Pymetrics Assessment 2026: The Complete Guide to Harver's Game-Based Test
What Is the Pymetrics Assessment 2026 Candidates Actually Face?
If a recruiter has just told you to complete a "Pymetrics assessment" before your first interview, you have probably already done what most candidates do: opened a search engine, typed the name into it, and found a confusing mix of forum posts calling it a "balloon game" and prep sites promising to help you "beat the algorithm." Here is the honest version. The Pymetrics assessment 2026 candidates encounter is a set of 12 short, neuroscience-based mini-games, usually completed in about 20 to 25 minutes, that measure how you pay attention, make decisions under uncertainty, tolerate risk, remember sequences, read emotional cues, and adapt when the rules of a task change partway through. There are no questions to answer and no instructions telling you what a "correct" response looks like. You just play.
Pymetrics was founded in 2013 by Frida Polli and Julie Yoo, built on research out of neuroscience and behavioral economics labs, and in August 2022 it was acquired by Harver, a talent-assessment company that has since folded Pymetrics' game engine into its own platform under the "Harver" brand while keeping the original games largely intact. That is why you will see the tool referred to interchangeably as "Pymetrics," "Harver games," or a "Harver game-based assessment" depending on which employer's careers page you land on — they are the same underlying technology. This Harver game-based assessment guide treats the two names as one system, because for a candidate, the experience is identical regardless of which logo appears on the loading screen.
This matters because Pymetrics is not a substitute or a variant of the personality questionnaires and reasoning tests you may already know. It sits in a genuinely different category of pre-employment assessment, and understanding that distinction is the single most useful thing you can do before you sit down to play.
How Pymetrics Differs From SHL, Hogan, and Traditional Reasoning Tests
Most assessment guides on the internet — including some of our own — cover tools like SHL's verbal and numerical reasoning batteries, the Hogan Personality Inventory, Watson-Glaser critical thinking tests, or cognitive screens like the CCAT, Wonderlic, or Predictive Index. Those tools share a common format: you read a prompt (a passage, a data table, a statement about yourself) and select an answer from a fixed set of options. Some of those answers are objectively correct, as with a numerical reasoning question. Others, like personality inventory items, are self-reported and gameable to a degree, since you are describing yourself in words.
Pymetrics abandons that format entirely. There is no reading passage, no multiple-choice item, no self-description prompt. Instead, you play games — pump a virtual balloon, sort cards, tap a key when a shape appears, decide whether to keep money or share it with a stranger — and the system captures dozens of behavioral signals per game: how quickly you react, how your reaction time changes as risk increases, how consistently you behave across dozens of trials, and how you adjust after a loss versus a win. It is measuring behavior directly rather than asking you to report on your own behavior, which is precisely why it draws on neuroscience and behavioral-economics research rather than the psychometric tradition that produced tools like SHL and Hogan.
If you have already read our guide to the SHL assessment test or our breakdown of the Hogan Assessment, it's worth holding two ideas at once: those tests reward preparation strategies like practicing question formats and pacing yourself against a clock, while Pymetrics rewards something closer to the opposite — showing up rested, curious, and unguarded, because the entire point of the tool is to capture your natural behavioral patterns rather than your ability to select the "best" answer under time pressure. We'll come back to what "preparation" honestly looks like for a tool built this way, but the short version is: you cannot cram for Pymetrics the way you cram for a numerical reasoning test, and trying to will likely backfire.
The 12 Games: What Each One Actually Measures
Pymetrics' library has evolved over the years and some employers deploy a subset rather than the full battery, but candidates most commonly report encountering a core set of games. Here is what each one is actually assessing, based on how Pymetrics and independent prep resources describe them.
Balloons (the risk game)
You are shown a virtual balloon and each pump adds a small amount, commonly five cents, to a running total. You can stop and "bank" the money at any point, but the balloon can pop, at which point you lose everything earned in that round. This is a version of the well-established Balloon Analogue Risk Task (BART) used in behavioral economics research. It measures risk tolerance and reward sensitivity — specifically, whether you keep pushing for more after a string of successes, and how you adjust your pumping after a balloon pops.
