Startup Interviewing is Fucked

January 28, 2016 holman

Silicon Valley is full of startups who fetishize the candidate that comes into the interview, answers a few clever fantasy coding challenges, and ultimately ends up the award-winning hire that will surely implement the elusive algorithm that will herald a new era of profitability for the fledging VC-backed company.

Most startups have zero users and are a glimmer of the successful business they might wind up being some day. But we’re still romanticizing the idea that programming riddles will magically be the best benchmark for hiring, even though technology is very rarely the cause for any given startup’s success.

Know what you need

There’s such a wild gulf between what gets asked in interviews and what gets done in the gig’s daily grind that it’s a wonder how startups make it out of the initial incubation phase in the first place.

I’m a product engineer. I don’t have a formal CS background, but I build things for the web, and I’m really good at it. Not once in the last ten months that I’ve on-and-off interviewed have I ever seen anything remotely close to a view or a controller or even a model. Not every company has insisted upon using programming riddles as a hiring technique, but the ones that do almost exclusively focus on weird algorithmic approaches to problems that don’t exist in the real world.

Interviewer: How would you write a method to do this operation?

Me: writes a one-liner in Ruby

Interviewer: Okay now what if you couldn’t use the standard library? Imagine it’s a 200GB file and you have to do it all in memory in Ruby.

Me: Why the fuck would I do that?

Certainly there are some jobs where being extremely performant and algorithmically “correct” are legitimate to interview against. But look around: how many small, less-than-50-person startups are doing work like that? The dirty secret is most startups for the first few years are glorified CRUD apps, and finding well-rounded and diverse people who can have the biggest impact tend to be the ones who are comfortable wearing a lot of hats.

My favorite few tweets from this week talked about this:

Worry more about whether you’re self-selecting the wrong people into your organization.

Power dynamics

A huge problem with all this is that it creates a power dynamic that virtually all but assures that people who are bad at technical interviews will fail.

Algorithm-based challenges typically come from a place where the interviewer, in all their self-aggrandizing smugness, comes up with something they think demonstrates cleverness. A reliable bet is to try solving it with recursion from the start; it’s goddamn catnip for interviewers. If that doesn’t work, try doing it all in one pass rather than in an O(n) operation, because the extra 1ms you save in this use case will surely demonstrate your worth to the organization.

When you come at it from this perspective, you’re immediately telling your prospective coworker than “I have a secret that only I know right now, and I want you to arrive at this correct answer.” It becomes stressful because there is a correct answer.

Every single product I’ve built in my professional career has not had a correct answer. It’s more akin to carving a statue out of marble: you have a vague understanding of what you want to see, but you have to continually chip away at it and refine it until you end up with one possible result. You arrive at the answer, together, with your teammates. You don’t sit on a preconceived answer and direct your coworker to slug through it alone.


This is why I so strongly advocate for pair programming at some point in the interview process. Take an hour and knock off whatever bug or feature you were going to work on together. Not happening to be doing anything interesting today? The bug is too “boring”? Cool, then why are you working on it? If it’s representative of the real work that the candidate will face in the job, then it’s good enough to interview on. Besides, you can learn a lot from someone even in the simplest of fixes.

Build something real together. The very act of doing that entirely changes the power dynamic; I cannot stress that enough. Whereas previously you had someone struggling to find out a secret only you were initially privy to, you’re now working together on a problem neither of you have a firm answer to yet. Before it was adversarial; now it’s collaborative. It’ll put your candidate at ease, and they’ll be able to demonstrate their skillset to you much easier.

No one has any idea what they’re doing

I’ve heard — and experienced — so many things happening in tech interviews that are just bonkers.

You have stories from people like Max Howell who get rejected from jobs ostensibly because he’s not a good enough developer to whiteboard out algorithms, even though he built one of most popular tools for software developers today.

I interviewed for a director of engineering role last year for a startup with famously massive growth that had fundamental problems with their hundreds of developers not being able to get any product shipped. I had a good discussion with their CEO and CTO about overhauling their entire process, CI, deployment, and management structure, and then when I went in for the final round of interviews for this non-programming leadership role the interviews were done almost entirely by junior developers who asked me beginner JavaScript questions. It just boggles my mind.

Look, I get it. It takes time and effort to interview someone, and most of you just want to get back to building stuff. Coming up with a standard question lets you get away with doing more with less effort, and gives you a modicum of an ability for comparison across different candidates.

But really take a long look at whether this selects the right candidates. The skill set needed for most early startups — particularly of early employees — is a glorious, twisted mess of product, code, marketing, design, communication, and empathy. Don’t filter out those people by doing what a Microsoft or an Apple does. They’re big companies, and let me be the first to tell you: that ain’t you right now. You have different priorities.

It’s more work, but it makes for better companies and better hires, in my opinion. But what do I know; I failed those fucking tests anyway.