For hiring teams
Levvy vs. take-home tests, skills assessments, and job simulations
Levvy is a work-sample (job-simulation) assessment platform that generates a custom challenge from a description of your actual role. Candidates apply by doing a few hours of the real work inside a simulated company — consulting AI colleagues, digging through materials, shipping deliverables — and you review how each person worked, not just what they produced. That combination, custom-per-role plus process-level evidence, is what separates it from templated tests and take-homes.
The category, plainly
Hiring teams trying to move beyond résumés usually reach for one of four things: a hand-written take-home test, a test-library platform of standardized skills assessments, a pre-built job-simulation product, or simply more interviews. Levvy sits in the work-sample/job-simulation category, with two deliberate differences: the challenge is generated from your role rather than templated, and the evaluation target is the candidate’s process — what they investigated, who they consulted, how they handled ambiguity, what they decided to ship — alongside the deliverables themselves.
Comparison at a glance
The table compares approaches as categories on the dimensions hiring teams actually weigh. It deliberately avoids vendor-by-vendor feature claims; individual products within a category vary.
| Approach | Custom to the actual role | What it evaluates | Candidate experience | Setup effort |
|---|---|---|---|---|
| Résumé screening | No — filters on credentials, not the role’s work | Self-reported history and proxies (titles, schools) | Low effort, low agency — most are filtered unseen | None, but every downstream interview inherits the noise |
| Traditional take-home test | Sometimes — if the team hand-authors it per role | The finished output only; process is invisible | Hours of solo work into a void, often unacknowledged | High — someone has to write, calibrate, and grade it |
| Test-library skills assessment | No — standardized tests picked from a catalog | Isolated skills via quiz-style items and scores | Fast but generic; feels like an exam, not the job | Low — pick tests, send links |
| Pre-built job simulation | Partially — realistic scenarios, but templated per job family | Behavior in a fixed scenario; mostly output-scored | More engaging than tests; same scenario for every company | Low to medium — configure from the vendor’s catalog |
| Levvy | Yes — generated from a description of your actual role | Process and deliverables: how the candidate investigated, decided, and shipped | A few hours of real work in a simulated company; the work is theirs to keep | Low — guided intake generates the challenge; you review and publish |
Named vendors — test libraries such as TestGorilla, or interview and coding platforms such as Karat and Codility — each fit one of the category rows above; evaluate any specific product on its own documentation.
Why work samples, and why process
The research case for work samples is old and strong. In Schmidt and Hunter’s 1998 meta-analysis of 85 years of selection research, work-sample tests showed a predictive validity of .54 for job performance — among the strongest single methods examined (later re-analyses put the figure lower, at .33, but still well-validated). The market has followed the evidence: TestGorilla’s State of Skills-Based Hiring 2025 report found 85% of employers now use some form of skills-based hiring, and 90% of those report a measurable reduction in mis-hires. The stakes make that reduction matter: a commonly cited U.S. Department of Labor estimate puts the cost of a bad hire at roughly 30% of the employee’s first-year earnings — before counting team drag and opportunity cost.
Process matters because output stopped being trustworthy. Any candidate can now produce a polished artifact with AI assistance; the artifact alone no longer tells you who can think. Watching howthe work happened — what a candidate investigated, which questions they asked, how they revised — is the signal that survives AI polish, and it is the layer templated tests and output-scored take-homes don’t capture.
What Levvy deliberately doesn’t do
- No scores or rankings. Dossiers carry behavioral evidence and per-skill signal, not a leaderboard. Numeric ranking of humans invites false precision and bias amplification.
- No AI hiring recommendation yet. A matching/recommendation layer is in active development; it ships when its predictive validity and adverse-impact properties are established, not before.
- No coaching or grading inside the workspace.The workspace AI never evaluates a candidate’s reasoning mid-challenge — validation would contaminate the very signal being observed.
Frequently asked questions
Is Levvy a take-home test?
It occupies the same slot in a hiring process, but the mechanics differ in two ways. First, the challenge is generated from a description of your actual role rather than picked from a template library, so it reproduces the real work. Second, candidates complete it inside a simulated workspace — with AI colleagues to consult, materials to investigate, and events that unfold — so you see how they worked, not only the final document.
What does the hiring manager see for each applicant?
A dossier of captured activity: the deliverables the candidate produced, how they were built, which resources the candidate unlocked through conversation, who they consulted, plus a per-skill readout of observed evidence against the role rubric. There are no scores or rankings — the evidence is behavioral and specific.
How much setup does a role challenge take?
You describe the role in a short guided intake; Levvy generates the skill focus and the challenge — the simulated company, colleagues, resources, and deliverables — and you review and edit before publishing. Most of the effort is reviewing, not authoring.
Is the candidate experience respectful of their time?
A challenge is a few focused hours of realistic work with a visible brief and concrete deliverables. Candidates work in a real environment rather than answering quiz items, and the work they produce is genuine work product — not throwaway puzzle answers.
Does Levvy rank candidates or recommend who to hire?
Not today. Levvy surfaces process evidence and per-skill signals; the judgment stays with the hiring manager. An AI matching and recommendation layer is in active development, and it is deliberately not shipped until its predictive validity and bias properties are established.
What does Levvy cost?
Levvy is in pilot and free for hiring managers during the pilot, and free for candidates. Sign-ups are invite-gated on the hiring side while the pilot runs.