Money Exchange 1 and 2 (trust and generosity games)
These are built on the classic "trust game" and "ultimatum game" from behavioral economics. In one version you are given a sum of money and decide how much to send to an anonymous partner, knowing the amount you send will be multiplied before it reaches them; in the other, you respond to offers from a partner and decide whether to accept a split you consider unfair. There is no "correct" allocation. These games are reading signals related to trust, fairness sensitivity, and generosity, and by design should never generate a pass/fail score — they contribute to a broader trait profile.
Digits / Memory Cards (working memory game)
A sequence of numbers or cards flashes briefly and you must recall or reproduce the order. This measures working memory capacity and how it holds up as sequences get longer and more complex, which is a proxy for attention and cognitive load management.
Easy or Hard (effort allocation game)
You repeatedly choose between an "easy" task offering a small reward and a "hard" task offering a larger one, with the actual difficulty and payout ratios shifting throughout. This measures how you weigh effort against reward, and whether you gravitate toward or away from harder challenges when the stakes rise.
Stop Game (impulse control task)
A classic "go/no-go" task: you are told to press a key rapidly for most trials but withhold the press when a specific stop signal appears. This measures impulse control and the ability to inhibit an automatic response, which correlates with attentiveness and self-regulation under time pressure.
Cards (learning and reversal game)
You repeatedly pick from several decks of cards, some of which are set up to be net-positive and some net-negative over time, and the "good" and "bad" decks quietly swap partway through. This is a reversal-learning task: it measures how quickly you adapt your strategy once the rules change, a trait closely tied to learning agility.
Arrows (attention and sequencing game)
Arrows or shapes point in different directions and you must respond according to a rule that periodically flips (respond to the direction the arrow points, then later respond to the opposite direction, or a related switching rule). This captures sustained attention and cognitive flexibility.
Faces (emotional intelligence game)
A face flashes on screen for a fraction of a second and you identify the emotion being displayed — happiness, fear, anger, surprise, disgust, sadness. This is a direct read on emotional-recognition accuracy and speed, often reported as one of the more memorable games because it feels the most like a "test" in the traditional sense.
Lengths / Keypresses (fine-grained motor and perception games)
Short games asking you to judge relative line lengths or tap keys in response to visual cues, generating additional data points on perceptual speed and consistency that feed into the broader trait model alongside the higher-profile games above.
Towers (planning game)
A Tower of London-style puzzle where you rearrange disks or blocks to match a target configuration in the fewest moves. This measures forward planning and problem-solving strategy rather than raw processing speed.
Taken together, these 12 or so games are not scored individually the way a single SHL question is marked right or wrong. They generate a behavioral fingerprint across dozens of micro-metrics — reaction time variance, risk-adjustment curves, learning speed after a rule change, emotional-recognition accuracy — and it is that combined fingerprint that gets compared against a benchmark. Which brings us to the part candidates most want to understand: how is any of this actually scored?
How Pymetrics Scoring Actually Works
There are two distinct scoring approaches Pymetrics/Harver can run, and which one applies to you depends entirely on how the hiring employer configured the tool.
Model 1: Comparison against a high-performer profile. Many enterprise clients — historically this includes firms like Kraft Heinz, McDonald's, and various consulting and banking employers — first ask a sample of their own existing top-performing employees in a given role to play the same 12 games. Pymetrics' algorithm identifies the behavioral patterns those high performers share (for instance, a particular risk-tolerance range combined with fast reversal-learning) and builds a role-specific benchmark profile. When you play the games as a candidate, your behavioral fingerprint is compared statistically against that benchmark, and the system flags how closely your patterns align with the traits that predicted success in that specific role at that specific company. This is why the "right" way to play Pymetrics genuinely differs by employer and by role — there is no universal winning profile, because the benchmark itself is built from a different group of people each time.
Model 2: Direct trait measurement against a general population. For some roles, or where no internal high-performer sample exists yet, Pymetrics instead measures your traits directly — your risk tolerance, attention span, emotional-recognition accuracy, and so on — and reports where you fall on established behavioral-science scales relative to a general population, without necessarily "passing" or "failing" you against a single number.
