Maximizing resume SNR

This is a brief note about resumes and hiring. At Terraform Industries, I’ve been doing a lot of recruiting recently and it’s helped me crystallize a few ideas I’ve had in this area. This blog will change over time.

Terraform Industries is hiring! Join us as we decarbonize our industrial economy by getting carbon from the air more cheaply than from the ground.

First, the disclaimer. While I’ve been on both sides of the table with recruiting, I’m not an expert and I’ve only worked in a few narrow fields. My father happens to be a professional recruiter, though, and we sometimes talk shop. I’ve written a book about academia which has a chapter on adapting academic skills to industry – a path I’ve personally trod.

Before we jump into resumes, some generic advice for job hunters who have stumbled upon this post. The best way to get hired is always through your network. This usually means former colleagues refer you to their colleagues who know about opportunities. Think about personal branding, and if you’re not getting >10 rejections a week, you haven’t tried hard enough to fail yet.

Resumes suck. As instruments to help workers find jobs or recruiters find workers, they contain zero useful information. Their signal-to-noise ratio (SNR) is zero. I’m focusing here on hardware engineering but my friend Aline crunched the numbers in software and found the same thing. Literally the only thing in a resume that was slightly predictive of on-the-job success was grammar and spelling, which in turn means that the candidate can either proof read or knows someone who can.

In the hiring process, a job req generates a flurry of applicants. As with any open selection process, inbounds are dominated by poor candidates – people who have been on the market for a while, people who spam job postings, people who can’t do the job. Not bad people, just not right for the job. The recruiter has to reject most of the candidates and pass a few on to the next stage, which is usually a screening call. The candidate.RejectBool() function takes as its argument the only available information – the resume.

The resume is stuffed with data. A mostly complete list of jobs, skills, certifications, degrees. The candidates name and contact information. Intertextual information about writing style and literacy. The RejectBool() function has to process this data in a few seconds and move most to the trash.

As a result, we see selection focus on red flags or disqualifiers, or particular conditions that are just relatively rare on the job market, such as affiliation with a prestigious school or company. But what can we say about someone based on their resume? What information on there is unique to the candidate? Literally everyone’s resume is a chronological list of similar-sounding positions at similar-sounding companies. If you calculated the principal components of a set of 10,000 resumes, you could span the whole set with about 3 vectors. The data is highly compressible, which means the information content is low.

At Terraform Industries we’re hiring for a range of skills, but the ability to obtain brief proofreading services from a relative isn’t near the top of our list. I know plenty of brilliant engineers and technicians who will probably never write the Great American Novel. Some are brilliant at spoken communication, some not so much, but neither affects either the resume or the job very much. This is why our recruiting hook asks only for a page of evidence of execution capacity.

Since we’re not selecting by typos, what other information in a resume is also irrelevant for us?

We don’t care if your name sounds foreign or not male. But the name is literally at the top of the resume so it’s not surprising to find that study after study shows more resumes are rejected for, eg, Hispanic names than any other reason. Similarly, resumes with names and affiliations removed show completely different selection patterns.

We also don’t care that much about which school you went to. It’s true that some schools have excellent education programs in certain skills, but as far as I know, there isn’t (yet) THE school for carbon capture. We don’t care what GPA you got. We’re happy you’re proud of your 4.0 and plenty of us got good grades too, but not all. Our work environment is quite different from memorizing facts for an exam, so grades aren’t predictive of much.

We don’t care that much about what your job title was, which city you worked in, or which company you worked for. It is true that some companies have a reputation for a level of rigor that is a cut or two above the average, and we’re happy to see that in a resume. But we also know plenty of people who are just as smart who didn’t work there, or smart people who did work there who are still a bad fit for our company. I can usually tell within a couple of minutes of talking to someone if they’re mechanically intuitive and have valuable insights regarding our business, but I won’t find that in a resume. I would be lucky to find it in a cover letter.

I don’t care that much if you were out of work for a while. It’s one of the only comprehensible things about a resume’s list of jobs so recruiters will often reject candidates who were out of the workforce, but that’s silly. Even if you make it to screening, why would I want to waste a question on “What were you doing in 2016?” as though they were robbing banks or getting a high score in Halo? Is that relevant to solving my company’s problems? Not unless my company’s problem is avoiding formal employment (it’s not). In reality, plenty of great people take breaks from the work force. Raising children is labor intensive and yet if we don’t make more humans, we will run out of humans, and that will be Bad. I’ve failed plenty of technical interviews myself, though fortunately was not out of work at the time. Again, this is not relevant unless the candidate’s job will be to successfully interview for jobs.

