How I Hire
Picture this: A hiring process in which candidates thank you for the refreshing experience, and you achieve a 90% reduction in interviews while snagging the perfect hire nearly every time.
Sounds too good to be true? This is the reality of a hiring process born in a crisis.
I was running a small, early-stage company, and a few weeks before a critical deadline, most of the team left on the same day. It was a disaster. I had to hire new people immediately, but I also had to do sales. I could spend my time interviewing candidates or selling, but not both.
Out of necessity, I devised a new hiring approach that revolutionized how we recruited. This post will take you through that process—from establishing first principles to crafting a job ad that excites candidates, designing a form that filters out unsuitable candidates, to interviewing techniques that identify the top performers.
This hiring process is built on a set of principles. Internalizing and applying these principles will help you shape your strategies and tactics. If you take one thing away from this post, it should be the principles, not the tactics.
The employer and the candidate should invest a comparable amount of effort and take risks. As long as both parties continue to make investments in the process, it aligns their interests and sets them up for a balanced relationship.
There is a cost to evaluating every application, so you do not want to maximize the number of candidates. Instead, think about how you can have as many people drop out of the process while still having the best candidates apply.
Hell yes or no
If you're uncertain, the answer is no. Only extend an offer when a candidate gives you that undeniable "hell yes" feeling. The cost of hiring the wrong person far outweighs running the process longer.
Hire for trajectory
Focus on where a candidate is going, not just where they’ve been. Assuming the candidate will stay with your team for a while, it matters more how capable they will be in a few months or years than today.
How to write the job ad
Writing the ad should not be an afterthought. You should not delegate it.
Writing is thinking; writing the listing is your opportunity to think through what problem(s) you want this role to solve for your organization. You may not actually need to hire for this role if you struggle to identify the problem.
It will surprise you how often this is the case.
Have you articulated the problem?
Great, now what is challenging about this problem? What makes it hard? Don’t worry about scaring some people off (remember, self-selection is your friend). If you make it sound hard, you will put off some people but excite others. The ones who get excited are the people you want to apply.
Now, what impact is solving these problems going to have? Write it out in detail. Ambitious people want to know that their work makes a difference. Tell them how solving these problems will impact the organization.
And why should they care about the organization in the first place? What do you do, as a company, that matters? You don’t have to have a fancy mission. Just explain why what you do is important, who it impacts, and in what way. Some work to live, others live to work. You may prefer to be friends with the former, but you want to hire the latter. We all want to feel that our lives have meaning; people who find meaning in work care a lot about what they work on.
Finally, articulate as clearly as possible what the job is not about. You do not want to waste people’s time applying for a different job than they had in mind because they will waste your time interviewing for it.
How to design the form
I find CVs almost redundant, and I hardly ever look at them. If I do, it’s only after reading the form. Initially, it was to avoid bias. Over time, I have learned that leaving CVs out at this process stage is more efficient. I can tell a good candidate apart within seconds of reading the answers. The CV blurs the picture, adding information I do not need.
However, this requires having a good application form. Designing a good form is hard. The form questions should reflect the problems the role is designed to solve as closely as possible.
A marker of a good question is if, reading the answer, you immediately know (within 5 seconds) whether it is a good candidate. If you read 2-3 applications and each time you read an answer to this question, you do not have this clarity – toss the question out.
Examples of questions
Below are examples of questions we successfully used recruiting for Product Manager, Data Scientist, and CTO roles.
We were seeking a Product Manager for hive.one, a product based on a ground-breaking community detection algorithm that could find communities on Twitter purely by analyzing the social graph. The Product Manager would spearhead productizing the technology, so we needed someone deeply analytical with a good sense of the product and the ability to understand the underlying tech.
What metric should we use to best assess the level of satisfaction with our lists (e.g., Bitcoiners, Ethereum) by each community?
Please take into account that your idea should be technically feasible. It should also correspond to where the product currently is right now. Make sure to take some time to play around with hive.one before answering this question.
