Give Me a Reason Not to Hire You
I don't interview people to find out whether they are good. That is a strange thing to say about hiring. Usually, when someone applies for a job, you talk to them for a couple of hours, ask some questions, perhaps give them a practical exercise and eventually decide whether they are good and you want to hire them. Most interviewers I have seen start with no expectation and wait for the candidate to impress them. They have to climb a metaphorical ladder. They demonstrate enough competence to get to the next rung, answer another question, demonstrate another skill and eventually, if they have climbed high enough, the interviewer decides they are good (enough), even if they wrap the decision in competency matrices, structured scoring and carefully calibrated interview questions.
I usually enter an interview with the intention to hire the person. I genuinely hope I can hire them. Then I look for reasons why I shouldn't. That doesn't make the interview particularly easy. If anything, it makes it more adversarial, just not in a very obvious way. I am not asking the candidate to prove that they deserve the job. I expect that they are good and I am trying to find things that make my initial hypothesis less convincing. Something doesn't fit their CV? I ask about it. Something they say doesn't make sense given something they said ten minutes earlier? I will come back to it. They claim they did something interesting? I want to know why. "Because that is how we did things" is technically an answer, but unless there is something behind it, it is usually a fairly terrible one.
I think the hiring job starts long before I read the first CV. I regularly see organizations hire for positions they don't really understand. There is a copy pasted job description, a cookie cutter list of responsibilities, some technologies and perhaps an old description of what the previous person did. Then they start looking for someone who matches that. I find that backwards. Before I can decide whether someone is a good candidate, I need to understand what I am actually hiring them into. What is missing from the team? What needs to change? Is the person supposed to build a capability, stabilize something, challenge an existing way of working, scale something that already works, or simply provide capacity?
I think this is where hiring becomes more interesting than simply finding someone who can do a job. A candidate does not exist in isolation from the system they are joining. I need to find someone who is a good fit for the specific problem I am trying to solve, who has enough potential to grow into the role and who is unlikely to be destructive to the system they are joining. The same person can be an excellent intervention in one team and a terrible one in another. I have occasionally recommended people for different positions after rejecting them myself, because I wasn't deciding that they were a bad candidate, just that they weren't a good answer to my particular problem.
That also changes what I hire for. I don’t dwell on hard skills, I hire for potential. Not potential in the corporate sense of "ambitious, good communicator, ready for the next challenge", but the machinery required for someone to keep getting better. I don't actually have a precise definition of potential. I have accumulated enough experience with people that I usually recognize it without consciously breaking it into components. However, there are certain things that appear over and over again, even if the list is not exhaustive by any means. I can see potential in someone who is curious, capable, creative, adaptable, who shows ownership. They have the ability to learn something new and change their mind when the new information actually warrants it and to not change it when it doesn't.
There is an obvious danger in this approach. "Good fit" and “potential” can very easily become a polite way of saying "someone I personally like working with." I am aware of that bias, and I don't think hiring people who think like me is the answer. I am not looking for a yes man and a bootlicker. Disagreement is useful. Different experiences and different ways of approaching problems make teams stronger, as long as the disagreement remains productive. I do care whether I would want to work with someone, but this is a pretty weak signal on its own. Someone can be charismatic, agreeable and extremely pleasant to work with and still be a terrible hire. This is a mascot problem: hiring the person who feels good instead of the person who can actually change the system for the better. I try to separate the two by looking for capability underneath the personality. I want someone who can challenge me without being an asshole, disagree without turning disagreement into friction for its own sake, and work differently from me without making collaboration impossible.
I don't care too much about whether someone has already solved the problems I have. I am much more interested to learn whether they can figure out problems neither of us has seen yet. Creativity often beats knowledge when I consider candidates. This doesn't mean I want to hire philosophers who can't execute. Execution is part of potential. There is little value in someone who can construct an elegant model of every problem in the universe but cannot make anything happen.
