Run online recruitment tests for your candidates

CV screening and the pre-screening questions in your ATS answer constraints: commuting distance, availability date, software already used. They do not say whether a candidate can write a clear summary, handle an ordinary situation of the job or follow an instruction. That is what a test measures, and it decides who you meet.

A test first protects the time of whoever makes the decision. When forty candidates are still in the running after screening, meeting everyone is impossible and choosing on instinct is risky. The same paper, the same questions and the same marking scheme for everyone, and you meet the top five or six knowing what they are top at.

It also keeps you from losing candidates. In a tight market, a process that drags costs you the best profiles: they accept elsewhere. A remote paper, open for a week and marked in one morning, replaces a group session to organise, a room to book and no-shows to chase.

An ATS posts, collects, screens and follows the pipeline. Evalmee runs the test and marks it. The two touch at a single point: the test link, pasted into the message your ATS sends to the candidates kept after screening.

In practice it takes four moves. You create one exam per posting, you set its share link to "anyone with the link", you set a closing date a week or two out, and you copy that link into your ATS template message. Candidates sit the test whenever they want inside that window, and you find them all in the same exam, ready to be compared.

There is no native connector to an ATS or an HR information system, and saying so avoids an unpleasant surprise on day one. What does exist: a public REST API on OAuth 2.0, Make and n8n connectors, and the Excel and PDF exports. The API is read by request, with no webhook before the top plan: it suits periodic synchronisation rather than an immediate trigger.

You write the test. Evalmee sells neither a ready-made library of job tests nor psychometric tests: it provides the question types, the marking scheme and the marking. Most recruitment tests are built from three or four blocks, much the same from one role to the next.

Start with the format that is missing almost everywhere else: the spoken question, which replaces the first phone call. The candidate reads the brief, gets a preparation time you set, then an answer time during which they record themselves, in audio or in video. A two-minute presentation after ten minutes of preparation is enough to judge ease of speaking, delivery and the ability to get to the point, which no written answer shows. A single take by default, and no way for the candidate to replay their answer: that is what makes the exercise comparable from one candidate to the next, and what you hear is an answer that was not rehearsed. That question type is in beta: it exists in the editor and can be used, and we would rather write that down than present it as settled.

Closed questions and their marking scheme, random draw from a question bankQuestion bankA reservoir of indexed, reusable questions, classified by topic, level or competency. It allows an exam to be assembled by random draw, papers to vary from one session to the next, and the statistical history that makes a question reliable to accumulate, item by item.See the full definition and importing a test already written in Word, PDF or Excel are described on the online multiple-choice test page. A test built once can be duplicated from one hire to the next: the same paper, with the three or four questions the role changes.

Three settings are enough, and they speak to everyone: a marking scale, a pass threshold, and criteria on the written questions. The rest is optional, and the vocabulary of competitive exams can stay on the shelf.

The scale sets the total, most often out of 100. The pass threshold is the point above which a paper counts as passed. It sorts nobody on your behalf: it colours the list and saves you the time of the first glance.

On a written question, the rubricRubricA grid crossing assessment criteria with performance levels described in full sentences, from the weakest to the strongest. The marker places the paper in a cell rather than awarding a figure: the mark becomes the consequence of a description shared in advance.See the full definition is what replaces the general impression. You list what the answer has to contain, you put points against each line, and the paper comes out with a mark per criterion next to the overall mark. Two candidates on 12 out of 20 stop being interchangeable: one has the right information and clumsy wording, the other the reverse.

Tags on questions let you read results by topic: you export the marks to Excel and look, section by section, at who missed the questions that matter for this role. One useful clarification, because the question comes up often: there is no knock-out question and no automatic alert when one specific question is failed. The threshold applies to the overall mark; spotting a weak question is done by reading.

Closed questions take no time at all: they are marked automatically by the scheme, with no marker and no AI. What takes time is the written answers, and that is where the marking screen changes the pace.

Marking is done question by question rather than paper by paper: every candidate's answer to the same question fits on one screen. You apply the same criterion forty times in a row instead of rebuilding it on every paper, and the gap between the first and the last paper marked narrows accordingly.

On those answers, AI-assisted markingAI-assisted markingPre-marking of written answers by a language model, which proposes a mark and a justification that the marker validates, amends or rejects. The decision stays human: what the AI saves is the first pass over the paper, never the responsibility for the mark.See the full definition suggests points and feedback from the rubric you locked before the test. Nothing counts towards the mark without your approval click, answer by answer. The data sent to the AI provider is anonymised, and the processing takes place in the European Union. The full sequence, from rubric to mark, is on the marking page.