Notice that answering this question requires a holistic understanding of the product. There was no one right answer to this question. We kept struggling internally with formulating the right metric. To answer this question, the candidate had to understand the product, its value proposition, and the target audience and only then start thinking about how their satisfaction could be measured. In other words, by asking about the metric, we quickly evaluate a whole set of abilities, knowledge, and skills necessary for the job. Since the answer starts with a proposed metric, we could quickly spot bad answers – a metric that did not make sense. We saved a lot of time reading justification when a candidate proposed a metric that did not make sense. We disqualified most candidates based on the first few sentences of the first question (within a few seconds).
What interface other than a website like hive.one could we build to make our scores useful to the end users?
This could be anything, but the idea needs to make sense. Please make sure that you justify your answer – why this idea, not something else. What problem does this address for the user, and how would we solve it? Please be as detailed as you can.
This question was designed to show whether the candidate has a product sense and is creative. Again, there was no one right answer. What we were looking for were ideas that made sense to us. A “hell yes!” candidate would propose an idea we either considered or would consider implementing.
If we implemented this idea, what metric would you use to measure the success of this project?
We're not looking for a 2-3 word answer here. The point of this question is to help us understand how you think about measuring success and how you design metrics. Please be specific and give an exhaustive answer.
This question aimed to increase our confidence with “hell yes!” candidates. If a candidate answered the two previous ways well, then answering this question was a cherry on top and made a “yes” easy.
What made you want this job?
What makes you the right fit for this job?
We end every form with these questions. Articulating the reasoning helps us understand the candidate's motivation and how they think about themselves and their career. A nice side effect is that the candidate also sells him herself on the job even more in the process of answering.
What does reducibility/irreducibility mean in the context of directed graphs? Which algorithm allows an approximation of the stationary distribution of a random walk on a reducible directed graph, and how?
Our questions are intentionally hard. We know that you can't fully solve the problems presented – the goal is to see how you approach solving them. We want to see how you think.
Describe your favorite community detection algorithm on graphs. Is it applicable to directed graphs? If not, suggest an extension to make it applicable.
Suggest an efficient algorithm to compute the number and vertex sets of all closed triangles / 3-cycles on an undirected graph on n vertices. How does the runtime scale with n?
What made you want this job, and what makes you the right fit?
How do you imagine that the system we are building would impact the world and the internet?
Play a thought experiment with us: assume that we have successfully built what we intend to build. How is the world different? How is the internet different? What are the implications for the tech stack?
What are the key challenges we are likely to encounter?
What would keep you up at night if you joined us as the CTO? Where do you see us facing the main risks? Please elaborate on them and how they should be mitigated.
What is a high-level technology trend we need to anticipate for this project?
Why do we need to anticipate it, and what implications it may have for our plans?
What made you want this job?
What makes you the right fit for this job?
Bad candidate: you know within the first 5 seconds that it is a “no”; you don’t need to read all the answers or even the whole answer.
Good candidate: You have a “hell yes!” rush of excitement when you get to the last answer.
Each answer should increase your level of excitement.
If the above is false, you have not created a good questionnaire yet. Keep trying.
It sometimes takes 10-15 iterations to get it right. But it’s worth it.
I find the experience of interviewing weak candidates depressing, and I get exhilarated about interviewing great candidates. With a well-designed form, you hardly ever speak to the former.
How to Interview
Move fast through interviews – we aim to complete them within a week. The best candidates scrutinize you as much as you scrutinize them. You make a great first impression if you do not waste their time and set clear expectations. Many companies drag this process on and require an unnecessarily high number of interviews. This is a great way to stand out.
During the first interaction, we state that the candidate should expect around three interviews, but not more than four. We explain that if the conversations go well, we will ask them to do a paid trial project with us that would take about a week, so the whole process will take two weeks if the candidate can make our timeline.
We can move so fast because we have filtered aggressively at the previous stage. With just a few people to talk to, it is possible to fit all the conversations in without distracting the team. This matters because the best candidates often have multiple offers. If you move too slowly you may not get a chance to make yours.
