Brain Teasers Questions For Interview

Why We Keep Asking Riddles To People Who Want Jobs

I started noticing it around 2016. Someone at a former company would throw out the Google "how many golf balls fit in a bus" question and look pleased with themselves when the candidate did actual volume math on a whiteboard. It felt like some kind of proof of intelligence. Then I watched three perfectly good engineers freeze up because they'd never been told to estimate the surface area of a curved object under time pressure, and I started wondering what we were actually measuring. The answer, as it turns out, is complicated.

Brain teasers in interview settings fall into a few buckets that everyone recognizes, but the distinctions matter more than people admit. There's the Fermi estimation problem where you're supposed to calculate something like "how many piano tuners are in Chicago." There's the logic puzzle with doors and switches and light bulbs. There's the lateral thinking riddle where the answer requires realizing you've been asking the wrong question the whole time. And then there's the brainteaser that's really just a coding problem wearing a costume, like "write a function that determines if a bracket sequence is balanced" disguised as "a stack of plates needs to stay upright." Here's what I learned after conducting probably two hundred technical interviews across three different companies. The candidates who perform best on these questions aren't the smartest people in the room. They're the ones who've heard them before, or who have a specific cultural fluency with the kind of puzzle-society that produces them. A candidate from a rural background in Henan might not recognize the joke structure of a riddle involving a monk crossing a river with a wolf, a goat, and a cabbage. That doesn't mean they can't design a database schema. It means their problem-solving training didn't include this particular genre of entertainment.

Brain Teasers Questions For Interview

The practical framework for using them effectively is deceptively simple. Before you ask any brainteaser, decide what signal you're actually trying to extract. If it's quantitative reasoning, pick a Fermi problem and let the candidate talk through their assumptions. If it's structured thinking, use a logic puzzle and observe how they break it down. If it's creativity under constraints, go with a lateral thinking riddle and see whether they ask clarifying questions or start guessing. The moment you can't articulate which cognitive skill you're testing, you're not administering an assessment. You're playing a game, and the candidate is the props. I ran into a specific edge case last year that I still think about. We were interviewing a senior backend engineer who had twelve years of experience shipping payment systems at scale. I asked the classic "two ropes, each burns in exactly one hour but not uniformly, measure forty-five minutes" problem. She stared at me for maybe six seconds, then said "I don't know how lighting two ropes at different ends helps me understand distributed transaction handling." She was right. The problem tests nothing about the actual work she'd be doing, and her honest rejection of the false premise was more informative than any correct solution would have been. I hired her anyway. The brainteaser didn't help me decide, and that's the point I keep trying to make to hiring managers. Here's a counter-intuitive finding from the research, because there actually is some of it now. A meta-analysis published in the International Journal of Selection and Assessment found that brainteaser performance correlates weakly with job performance for technical roles, around 0.15 to 0.20 correlation coefficients. For sales and client-facing roles, the correlation drops further. The only category where brainteasers show meaningful predictive validity is roles that explicitly require the kind of abstract pattern recognition these puzzles measure, which is to say, not very many roles at all. Yet companies keep using them. The most likely explanation is not that hiring managers don't read research. It's that brainteasers serve a social function that has nothing to do with prediction. They signal in-group membership. They create a shared experience between interviewer and candidate. They give the interviewer something to do with their hands while waiting for the real assessment to begin.

Another thing nobody tells you about these questions is the noise problem. Two candidates with identical problem-solving ability can score very differently depending on whether the question uses a context they find boring versus one they find interesting. Ask someone to estimate the number of tennis balls in a Boeing 747 and they might zone out. Ask them to estimate how many Airbnb hosts are in their city and suddenly they're doing demographic math with genuine engagement. The variance introduced by interest-level matching can easily exceed the variance introduced by actual ability, which means your ranking is mostly measuring whether you picked a topic the candidate happens to care about. Now let me give you some actual questions that work, with the kind of specificity that makes them useful rather than decorative. The Fermi estimation category: "How many windows are there in New York City?" The key is not the answer, which is impossible to verify, but the chain of reasoning. A good candidate will say something like "I'll start with population, estimate households per capita, then windows per household, then adjust for commercial buildings." If they jump straight to a number without showing the decomposition, they're either guessing or they've memorized an answer. Both are detectable if you listen carefully. For logic puzzles, the jar-and-coins problem is a solid choice. You have three jars labeled incorrectly, one with only pennies, one with only nickels, and one with a mix. You can draw one coin from one jar. Which jar do you choose and how do you relabel all three? The solution requires realizing that since all labels are wrong, drawing from the jar labeled "mixed" guarantees you get a pure coin type, which then cascades into the full solution in three steps. This question tests systematic deduction under constraint, which is closer to actual debugging work than most people admit.

The lateral thinking category needs more care. The classic "a man pushes his car to a hotel and loses his fortune" riddle has the answer "he's playing Monopoly." These work only if the candidate enjoys wordplay at all. If you ask this to someone who takes language literally, you're not testing creativity. You're testing whether they share your sense of humor, which is a completely different skill with completely different predictive validity for most jobs. Here's what I recommend instead of brainteasers, because I've spent enough years watching good candidates get filtered out by bad questions. Use a work sample. Give the candidate a real problem from your actual product, stripped of proprietary details, and ask them to think through the solution out loud. A backend role gets a system design prompt about rate limiting. A frontend role gets a component architecture challenge. A data role gets a messy CSV with unclear columns and an ambiguous business question. This takes longer to prepare, roughly twenty to thirty minutes per unique prompt, but the information yield per minute is orders of magnitude higher than any riddle about burning ropes or moving monks. For companies that insist on keeping brainteasers in their process, here's the mitigation I've found useful. Ask the question but explicitly tell the candidate that you're not grading on getting the right answer. Say something like "I'm interested in how you approach this, not whether you solve it." Then listen to the approach. A candidate who says "I'd start by clarifying what I'm optimizing for" while facing an ambiguous riddle is demonstrating the same skill a candidate who says the same thing while facing a real work problem. The context changes. The cognitive move is identical.

The biggest blind spot with brainteasers is the feedback loop problem. When a candidate solves your riddle, you feel satisfied. When they fail, you feel confirmed in your suspicion that they're not sharp enough. But you almost never get external validation that your judgment was correct, because you rarely track whether the people you rejected based on brainteasers would have outperformed the people you hired who solved them. The only way to break this loop is to run a controlled comparison, which most companies don't do because it requires institutional honesty about decisions made months or years ago. I'll leave you with one more specific observation from my own process. I once gave the "how many gas stations in the US" Fermi problem to a candidate who worked in logistics. She didn't know the population of the US to two significant figures, which is fine, but she did know something I found more valuable: she asked whether suburban gas stations and highway service plazas should be counted separately. That question revealed more about her understanding of service area decomposition than any numerical estimate would have. I noted it in the debrief and we hired her. The brainteaser didn't help us decide, but her response to the brainteaser's ambiguity did, and that distinction is everything.