Hiring Engineers, another Process

Monday 25 February 2013 at 16:21 GMT

This post was prompted by Eran Hammer's excellent article on hiring what he describes as "non-conforming great talent". It's part rebuttal, part agreement. It's not about how Palantir or TIM Group hire—they both diverge from this on a number of things, but simply my thoughts on the interview process.

"If you don't have time to read this, we are not a good fit."

I think this is fair, but I also think it's pretty specific to Eran. You could interpret it one of two ways: he could be talking about reading articles on hiring, or this article specifically. The former, I think, is valid: people in a position to interview and hire should be open to new ideas, and shouldn't be ignoring posts simply because they're not cited in a textbook or written by someone more traditional. The latter reflects Eran's reality, but not the realities of most of us: he's a notable individual in a position to dictate the means of negotiation, whether he's the one in the hot seat or the one grilling.

"I got up and said, "I'm going home now"."

Eran goes into quite a bit of detail on "What Not to Do", but it can be summed up fairly simply: know what you're looking for and don't compromise. We in the tech industry are in the enviable position of having more jobs than people able to fill them by an order of magnitude, and you should exploit this: if you don't like the way a job is explained to you, feel the questions you're asked demonstrate a mismatch between you and the role, or even get a bad feeling about the organisation or your interviewers, walk out. There are plenty of other jobs. You'll have learned something about yourself, and you should use this information to find something that's a better fit to your expertise and goals.

Jason Wright pointed out that this should go two ways: the interviewer should remind the interview candidate that the candidate is free to leave if he or she feels this is going to be a waste of time.

In short, have some standards. Your interviewers do.

"I don't phone screen. I don't read resumes."

Instead of resumes (or CVs), Eran favours your GitHub account. I love this attitude: I do too. I often tell people that the first four things I want to see on a CV are links to their GitHub account, Stack Overflow profile, blog and Twitter account. (And it's OK if some of those are missing.) The problem is that there are plenty of excellent software developers who do other things in their spare time. They don't talk about code outside of working hours. They arrive at the office at 9am, and leave at 5pm sharp. And in the thirty-five hours they spend working every week, they produce a ton of value. I think that passing on people because they don't have a public record or "an attitude" shows as much of a bias toward a certain type of person as the traditional methods of deciding whether someone is a viable hire. I'd never cold call someone to attempt to get a job, and would have a hard time taking anyone seriously if they called me out of the blue.

I've seen excellent engineers appear from all corners, with extremely diverse personalities and backgrounds. To be honest, I don't think I'd ever say no to a first interview. Of course, with that comes volume. And that leads me to my next point.

"I suck at writing code on demand."

That's OK, because the first round is simply a technical test: write some code for me to solve a very simple problem, then send it back. If it's good, I'll ask you to come in, and we will talk about your code in the interview. Eran explains this way better than I can (if you haven't read his article yet, I don't know why you're reading this).

After we've talked about that, we will write more code.

Sure, no one is great at walking into a room and coding on a whiteboard, but I'm not asking you to build a system to operate a space shuttle. I'm asking you to solve a simple problem. I don't really care if you finish it, and I definitely don't expect you to produce code worthy of shipping. If no code gets written but we have a great conversation, that's brilliant. In fact, the code is simply a driver for conversation. I'm going to be interrogating you about everything you do, and you should be confident enough in your decisions to defend them.

More than that, though, you should be good enough to solve this simple problem in no time at all. You should be better: you should be challenging my original proposal, explaining how tweaking the parameters would make the problem go away altogether. You should be discussing every area of the problem space with me before delivering a one-liner that does exactly what's needed (and probably not what I originally wanted). The coding should be a tiny part of the interview compared to the discussion before and after writing it. You're writing all the code, but the interview is far more reminiscent of a pair programming exercise.

One thing that's really important to me is your ability to react to changes. It's something that's probably needed much more in client-facing jobs such as mine than in Eran's. If I change the requirements afterwards, you should be able to cope gracefully. This isn't something that can be tested well in any situation aside from a real-world scenario, but I can simulate it pretty well in the interview room. What's more, it requires a product to be produced there and then, after the discussion. I'm a fan of agile software development: you should be delivering quickly, showing me the end result, and reacting to any and all feedback. And being able to take a step back, realise that a previous decision doesn't make sense given new information, and fix it, rather than letting the problem fester.

"I'm not going to make shit up."

Eran has definitely convinced me about a few things. The first: whiteboards suck. From now on, I will ask someone to write their code on a real computer, preinstalled with editors, IDEs and support for every language imaginable. I'll even set up some VMs specifically for this, so they can use their favourite operating system. It's not as good as asking them to write it on their own computer with their own .vimrc, but hopefully it's close enough.

Secondly: real problems are the best kind of problems. Most of our development work is not about writing new code: it's about changing old code. Usually badly written, poorly maintained, untested code. So I'm factoring in with an idea some of the members of the London Java Community have been discussing: interviews focused on critiquing and refactoring are far more informative than those focusing on solving a specific problem. I'm going to grab some real code, anonymise it, translate it into five or six different languages, and get the candidate to work on cleaning it up, adding new features and changing old ones. I would prefer to give them the original problem as the offline test and have them continue working on it in the interview, but that's a much bigger change to our interview process.

Thirdly: longer interviews are better. I've heard this one from a few people, including Sandro Mancuso and a report of Reevoo's interviewing practices. Eran asks the candidate to write code beforehand, which is a great way of extending the interview time without a huge cost, but I've seen some information from Barry Cranford of RecWorks that this rapidly degrades when everyone is doing it—no one puts much effort into solving the problem. Sandro just takes a whole afternoon to interview someone. Reevoo actually spend a few hours pair-programming with them on a real feature or bug fix. I can't pull this off at my current job right now, but I think I'll be pushing for something like this in the future.

Do's and don'ts.

Just go read these. They're so important. My favourite, which I'm going to adopt immediately:

Begin the interview describing your role in the team and allow the candidate to ask questions.

This is so obvious in retrospect I feel like an idiot for not doing it.

"You references and past work is what we should really rely on for verifying your claims."

I agree with this statement completely, but I don't think we're there yet. Most programmers don't build portfolios like designers or artists do. That should change, because more and more organisations are starting to think like Eran, and if you don't have some work on display, you're going to be left behind.

"Inspire others to break away"

Eran's definitely inspired me to think more about recruiting. I wrote this post partially as a way to try and get some concrete thoughts out. I'd like everyone else to do the same. The status quo is awful, and the only way it's going to improve at all is if people tell others how they're making it better.