In both cases, the underlying mechanism has drawn serious scrutiny, and Pymetrics has responded to that scrutiny in a way that is genuinely worth knowing about if you are nervous about being judged by an opaque algorithm. The company built and open-sourced Audit-AI, a bias-testing tool, and in one of the more unusual moves in this industry, commissioned an independent audit of its own hiring algorithm from computer scientists at Northeastern University, published as a peer-reviewed paper on fairness in algorithmic hiring assessment. The audit tested Pymetrics' models against the "four-fifths rule," a longstanding U.S. Equal Employment Opportunity Commission benchmark requiring that no demographic group pass a selection process at a rate below 80% of the highest-passing group. As reporting from MIT Technology Review on the audit noted, the Northeastern team concluded Pymetrics was "doing a really good job" against that specific fairness threshold, while also cautioning that the audit could not confirm the tool's predictions actually correlate with real job performance, nor rule out every possible form of bias. It's a genuinely more transparent posture than most assessment vendors take, even if it isn't a perfect guarantee — and it's a detail worth knowing if the idea of "an algorithm judging my behavior" is what's making you anxious about the test in the first place.
Which Companies Use Pymetrics in 2026
Pymetrics and its Harver-branded successor are used across a wide swath of industries, and the specific traits an employer emphasizes tend to track the nature of the role. Some of the most prominent users worldwide include:
- Boston Consulting Group (BCG) uses Pymetrics as an early screen in its graduate and experienced-hire recruiting pipelines across multiple regional offices, alongside its case interview process.
- JPMorgan Chase deploys it within early-career and campus recruiting, particularly for roles where risk judgment and steady decision-making under pressure matter.
- Accenture uses it at high volume for graduate and analyst-level behavioral screening, given the scale of its global hiring.
- Unilever was one of the earliest and most publicized adopters, using Pymetrics as part of a redesigned graduate hiring funnel intended to match candidates' soft skills to its culture and reduce reliance on CVs alone.
- Mastercard leans on the assessment's emotional-intelligence and learning-agility signals for roles that require adapting quickly to new products and markets.
- McDonald's has used it to assess fairness, consistency, and teamwork traits, particularly for high-volume frontline and management-track hiring.
- Blackstone uses it in identifying high-potential talent through cognitive and behavioral profiling for competitive finance roles.
- UPS has used it to assess attentional focus, reaction time, and multitasking capacity relevant to operational roles.
- LinkedIn and Kraft Heinz are also among the enterprise clients that have publicly used Pymetrics-based assessments in parts of their hiring funnels.
This list keeps growing, and availability by region and role varies year to year, so if your recruiter mentions Pymetrics or Harver, don't assume the exact same version of the test is running everywhere — configurations differ by company and sometimes by office.
The geographic reach is genuinely global. Because the underlying games are language-light (recognizing an emotion or pumping a balloon doesn't require reading comprehension the way a verbal reasoning test does), Harver states the platform is deployed in more than 100 countries and 30 languages. In practice, this means a graduate applicant in Mumbai applying to a BCG or Unilever program, a finance candidate in London interviewing with JPMorgan, and an operations hire in Chicago applying to UPS may all be playing a very similar set of 12 games, even though their interview processes otherwise look completely different. If you're an Indian candidate applying to a multinational graduate scheme and this is your first time seeing a game-based assessment rather than a written test, that's completely normal — it's simply a newer category of tool that happens to travel well across borders precisely because it doesn't rely on written language the way older assessments do.
Can You Actually Prepare for Pymetrics? Here's What Honestly Helps
This is the question everyone actually wants answered, and the honest answer has two parts: no, you cannot prepare for Pymetrics the way you prepare for a numerical reasoning test, but yes, there are several things worth doing beforehand that are legitimate and not remotely "cheating."
Why you can't cram. Traditional test-prep works because there is a stable, learnable skill underneath the test — vocabulary, arithmetic, logical inference — and practice genuinely improves your raw ability to answer correctly. Pymetrics isn't measuring a learnable skill in that sense. It is measuring behavioral tendencies that are, by design, supposed to be relatively stable across repeated play: how much risk you take when uncertain, how quickly you adjust after a rule changes, how accurately you read a fearful versus an angry face. Trying to consciously perform a "better" version of yourself — pumping the balloon less because you've decided low risk-tolerance looks safer, or slowing your reactions because you think it looks careful — usually just adds noise and inconsistency to your own behavioral signal, which is the opposite of what you want, since Pymetrics' fairness and validity claims rest specifically on measuring genuine, consistent behavior.