I do the last interview, I ask one question: “Please tell me your story chronologically. I’m interested in the projects you have worked on and anything you have done or achieved that you feel proud of. Focus on the problems and challenges you have faced and how you solved them.”
You will be surprised how many people fail to tell their stories in chronological order coherently. I have experimented with many questions over the years but eventually realized that this is the only canned question I need. I do ask follow-up questions, but they depend on what I hear.
“Why did you choose to work on this?”
“What did you find appealing in this problem?”
“How did you solve it?”
“Why did you solve it this way?”
“I don’t understand. Can you explain this to me?”
“I still don’t understand; I’m not an engineer/data scientist/mathemtician/etc. can you explain this in a way that I can understand?”
“If you were leaving this job/project and I was to replace you, and you only had two minutes to give me advice to set me up for success, what advice would you give me?”
“If you were to work on this project all over again, knowing what you know, what would you do differently and why?”
Many people struggle with “selling” themselves and do not do well in interviews. I am one of them. If you asked me to give examples of situations when I showed ingenuity or extraordinary ability, I would not be able to think of anything. That should not be a disqualifier unless I interview for a sales role.
As the interviewer, you want to have the right mindset. A wrong mindset is that the candidate’s job is to impress you. A good mindset is that you and the candidate work together to find evidence of their extraordinary ability, indicating they are the right person for the job. In this collaboration, you guide them with questions that help uncover this evidence.
I work as hard as the person answering the questions to be impressed. Titles and fancy brands do not impress me; I find deep thinking, ingenuity, and ability impressive. It has not always been the case, and I made some of the worst hiring mistakes, being blindsided by resumes with big names on them.
The solution is to keep listening until the candidate mentions something that grabs your attention, ideally a project you think is ingenious. And just keep asking “how” and “why” questions. If you dig long enough, you will either unearth insights and actions that impress you or learn that there is little substance behind the credentials.
Find your must-haves
I cannot tell you what you should look for in a candidate. You need to figure out what kind of people you work well with and what characteristics you seek. My must-haves are self-motivation and the ability to articulate one’s thinking clearly. I found that side projects are a great indicator of these qualities, so I seek them out in my process.
Building a company or a side project is a good sign alone. A project that solves a problem in a way that resonates with me is an obvious catch. It is a home run if the candidate also found a way to promote it well, and people are using it. If I come across someone who checked all three boxes, I move aggressively to close them before someone else does.
Give a quick no
I have developed strong intuition after interviewing hundreds of people. Occasionally, I will know this is not the right person within a few minutes.
When this happens, the kind thing to do to yourself and the candidate is to tell them immediately that it will not work. If “you are nice,” carry through with the interview despite knowing it’s a no; you will avoid getting called an asshole, but you are an asshole. Saying “no” when you know it’s a “no” is not easy, but it’s right.
I should never speak to a weak candidate in a well-running process if I do the last interview. I do a post-mortem whenever this happens. Understanding why this person got through the previous interviews is critical because this is a sign of the process malfunctioning and an opportunity to find flaws in the system.
Do not let it go to waste.
The trial project
Before making an offer, we always start with a trial week project. We pick the kind of project that the candidate would work on if they got the job. We onboard them on our tools, introduce them to the team, etc. The idea is to simulate what it is like to work together as closely as possible – both for them and us.
We always know within the week if we want to make an offer. Typically, we know within the first 2-3 days. Sometimes, it’s inconvenient to do the trial project, for example, when the candidate is still employed elsewhere. We have broken the rule a few times and made an offer without going through the trial project, and we regretted it every time. It doesn’t need to be a whole week; it can be broken over two weeks and done over weekends. Be flexible, but don’t skip this step. Do not make an offer unless you first worked together.
Finally, there is the question of how much you should pay for the trial. We experimented with different approaches, one of them erring on the generous. We’d pick the upper end of the salary range for the role and pay the week’s pay. This turned out to create an unhealthy dynamic. Some people went through the trial project without giving it all in; it seemed like they were doing it just to get paid for the week. It also sets up people for disappointment. If you set the benchmark at the upper end of the salary range, then in most cases, they will be disappointed when they get the offer.