No-one cares much about your hobbies. The last couple of lines of a resume are almost as important as the first couple, and should never be wasted on something that isn’t directly relevant to the job req and the problem the hiring manager is trying to solve. The resume should be designed to avoid common triggers for rejection, and unless you’re sure the recruiter and the hiring manager freaking love your particular hobby, why waste a line of your resume on it? Ideally, the last line should be the strongest evidence you have that you can be trusted to solve the particular problem articulated in the job req, to the hiring manager, whose identity you should have deduced by now.

A resume should be a single page. The candidate has about 350 words to make their case. Every word should be adding as much positive signal as possible to the mix. If in doubt, leave it out. A curriculum vitae (CV) can list every last position. At last count, mine was 10 or 15 pages long, and even more irrelevant for any job than a resume. No hiring manager in industry needs the abstract of every paper and talk you’ve ever published. For academics going to industry, no more than “five first author papers in field X” is required. Not only will the recruiter not care nor look up your paywalled journal article, you will be lucky if a single colleague asks you about what your academic research was about more than once per year.

Unfortunately for us and other companies in the hardware development space, resumes are also lousy ways to talk about execution capacity. I could look at two identical candidates with similar career paths, and have no way of knowing that one of them built a jet engine in their living room. Their resumes are very similar and have no ability to choose between them. But a single photo of them in front of their own jet engine would tell me 95% of what I need to know to make a job offer. It would pretty much instantly put them on the top of the screening pile too – and that’s an important point. Most companies hiring people outside their immediate network will hire in stages – resume, screening, deep call, on site, reference checks. As the process continues, more resources are spent and detail is uncovered. At any given stage, you only have to get the recruiter excited about moving you to the next stage.

Incidentally, this is one of the few ways that certain degrees from certain places can accidentally transmit information. If you got a PhD from Caltech (I did) it shows you have a certain degree of pain tolerance and probably autonomous research capacity. But the opposite isn’t necessarily true – it’s weakly selective at best. I know half a dozen people off the top of my head who didn’t finish high school who I would rather hire than a randomly selected PhD. A PhD program isn’t totally useless though – it’s a great way to meet people who go into a range of industries. So as usual, the school is mostly useful for networking rather than knowledge, at least beyond a certain point.

There are ways to honestly signal a degree of technical competence that can help you find career opportunities. I credit Patrick McKenzie and Vladimir Dinets for prompting my own evolutions in blogging. Through this and Twitter I build a personal brand of technical literacy and a culture of asking the right questions.

It’s also possible to project a semblance of technical ability through video, and I greatly admire work by Stuff Made Here, Xyla Foxlin, Simone Giertz, Mark Rober, Joe Barnard, Cody’sLab, and friends for its depiction of the challenges of making something in the real world. But not everyone can make videos, or wants to, and plenty of people can project a certain personality and level of confidence in media that may not be matched in reality.

Personal projects are a great way to show that you can actually manage a whole system. They’re a great jumping off point for getting into specifics, particularly if the project had failures along the way. Not everyone has the time or budget to do “free work” outside of their job, but what we’re looking for here is the highest leverage way to improve your job application’s signal to noise ratio. If you have to change your car’s oil anyway, you may as well photograph it!

So what DO I care about?

I want to see evidence that your former colleagues trusted you to not break the product. My favorite interview question is “What is the most expensive thing you’ve ever broken?” because it immediately tells me more about your job and its actual responsibilities than any summary could. It also begins a conversation about accountability and introspection.

I want to see evidence that you were matched with tasks within your ability level and that you succeeded at them. There’s not much point in crushing a series of trivial tasks – Terraform Industries is solving a hard problem. There’s also not much point in taking on an impossible task and failing, at least, not doing it over and over again. Once in a while is good for calibration and character building, but if your career is an unbroken string of ambitious failures, we don’t want to add to it.

Specifically, at Terraform I want to see evidence of insight, intuition, autonomy, integrity, low drama, execution capacity, and urgency. A good one pager will tell a brief story that indexes strongly on most of these without having to break it down. The best candidates mix them together in a compelling way.

Insight is important – and we select for this in investors too. If you can tell us something important about our business that we didn’t already know, that’s really really good signal. It shows you think deeply about these issues, you have expertise, and you’re able to translate between reality and some kind of cognitive knowledge structure.

Intuition, particularly mechanical intuition. Machines are friends. When something isn’t working properly you usually know what’s wrong before you break out the trouble shooter. Again, you have the ability to apply your cognitive mental models of the world to the actual real world in front of you, and integrate new information into this framework. For onsite interviews, I’ve seen this done with a bucket of random parts. “What’s this for?” “How does this go together?” “What do you think assembly yield for this part is, and how would you improve it?”

