Skills tests built fromyour job description.
Every candidate does a version of the role’s real work, from code and spreadsheets to a call with an unhappy customer, and is scored against the same rubric. Paste a job description to get one built for the role, or browse 36 tests to see what each one covers.
Free trial: 10 candidates or 14 days, whichever comes first. No card.
What candidates do
Nine ways to see the work.
The formats follow the job: an analyst gets a spreadsheet, a support hire gets a customer, an engineer gets code to fix. You choose which formats a role uses and the mix.
- 01
Written answers
Explain a decision or work through a problem. A polished answer with nothing behind it can get a follow-up that asks for specifics.
- 02
Multiple choice
Quick checks of what someone knows, with how confident they were in each answer.
- 03
Coding
Debug, finish, speed up or secure a snippet in the job’s language. The AI reviews the code line by line; it isn’t run against tests.
- 04
Spreadsheets
A working spreadsheet with Excel functions: build the formula, fix the model, answer from the data.
- 05
Email replies
Answer a realistic email thread, with its sender, subject and the people copied in.
- 06
Charts and dashboards
Read a chart, dashboard or diagram generated for the question, and say what it shows.
- 07
Chat simulations
Four to eight turns with an AI customer or counterpart: a role-play, an investigation with hidden facts, a teach-back or a negotiation.
- 08
Critiques
Find the planted flaws in a piece of work, a set of candidate profiles or a metrics readout.
- 09
Voice role-plays
A spoken call with an AI customer or buyer, up to ten minutes, scored from the audio on clarity, listening and handling objections.
Scoring
A score you can check.
Each candidate gets an overall score out of 10, a score out of 10 for every skill in the role’s rubric with the evidence for it, and a recommendation. Every score links to the answer behind it, so you can read the code, the spreadsheet or the call before you agree.
You also see four broader scores, for problem solving, expertise, professionalism and clarity, and a short written summary.
- Strong Hire · 9–10
- Hire · 7–8
- Consider · 5–6
- Not a Fit · 0–4
Data Analyst · candidate report
7.8/ 10
- SQL and data modelling8.5
- Analysis and judgement7.5
- Stakeholder communication8
- Excel and reporting6.5
Why 6.5 for Excel: the lookup was right, but the summary table hard-coded two totals that should have been formulas. See the answer →
An illustration of one candidate’s report.
Hard to game, by design.
No test is cheat-proof. These make the usual shortcuts stop working, and the security page lists exactly what proctoring captures.
- Questions are generated for each candidate, so there is nothing to share or look up
- Each question follows from the answer before it
- Copy and paste are blocked outside the code editor, and most questions run against a timer
- On paid plans, proctoring records the screen and flags tab switches, a second face or a phone
Test library
36 tests, explained.
Each page covers one test: what it measures, who should take it, the format and length, sample tasks and how it is scored. In Cohesyve, the assessment itself is always built from your job description.
Engineering
Data and finance
Business roles
Communication and thinking
Questions
Before you send one.
How long does an assessment take?
Can I write my own questions?
Is the code run?
Can candidates use ChatGPT?
Which languages can candidates take it in?
Do candidates need a laptop?
What does it cost?
Hiring for a specific skill? How to assess it, skill by skill.
Paste a job description. See the test.
The assessment, its rubric and its knockout questions are drafted in about a minute, for you to edit before anyone is invited.