Criterion-referenced assessment
Competency with confidence.
Criterium is a competency assessment platform for organisations whose results have to stand up to scrutiny. Every response is kept as evidence, and every decision can be reconstructed later.
Built on four commitments
Criterium is built as a governed assessment system, in the tradition of regulated assessment. Four commitments run through the whole platform.
The server decides.
Every screen a user sees is set by their role and the current state of the assessment, enforced on the server. Marks, timings, and access never depend on trusting the browser.
Evidence is captured once and never rewritten.
Candidate responses, marks, and moderation decisions are recorded as evidence with full provenance. A correction never overwrites history. It is added to it, attributed and timestamped.
Explicit states
Assessments and results move through defined states: draft, in review, approved, published. What a candidate sits, and what a result says, is always a published, governed version.
Every decision can be reconstructed.
Who marked which response, when, under which version of the mark scheme, and what changed afterwards: the record is there to answer the question before it is asked, whether by a candidate, an appeal panel, or a regulator.
Four roles, four bounded views
Access is role-based and strictly scoped. Each person sees exactly what their role needs, and no more.
Administrator
Authors assessments, governs their lifecycle, sets up sessions, and manages the organisation's question banks and reporting.
Invigilator
Runs live sessions: admits candidates, monitors attempts in progress, and handles interruptions with a recorded, auditable trail.
Marker
Marks responses against the published mark scheme, records feedback, and takes part in moderation. Every judgement is attributed.
Candidate
Joins with a short access code, sits the assessment in a focused environment, and receives results when they are formally released.
A governed lifecycle, end to end
Nothing reaches a candidate, and no result reaches the record, without passing through explicit states. The order below is the order the system enforces.
Draft
An assessment is authored and revised freely. Nothing in draft can be delivered to a candidate.
In review
Content is checked against the standard before it can progress. Changes send it back to draft.
Approved
The assessment is fixed and versioned. What was approved is what will be delivered.
Published
Live for delivery. Scripts, marking, and results all reference this exact version, permanently.
The same discipline applies downstream: scripts, marking, moderation, and result release each have their own governed states, so a session can always answer one question — where is this, and who moved it there?
Measurement, taken seriously
Criterium is grounded in criterion-referenced measurement theory and a checked body of research.
Criterion-referenced by design
Results express performance against a fixed, published standard. A candidate's outcome is never a function of who else sat the assessment, which makes results comparable across sessions, cohorts, and time.
Confidence-based marking, used diagnostically
Candidates can record how confident they are in each response. Criterium uses this to show how well confidence matches performance. It informs feedback and development, and does not reduce scores.
Fairness as an analytical question
Reporting is built to support fairness and quality analysis of questions and results, so that is this assessment working as intended? is answered with data, not assertion.
A criterion-referenced result answers the only question that matters in certification: can this person do this thing, to this standard? — the design premise behind Criterium