We learned that a wiser approach is to calculate the payment for the trial project on the lower end of the salary range. We explain this to the candidate by saying that the trial project is about finding out if it's a good fit for both sides. We do not want them to work for free, but the point is that we are taking a risk, and the candidate should be taking some risk, too – this signals that they care about getting the job. That creates a better setup. Candidates approach it with a better attitude and are often positively surprised by the offer.
Be generous with the offer but stingy with the trial project.
All of the above applies to inbound hiring, but I have hired about half of my team on Twitter, Reddit, and ProductHunt. The outbound process is unpredictable, and you cannot scale it, but I have made some of my best hires this way.
During the interview, I look for signs of ability, self-motivation, and thoughtfulness, and my favorite type of evidence is side projects. You can skip right to this step when you spot such projects online. Whenever I see an individual posting a project that grabs my attention, I send them a short message; it could go something like this:
I just came across [x]; I think it’s great; what I like about it is [y]. Have you considered [z]?
People who spend their free time working on passion projects love when others recognize and appreciate their work. Articulating what you like about the project and making suggestions shows that you took a close look and understood what they were going after. This goes beyond a simple compliment and sets you up for a warm interaction.
I get a friendly response and a brief exchange nine out of ten times. When this happens, I ask if they are open to job opportunities, and if they are, I give a brief idea of what I am working on and how it ties to what I saw in their project. From there, I suggest we jump on a brief call.
This call has the reversed order of the interview. I start by pitching the company, the team, and myself, and they ask questions. You mustn't spend the whole call talking about yourself and letting them ask questions. Most of the call should be about the candidate telling you about themselves, so you have to flip after the first few minutes. A good way to set the expectations upfront is to start the call by saying:
“How about I tell you a little about the company and myself, and then I’d love to learn more about you?”
This is a reasonable suggestion, and it formalizes the agenda. When you actively recruit someone, they may feel they should not undergo an interview. Focusing on pitching to them may be tempting, but this is a mistake, even if you are convinced you want to hire them. So far, you have been making all the effort. You must flip the dynamic to balance the relationship.
Remember the balanced investment principle.
The candidate will feel better if they had a chance to ‘win you over,’ even if it was unnecessary because you knew it was a “hell yes” in the first few minutes. You still need to put them through the process, but you can accelerate it. If the first conversation goes well, I say something like:
“I enjoyed the chat, and I’d love to talk to you about the next steps, but first, I want to know how you feel about exploring joining our team. Is this something you’d feel excited about?”
If they don’t tell me, they would feel excited; the hiring conversation ends here. I just invested by telling them that I would like to move forward, so they need to invest, too. If that happens, then I suggest how we move forward.
“Great, I would like you to speak to [x] and [y]; you will have a chance to learn more about our [z]. If these conversations go well, we will invite you to do a trial project with us, which typically takes about a week, and we pay you for it. We will either make you an offer or say it’s not a fit before the project ends; either way – you will know within that week. How does that sound?”
The process reversed; they started with the last interview (with the CEO) but still needed to go through the other steps. You may want to frame this as their opportunity to learn about the company, but you must complete your process. This also communicates to the candidate that you are selective and not anyone can join the team.
No one joins the Navy Seals without going through Hell Week; no one should be able to join your team without going through your process.
If you apply the ideas outlined here, remember that this should be just the starting point. The key idea I am trying to convey is that you should think about the outcomes you want and walk backward to the process.
Ask yourself what kind of people you want to work with. Why would you want them to want to work with you – what should motivate them? Who do you enjoy working with and not? What reputation do you want to develop as an employer in a decade from now?
Meditate on these and similar questions.
Write them down. Talk them over with your co-founder if you have one.
Once you gain clarity on what you want, only then think about strategies and tactics that can get you there. As you implement them, continue to evaluate whether you are achieving the outcomes you were looking for; iterate.
Your principles will come in time. You will notice that you have been making the same mistake repeatedly, and you put a rule in place that makes you stop. When they work, these rules make for great principles.