Autonomy is vital in start ups. You have to be able to see the big picture and how you fit into it, understand what’s missing, and go fix it. Essentially everyone is their own manager and the team isn’t large enough to specialize yet. If this works well it’s incredibly powerful but it doesn’t work if you’re unable to do anything without a supervisor spelling out your tasks in incredible detail. Many armed forces veterans get this intuitively – they understand command intent, the boundaries of their scope, their tools, and then OODA until it’s solved.

Integrity seems obvious but actually in the vast majority of jobs and careers, appearing competent is 95% of the job, where as actually being competent is only 5%. Put another way, competent workers who don’t bother to ensure they get credit will have a Bad Time in places that can’t actually measure exactly who is doing what. This is especially common for academic escapees. In a start up environment, it’s usually pretty obvious who can’t actually do the work, because the buck stops quickly. It’s damaging and expensive and time consuming to fix problems caused by people lying to other people and lying to themselves, so if you are unable to separate bullshit and reality, do us all a favor and go do counterproductive work for a company that’s making the climate situation worse.

Low drama. Speaking of expensive and time consuming avoidable issues, be prepared to check your ego at the door. At Terraform we’ll assume you’re new to carbon capture but we have very little tolerance for avoidable interpersonal strife, which my children are currently mastering in daycare. If you are feeling aggressive, you are encouraged to be aggressively helpful, or to aggressively rewrite onboarding documentation to help grow the team. We’re actively selecting against people who we think will make it harder to hire great people in the future.

Execution capacity. This is where personal projects come into their own. Lots of us work on enormous problems where we’re just a tiny cog in a huge machine, but being able to see the big picture is very valuable. A resume gives us precisely zero data on whether you’re actually able to execute on much more than sending an email. If you repair your own appliances, build stuff, organize community events, we want to know about it. This is Strong Evidence that you can Get Shit Done. Three lines in your resume about your 4th most recent job only says “I managed to not get fired for a while in 2012” while three lines about your Roomba mod practically guarantees a screening call.

Don’t assume the recruiter knows about the job req. Recruiting is a specific skill set but it doesn’t pay nearly as well as engineering, so most engineering-literate recruiters long since jumped over. When you write about former jobs in a resume, or discuss them in an interview, do the recruiter a favor and assume they know nothing about the company, the job, or the problem. Really spell it out.

Your summary must contain the most specific information you can manage about exactly what you were hired to do and the projects you owned, what was hard about the task, list the Key Performance Indicators (KPIs) and by how much you improved them. Focus on the bottom line improvement. The hiring manager doesn’t care particularly if the problem was academically interesting or what color your shirt was. They need evidence that you can solve problems that matter.

Let me give an example. I’ll give two resume entries for the same job at the same company that I did a few years ago.

15 September 2015 – 5 March 2018. Levitation Engineer, Hyperloop One. Reported to [name redacted]. Focused on success for my team. Managed relationship with BlandInc. Ensured quality and synergy. Won award for most improved in office pull up competition.

Sep 2015 – Mar 2018. Levitation Engineer at Hyperloop One. Responsible for clean sheet analysis and design of magnetic levitation system for full scale development prototype, delivered on time and budget for successful first flight in 2017.

Taking up the same space, the first entry is full of irrelevant, redundant, context-free information, jargon, and at least one lie. (Never lie on a resume!) The second entry obviously leaves out details (I worked on >100 projects) but it establishes a track record of autonomy, success, and technical mastery. The second entry could be even better, but it’s a good start.

It is 2022 but most jobs will still require a resume. Some will even require a boring-looking one, and that’s fine. The resume’s job is to be unoffensive. But think carefully about what you can write in yours that provides actual signal that you can do the job and that distinguishes you from an unending onslaught of similar candidates with similar backgrounds. Definitely don’t save the file as resume.pdf! The filename should contain your name and the company you’re sending it to.

For Terraform, send us your one pagers. Ideally they will contain photos of awesome hardware you personally created, together with a brief and informative summary of how the project relates to your desired role with us. We don’t want you to waste hours of your valuable time honing a one-pager to perfection. Think of it instead like an asteroid impact – imprecise but high energy, quick and dirty. The single page limitation is to ensure focus. Omit anything that degrades the signal. Omit anything that’s redundant. If you can tweet it at us, it’s about the right length!

As of March 2022, we’re particularly interested in finding an experienced chemical engineer to lead our Sabatier synthesis reactor, and an electrochemical engineer to lead the electrolyser team.

With candidates’ permission in future, I will include a few exceptional examples below.