The Harver Assessment Explained: Every Module
Harver is a platform, not one test. Which modules you may meet, how each is scored, how it relates to NOA and pymetrics, and how to prepare.
Harver was founded in Amsterdam in 2015 as a hiring service for call centres and grew into a platform for high-volume employers: supermarkets, restaurants, retail chains, security firms and outsourced customer service. The US assessment company Outmatch bought it in 2021 and took the Harver name, and Harver bought pymetrics in 2022. So an invitation that says Harver tells you who runs the platform, not what you are about to sit.
How a Harver journey works
The employer assembles a candidate journey from modules, and you move through them in order, usually at a page such as journey.harver.com carrying the employer’s own branding. Harver’s own candidate help centre points out that Harver is a separate company from the one you are applying to. An Albert Heijn store application, one of Harver’s best documented cases, takes about ten minutes in total.
Three rules apply to every module. You can log out between modules and your progress is saved, but leaving part way through a module means starting it again. You cannot go back and redo an assessment once your application is submitted; a retake is the employer’s decision and deletes your earlier results. And you only see your results if the employer switches that on, under a My Results button.
The modules you may meet
Personality Print is the one you are most likely to meet, because it is mandatory on every vacancy built from Harver’s standard job profiles. Beyond these, an employer can add a language test (licensed from LTI), a spoken language test, a writing assessment, a job knowledge test written by the employer, a content moderator assessment and a video interview. Older journeys may still carry a HEXACO personality questionnaire or a Learning Agility module, both now closed to new accounts.
How each one is scored
Harver uses different kinds of result for different modules, and knowing which one applies tells you what a number means.
- Cognitive subtests and the chat simulation are reported on a norm scale from 0 to 100% where 50% is the average candidate and most people score between 30% and 70%. That is not a percentile: on a percentile scale only 40% of people land between 30 and 70. The cognitive score depends on how many items you get right, and Harver describes no deduction for a wrong answer.
- The situational judgement test is scored per pick: +1 when your best or worst matches the key, -1 when your best is the keyed worst or your worst is the keyed best, and 0 for a neutral option. You see a percentage. Harver’s own example report labels an 88% retail SJT a Great Fit.
- The chat simulation gives an Overall plus four parts: Clear communication, Customer orientation, Problem solving and Handletime. The first three are about the quality of your replies, and Handletime about speed.
- The typing test is net words per minute, typed words minus mistyped words, divided by time. Harver publishes its norm: an average of 38 net words per minute with a standard deviation of 17.
- Personality Print has no better or worse. Harver’s report guide says high scores are not necessarily better and low scores are not necessarily bad, and the feedback report it can send candidates shows each trait in one of three bands with no numbers.
None of those is the decision. The employer screens on an overall matching score that combines the modules against the vacancy’s profile, with its own Good Fit and Great Fit brackets. On a standard job profile each cognitive subtest is roughly a tenth of that match, which is why a weak subtest is not an automatic rejection on its own, and why no pass mark is published.
Where NOA and pymetrics fit
Harver’s knowledge base calls its cognitive module "NOA Cognitive Ability", and its vacancy builder prints the module type "NOA" under each subtest. NOA is an Amsterdam test publisher, and Harver’s five subtests are five of its eight item types. The suffix you may see on a module name, such as "Cognitive Ability - Number Sequence (Bachelors)", is the education level you are compared with, which is how NOA norms its tests. The difference is the report: NOA’s own candidates see a stanine band, Harver’s see the 0 to 100% scale. If you want the NOA background, our guide to the NOA test covers it.
pymetrics is simpler to spot. If your journey contains short games, with balloons to pump, towers to move or money to share, you are sitting pymetrics, and the pymetrics practice games are what to prepare with. An employer can mix the two in one journey.
The Brain Game and other things you may read about
Older preparation material describes a Harver Brain Game with three puzzles: Path Finding, Crack the Code and The Puzzle. Harver’s own pages show it was replaced by the five cognitive subtests in 2019, so if you are reading about it, the source is out of date. The same goes for Harver CHAT, a recruiting chatbot that was not an assessment and was discontinued in May 2025, and for the multitasking card games, which no longer appear in Harver’s module catalogue.
How to tell your modules from the invitation
The invitation often lists the steps, and each module opens with a title screen that names it and gives an approximate time: "Approx. 12 minutes / Live Chat Support Simulation", for example. A cognitive module name starting "Cognitive Ability -" is one of the five subtests. "Situational Judgement Test" or, on older journeys, "Work Simulation Game" means best-and-worst scenarios. If the employer is a contact centre or a chat support operation, expect the chat simulation or the typing test; if it is a store or a restaurant, expect Personality Print and an SJT.
How to prepare
Our Harver practice tests cover the five cognitive subtests, the situational judgement test, the live chat simulation, the typing test and Personality Print, with feedback written in Harver’s own reporting terms. If you want to know whether any of it can be short-circuited, we answer that at can you cheat Harver.