Developing potential also has a cost. If I have six months to develop someone, I can hire differently than if I need them to be productive next month. If the team has capacity to mentor them, I can tolerate a larger gap than if everyone is already overloaded. I have been in situations where we simply couldn't invest enough into somebody's growth, but finding someone who was already exactly right would have taken an unreasonable amount of time. Neither of those options were sensible, I had to stop the hiring effort and try to fix the system in a different way.
During the interview, I am ultimately not really evaluating answers, I am evaluating epistemic behaviour. I don't particularly care if their answer is the one I would have given. I don't actually think there is such a thing as an "initially wrong answer" in the way interviews usually treat them. I am perfectly happy to accept an answer if the candidate can defend it when challenged and tell me what information would change their mind. The important part is the reasoning and the intellectual transparency. Eventually I will introduce something they hadn't considered and see what happens. Model revision is one of the strongest signals I know. Someone who can say "I hadn't considered that, and it changes my conclusion" is showing maturity. Someone who can say "I hadn't considered that, but I don't think it changes my conclusion, and here's why" might be showing even more. I am always careful about people blindly accepting the new information, as that isn't learning, but compliance.
I deliberately try to find the edge of someone's capability. I tell candidates this beforehand because reaching the edge is not a failure. Eventually I want to ask something they cannot answer, or cannot defend. If I don't, I am mostly measuring the part of their capability they already know how to display. The really interesting information starts when they run out of prepared answers. Do they start making things up? Do they become defensive? Do they ask for more information? Do they reason from first principles? Do they acknowledge uncertainty? Do they challenge the premise? Do they construct a way to find the answer? Do they update their model, and if they do, do they actually incorporate the new information later?
This is why some of my questions sound almost insultingly basic.
"What is quality?"
"Imagine you are in my shoes. How would you know you are working with a high-performing team?"
"What kind of KPIs would be interesting to you, and why?"
I am not particularly interested in the definition of quality they memorized or what KPIs they have seen in their previous company. I want to see what happens when I give them something that can be interpreted in several different ways to expose different models of approaching the problem. Then I can start asking why they have given this specific answer.
This ability to change abstraction levels is probably the most important interviewing skill I have developed. A candidate can tell me what they did, and I can ask why. They can explain why, and I can ask what problem that decision was solving. They can describe the problem, and I can ask what would make their understanding of it wrong. I can keep moving between the concrete thing they did and the model they were using to decide what to do. Sometimes I go the other direction and force the abstract answer back into something specific. I don't necessarily need to know more about the domain than they do. I need to be able to reason about the structure of their reasoning. That is also why genuinely excellent candidates occasionally expose my own limits. If they can follow me through the abstraction layers and then start doing the same thing back to me, I eventually run out of places to push. I consider that to be a very good outcome.
There is a catch to abstraction, of course. Someone can be exceptionally good at it and still be useless. You can construct beautiful theories while being completely incapable of executing anything. It is also why the second round is different. I usually give candidates an exploratory testing exercise. It is not particularly domain specific. I want to see what happens when they are given something unfamiliar and have to figure out what matters. Do they ask questions before starting? Do they create structure? Do they take notes? What do they notice? How do they decide what is worth investigating? How do they communicate what they found? What would they automate? Why? Should it be automated at all? I don't need to ask them whether they are systematic. If the situation calls for being systematic, I can watch them either do it or not do it.
I am not against a structured approach to hiring. Quite the opposite. I have a fairly good idea of what I want to learn about every candidate and I have a small collection of questions that tend to produce a lot of useful information. There are questions I ask almost everyone because they tend to expose useful things. There are questions I build from the candidate's CV. There are areas I want to understand in every candidate: adaptability, systematic thinking, whether they ask questions before acting, whether they understand business value, whether they can learn, whether they take ownership. But my aim is to standardize the signals I am looking for, not the conversation itself. Two candidates can arrive at the same useful evidence through completely different conversations, and forcing them through the same sequence simply because it makes the interview easier to compare doesn't strike me as an improvement. Experienced interviewers should not be prevented from following something interesting when it appears.
