Measurement, not assessment.
How every number is made.
A test gives you a mark. A measurement tells you how much to trust the mark. This page explains, in plain words, how every number on the Card is made.
Skills a job actually needs.
Every role is broken into the skills it needs, grouped into skill clusters. Each skill and its bar were defined with people who hire for that role, at fresher, mid and senior level.
- Skills
- 8
- Clusters
- 3
- Must-haves
- 3
Switch the level. The skills stay the same. The bar moves.
Thinking
- Problem solving70
- Analysis72
- Data literacy58
People
- Communication65
- Stakeholder handling62
Delivery
- Requirements60
- Documentation55
- Prioritisation56
Illustrative bars for one role. Real bars are set per role with people who hire for it.
Every question, one skill.
We do not score a test. We score skills, using every question, answer and practice session that touches them.
- AI Assessment Case studyProblem solving
A customer’s churn doubled in one month. Which driver would you check first, and why?
- AI Assessment Multiple choiceAnalysis
Here is last quarter’s sign-up funnel. Where is the biggest drop?
- AI Interview Managerial roundStakeholder handling
Explain a two-week delay to a client who is already upset.
Sample questions. The skill gets the score, not the test.
Harder questions count more.
The engine uses item response theory and Bayesian updating. In plain words: a hard question answered well counts more than an easy one, a lucky guess counts less, and each new answer updates the score rather than replacing it.
- EasyCorrectCounts a little56
- HardCorrectCounts more66
- MediumWrongPulls it back, not to zero62
- HardCorrect in 4 secondsLikely a guess, counts less63
- HardCorrectCounts more, range narrows70
Before any answers, the engine knows little. The range is wide.
The rules every score follows.
Written down so that no page, pitch or product change can quietly break them.
- Every attempt gets its own fresh score.An assessment, interview, mock interview or Communication Studio session is scored on its own. The Card then builds up across all of them for that role.
- Employer-set, proctored work counts in full.Toward fitment for the role the employer is hiring for.
- Practice you start yourself counts at half weight for employers.It counts in full toward your own readiness, and it is always labelled as practice.
- Coder Arena builds readiness only.It never feeds employer fitment.
- AI Career Coach only reads.It produces no evidence and never changes a score.
- A shortlist needs enough evidence.Any activity that feeds a shortlist has at least 8 questions per skill cluster.
- No separate personality test.Self-report questions live inside AI Assessment.
Confidence, not accuracy.
Every score comes with a range. A 74 with a range of 70 to 78 is ready to act on. A 74 with a range of 55 to 90 means measure more. We always show which one you are looking at.
74 likely 70 to 78. Narrow range. Act on it.
74 likely 55 to 90. Wide range. Measure more first.
Evidence, not results.
Every number links back to the questions, answers and sessions that produced it. A candidate can question it. An employer can audit it.
likely 70 to 78
- Problem solving
- Analysis
- Communication
What produced this score
“Broke the churn question into three drivers and tested the cheapest one first.”
“Summarised the trade-off in two sentences before going into detail.”
“Agreed to the new deadline without checking what it would push out.”
“Sized the market bottom-up and stated one assumption clearly.”
Who can check it
- The candidate, who can open any point and question it.
- The employer, who can audit any decision after it is made.
People decide.
The engine informs a decision. It never makes one. Hold for more evidence is always an option.
The range is wide, so holding for more evidence is suggested. Pick an action to preview it. Nothing is saved.
What we will publish.
Our scoring method, open for anyone to check. Coming soon.
- Scoring methodHow skills, weights, ranges and bars are calculated, written for anyone to check.
- Expert panelThe people who defined the role bars, named here once they agree to be listed.