What legitimately helps instead:
-
Get a real night's sleep beforehand. Attention, working memory, and impulse control — three of the core traits these games measure — are all measurably degraded by poor sleep. This is not a trick; it is the single most evidence-backed thing you can do to make sure the version of you the test sees is your normal, functioning self rather than a sleep-deprived outlier.
-
Understand the format in advance so it doesn't rattle you. A meaningful amount of "bad" performance on games like Pymetrics doesn't come from bad underlying traits — it comes from candidates freezing because they didn't expect a balloon-pumping game to be part of a job application and spend the first 60 seconds confused rather than playing naturally. Simply knowing that you'll see a mix of memory games, a risk game, a face-emotion game, and a card-sorting game removes that shock factor, and lets your actual behavior come through cleanly rather than being muddied by initial confusion.
-
Play in a quiet, uninterrupted environment on a reliable device. Several of the games are timing-sensitive at the millisecond level. A laggy trackpad, a spotty Wi-Fi connection, or a sibling interrupting you mid-game can distort reaction-time data in ways that have nothing to do with your actual attentiveness. Treat the environment with the same seriousness you would a proctored exam.
-
Don't try to "solve" the trust and generosity games. There's no scoring advantage to being maximally generous or maximally self-interested in the Money Exchange games — remember, there's no universal "correct" allocation, and companies that use these signals responsibly are matching for role fit, not virtue. Respond the way you actually would.
-
Treat your overall interview prep — not just this one tool — as the priority. Pymetrics is typically an early-stage screen, not the final word. The behavioral interview, technical round, or case interview that follows still rewards genuine preparation. If you're heading into behavioral interview rounds after clearing a game-based screen, it's worth using a structured method like the STAR framework to organize your stories — ClavePrep's STAR Builder is built for exactly that step, and our broader library of free interview prep tools covers everything from mock interviews to resume alignment for the rounds that come after Pymetrics.
A Sample Game Walkthrough: What the Experience Actually Feels Like
To make this less abstract, here's a realistic walkthrough of what a first-time candidate typically experiences, stitched together from how these games are commonly described by people who've taken them.
You click the assessment link from an email and land on a welcome screen explaining, in a sentence or two, that you'll play a series of short games and that there are no right or wrong answers — just play naturally. There's a brief practice round for each game before it "counts," which is worth taking seriously rather than skipping.
The first game might be Balloons. A cartoon balloon sits in the middle of the screen with a "pump" button. You click it once; the balloon grows slightly and a small dollar amount ticks up in the corner. You have no idea, and are not told, when it might pop. After a few pumps you get nervous and click "collect," banking a small amount. On the next balloon you push further before collecting. On the third, it pops before you can collect, and your total for that round resets to zero. This repeats for a couple dozen balloons of varying (and randomized) "pop points." There's no way to learn the "trick" — the point is simply to see your natural pattern of risk-taking and how you adjust after a loss.
Next might be the Faces game. A face flashes for under a second, then disappears, and you're asked to select which emotion it displayed from a short list. The exposure is quick enough that guessing purely on gut reaction is the intended mode of response — overthinking often hurts your speed without helping your accuracy.
Later you might hit the Cards game: four decks appear on screen, and you pick a card from any deck, revealing a small win or loss. Over 20 to 30 picks, you start to notice some decks feel "luckier" than others — until, without warning, the pattern flips. The game isn't testing whether you find the best deck; it's testing how quickly you notice and adapt once your working theory stops being true.
By the end — typically 10 to 12 games, 20 to 25 minutes total — the platform has recorded hundreds of individual data points: reaction times down to the millisecond, choice patterns, adjustment behavior after wins and losses, emotional-recognition accuracy, and working-memory span. None of it was presented to you as a quiz, and that's deliberate.
