trigger full interviews- see "Chaos Monkeys" for a description how this happened at Twitter/Facebook.

B-players know an A-player when they see them, they don't hire them because they feel threatened.

Off-topic pet peeve but why is "OK" now apparently spelled "okay" these days? (especially in bandwidth-limited situations such as SMS or IM). OK is not short for "Okay", OK?

I got these questions when being interviewed for SWE and SRE. My answers suggested to the interviewer that I should go down the SRE route. I passed in the end, because I felt the culture regarded everything as a technical problem (which has worked out for them so far) and Google was a company where you did things their way, and don't rock the boat.

That only applies if you want to learn. If you just want to coast and be "the best" at something at your company, you dont look for people better than you. At best you look for people that are better at the tech than you, but who are passive or easily browbeat so you can claim their work as your own.

It has always been an alternative spelling.

Having a high degree of false negatives doesn't mean the positive signal isn't reliable.

Is this an US thing?

I went through three acquisitions and never had any interview, besides the set of meetings for each of us to decide to go along with the acquisition or get a severance package.

Are SWE and SRE interview the same? I thought they were different enough job descriptions that it would require different interview questions.

I think "okay" looks better than "OK".

But it is ...wrong. OK originated as an abbreviation so why spell out the pronunciation of the letters? Makes no sense to me.

"These days"? It's been spelled that way for nearly 100 years.

> Spelled out as okeh, 1919, by Woodrow Wilson, on assumption that it represented Choctaw okeh "it is so" (a theory which lacks historical documentation); this was ousted quickly by okay after the appearance of that form in 1929.

http://www.etymonline.com/index.php?search=okay

Not so; as you can see in my other comment's link, we can cite OK about 90 years before we can cite "okay", and more reliably than that we can cite the alternative spelling "okeh" to 1919, which establishes pretty well that "okay" was not standard then.

Because the connection to a 180-year-old fad for misspelling "all correct" is so obscure and non-obvious that it was lost long ago? Do you think we should be capitalizing LASER, too?

LOL. That's happened to me before too. And I've been on the other side, where I reach out to people to ask if they'd be interested in a role on my team.

However, I immediately tell them the role that I have, to avoid the whole, "We have no vacancies that would suit you right now" answer. Seems like their recruiters should have done that up front, to save you and themselves some time.

YouTube is pretty successful (even if they acquired it).

Google's interview process, and interview processes modeled on it, do not select for "math-oriented CS-conscious" engineers.

These processes select for recent CS graduates from a handful of universities where Google expends recruiting effort, and anyone not from that background mostly only gets in by blind luck or by knowing someone already in Google who can navigate them through getting hired there.

I think it all depends on the requirements of the company doing the acquisition. I've gone through several myself and did not have to re-interview however I have heard of a employees claiming Google required them to re-interview which has got to be nerve racking considering Google's heavy focus on avoiding false positives a lot of good people don't pass them.

by talking about this you've wasted all the bandwidth your two bytes would have saved a year.

As far as I understand this has been common vernacular since before my lifetime, I'm not usually one to welcome evolution of the base language but this one is before our lives we need to let it be.

If that's true, making non-technical people conduct technical interviews is also a pretty big failure.

They are different, but a lot of the skills are similar. SREs need to problem solve and while they may need to have more domain-specific knowledge I'm not sure a facts quiz administered by someone non-technical is the best way to do that.