I am also not particularly fond of structured scoring. Candidates have bad days. Interviewers have bad days. Someone can be nervous, someone else happens to be an excellent public speaker. I have almost never had five fantastic candidates competing for one position anyway. Usually I'm happy if I get one or two people I would genuinely like to hire, and one of them may refuse the offer anyway. I don't need to know whether candidate A is better than candidate B. I need to know whether candidate A is a good answer to the problem I have. Comparing them against each other solves a non-existent problem.
Pattern recognition is a very useful tool here. It is also the thing I have to be most suspicious of. After enough years of interviewing, managing and coaching people, a lot of the small patterns disappear from conscious thought and they become hunch. Something doesn't fit. The answer is too polished. Two parts of the story don't quite line up. Someone asks a question that tells me much more than the answer they just gave. I follow the thread without necessarily being able to explain why that particular thread looked interesting. That is where hubris becomes dangerous. I have no way of knowing from inside my own head whether every instinct is an accumulated heuristic or simply a bias I have become very comfortable with. The only reasonable response I have found is to make sure other people can challenge me. I allow the candidate to challenge the things I say. The second interviewer can challenge my opinions. A game we play can challenge my biases. Somebody I trust can tell me that I am wrong. To make sure we are choosing the right person, I want other people involved and I want them to be able to push back. My coworkers have successfully pushed me into moving the candidate to the second round before. I am quite happy when that happens, as the best protection against becoming convinced that your intuition is infallible is having someone who is willing to tell you to stop being an idiot.
I don't think you can scale my approach by giving somebody my interview questions. I have trained people to interview with me, and some of them became very good at it. They didn't become good because they memorized my questions. They sat in interviews with me. Sometimes I was driving and they watched. Sometimes they took the lead while I stayed in the background and only intervened when necessary. Over time they learned what I was looking at and, more importantly, why I was following one answer rather than another. Eventually they developed their own heuristics. That is good enough for me. I don't need clones. The transferable part isn't "ask what quality means." Anyone can ask that question. The transferable part is noticing what the answer tells you, what it doesn't tell you, which competing explanations are still alive and what question would distinguish between them.
The method is teachable, but it takes time and effort, it requires apprenticeship, not lectures. You need to learn to operate at least one layer of abstraction above the person you are interviewing. Once you can do that, a lot of the other things start happening naturally. You notice contradictions because you're comparing models rather than answers. You ask follow-up questions because the interesting part of an answer is usually somewhere in what remains unexplained. You change abstraction levels because you can see when the current level isn't producing useful information. You push towards the edge because you know that the candidate's most polished answers are usually the least informative ones.
There is a temptation to solve this problem organizationally by taking judgement out of the hands of people and replacing it with process. Use a scorecard, couple mandatory questions, and require rigid independent scoring. Collect enough documentation that nobody can later question why someone was hired. I understand why organizations do this. It makes decisions easier to audit and easier to defend. I don’t think it makes them better. There is an uncomfortable trade-off to be had here. If you want excellent people to exercise judgement, you have to trust them and accept that judgement is going to be imperfect every now and then. You can easily govern mediocre judgement into consistency, but with the same process you will also govern exceptional judgement into mediocrity. I admit I will occasionally get things wrong too. There are a couple peopleI interviewed who I would not re-hire. None of them were bad, they were just not excellent. However, I would extend the offer again to most of the people I hired in a heartbeat.
I don't see probation as a separate thing from hiring. It is simply a much more expensive and much more realistic diagnostic environment. By the time somebody joins the team, I should have a hypothesis about them. There are things I can learn in two hours and things I simply cannot. I can observe how someone reasons about unfamiliar problems. I can see whether they update their model when I introduce new information. I can get a reasonable idea of whether their weaknesses look coachable. I cannot know exactly how they will behave six months into a difficult project, how they will react when a colleague repeatedly frustrates them, or whether their carefully constructed interview answers survive contact with reality. Then they get to work and reality starts producing evidence. Sometimes the hypothesis was good. Sometimes it wasn't. Sometimes the person turns out to be even better than I expected. The important thing is not pretending that the two-hour interview gave me certainty. The decision was made with incomplete information. I am comfortable with that. Most of the interesting decisions I make have the same property. The big difference is that this time the person being diagnosed talks back.