The Days Before Your Assessment: A Practical Checklist
- Confirm the deadline and give yourself a buffer day; most Pymetrics/Harver links have a window of several days to a week, and rushing through it late at night after a long day is the single easiest way to underperform on attention and impulse-control games.
- Test your device and connection beforehand if the platform allows a practice run; use a laptop or desktop rather than a phone where possible, since several games rely on precise key-press timing.
- Get a full night's sleep the night before — this is worth repeating because it is the most evidence-backed lever you actually control.
- Find 30 uninterrupted minutes in a quiet space; close other tabs and silence notifications, since several games are timed at the individual-response level.
- Skim a game-type overview like the one above once, so the format itself isn't a surprise, then stop reading about it — over-researching or trying to reverse-engineer "ideal" answers tends to backfire on tools designed to read natural behavior.
- Remind yourself this is usually one stage among several. If you clear it, the next round is likely a behavioral or case interview where structured preparation genuinely pays off; that's a good moment to revisit our guide on how ClavePrep's interview practice tools work so you're not starting cold on the next stage.
Frequently asked questions
Is Pymetrics the same as an IQ test? No. Pymetrics measures behavioral traits like risk tolerance, attention, and emotional recognition rather than crystallized knowledge or fluid reasoning ability the way an IQ test or numerical reasoning test does. Some of the underlying traits, like working memory, overlap with cognitive testing, but the format and purpose are different.
Can I fail a Pymetrics assessment? Not in the traditional sense of a wrong answer. What can happen is that your behavioral profile doesn't align closely enough with the benchmark a specific employer built for a specific role, which can result in not advancing. That's a fit signal for that role at that company, not a universal judgment of your abilities.
Do Pymetrics and Harver refer to the same test? Yes, functionally. Harver acquired Pymetrics in 2022 and has since integrated the original 12-game engine into its broader assessment platform. You may see the games branded as "Harver" on some employer career pages and "Pymetrics" on others, but the underlying games and mechanics are the same.
How is Pymetrics different from SHL or Hogan assessments? SHL's reasoning tests and Hogan's personality inventory both use written items — questions, passages, or self-description statements — where you select from fixed answer choices. Pymetrics uses no written questions at all; it captures behavior directly through gameplay. If you want a deeper look at those other formats, our guides to the SHL assessment and Hogan Assessment cover them in detail.
Should I try to appear low-risk or highly emotionally intelligent on purpose? It's generally not advisable. The games are designed to be difficult to consciously game, and inconsistent or artificial-feeling behavior patterns can themselves read as noise rather than as a deliberate "good" signal. Playing naturally and rested is the more reliable strategy.
Is Pymetrics biased against certain groups? Pymetrics has taken a more public stance on this than most vendors, building an open-source bias-auditing tool and commissioning an independent academic audit of its own algorithms, published through Northeastern University researchers. That audit found the tool met a common U.S. fairness benchmark (the four-fifths rule) for the groups tested, though it also acknowledged the audit could not verify every dimension of fairness or confirm real-world job-performance correlation. It's a genuinely more transparent track record than most assessment vendors, though not an absolute guarantee.
How long does the Pymetrics or Harver assessment take? Most candidates report finishing the full set of games in 20 to 25 minutes, with individual games lasting one to three minutes each. Give yourself a buffer of about 30 minutes so you're not rushed near the end.
What should I do if I don't understand a game's instructions? Each game typically includes a short practice round before the scored version begins. Use it. If instructions are genuinely unclear, most employer-facing versions include a support contact for technical issues — reach out rather than guessing blindly, since data collected from a confused first attempt won't reflect your natural behavior.
Moving Forward After the Games
Clearing a Pymetrics or Harver screen is often just the entry point into a longer process — a recruiter call, a behavioral interview, a case study, or a technical round, depending on the employer and role. That's where deliberate practice actually does move the needle, and it's worth putting your energy there once the game-based stage is behind you. If your next step involves telling structured stories about your past experience, ClavePrep's STAR Builder can help you turn rough memories into clear, interview-ready answers, and our full set of interview prep tools covers mock interviews, resume feedback, and more for whatever comes next in your pipeline. You've already gotten through the part nobody can coach you on — the rest is where preparation genuinely counts.