Spoken questions count separately: a recording is watched in real time. Forty two-minute answers make eighty minutes of listening, and they cannot be skimmed the way a text can. So a spoken question belongs at the end of the test, and on an already short list: ten candidates rather than forty, with a brief precise enough that two minutes are enough.

What comes out: the marks in Excel, the marked papers as PDF, one per candidate or all in a single document. That Excel file is what you paste back into your ATS, or hand to the company you are hiring for.

A test taken at home can be cheated: a second tab, a phone, someone sitting next to them. Evalmee does not claim to prevent it. It detects the events listed below, timestamps them and hands them back in an integrity report reserved for the recruiter.

The risk score closes no exam and disqualifies nobody: you are the one who opens the report, reads the events and their timestamps, and decides. Exam misconductExam misconductAny conduct aimed at obtaining a mark that does not reflect the candidate's own work: unauthorised documents, communication with a third party, impersonation, undeclared use of a generative tool. How it is classified and what follows from it are set by the institution's regulations.See the full definition is established, not computed. No behavioural analysis is performed, neither gaze, nor emotions, nor typing dynamics. What is recorded, who can see it and how long it is kept is written on the online exam proctoring page.

The limit is worth stating plainly, because it sometimes decides a choice of tool. Evalmee does not impose a locked-down browserBrowser lockdownRestriction of the candidate's software environment during the exam: full screen imposed, tab switching blocked or flagged, copy-paste and printing disabled. The measure raises the cost of opportunistic cheating, but can do nothing about a second device sitting next to the computer.See the full definition and does not lock the candidate's machine. A platform such as TestWe goes further on that point: it locks the computer for the duration of the test. Evalmee detects and reports. For a screening test before an interview, that level is usually enough; for a paper a qualification depends on, the question is a different one. The French platforms are compared criterion by criterion in the comparison of online exam platforms.

An outsourced HR consultant handles candidate data that is not their own, and answers to their clients for the tool they chose. Three facts matter then: where the data sits, what contract covers it, and how long it stays.

Papers, results and proctoring recordings are hosted in France, with Scaleway and in the AWS eu-west-3 region in Paris, with no replication outside Europe, backups included. A data processing agreement compliant with article 28 of the GDPR is provided at no extra cost, and is better asked for during evaluation than on the eve of signing.

The retention periodData retention periodThe length of time personal data may remain in an active database. The CNIL points out that it is determined by the purpose of the processing: at the end of that period the data must be deleted, anonymised or archived, which bears directly on remote proctoring recordings.See the full definition for the webcam and microphone recordings produced by proctoring is three months by default, the time it takes to handle a dispute; screenshots, attached to the candidate's submission, follow the retention period you set for submissions and are deleted along with them. The standard product uses neither facial recognition nor biometric identification, and does not analyse the candidate's behaviour: automatic analysis of gaze, emotions or typing dynamics is not part of the product.

What that does not say: compliance is judged processing by processing, and you, or your client, are the one accountable for it. What Evalmee documents on its side, hosts, sub-processors, transfers and retention periods, is set out on the security page.

A short list, worth reading before you try it. Evalmee covers one step of hiring, not hiring.

It is not an ATS. Evalmee does not post jobs, does not source candidates, does not screen CVs, and manages neither pipeline, nor statuses, nor reminders. There is no native connector to an ATS or an HR information system: the handover happens through the test link on the way out and an Excel export on the way back, or through the API for anyone who wants to automate it.

It is not a test catalogue. You will find neither ready-made job tests, nor psychometric tests, nor personality tests. Evalmee provides the question types and the marking; the content of the test is yours to write, alone or with the person you are hiring for. If what you want is a catalogue to buy, other vendors have made it their trade.

It is not an interview tool. Evalmee does not schedule slots, does not run video calls and holds no live interview, neither one to one nor in front of a panel. The spoken question, still in beta, records an answer that is prepared and then timed; it does not replace the conversation, it decides who you invite in to have it.

It is neither a locked-down machine nor an automatic decision. Proctoring detects and reports; the risk score rules nobody out; the mark hires nobody. When the stakes call for a closed environment, software that locks the computer remains the right choice, and it is better to know that before the trial than after.