Seven assessment types. One grading queue.
Pick the type, and the delivery, integrity, grading, feedback and retake rules arrive already set. Everything a person has to mark — assessments, lessons, programme work — lands in one place.
What Lurno does when a grade has to hold up
Lurno handles assessment end to end: authoring, secure delivery, marking and results. You choose one of seven assessment types — quiz, exam, practice, diagnostic, survey, portfolio task or psychometric instrument — and the type pre-configures delivery, integrity, grading, feedback and retakes, so an exam arrives locked down and timed while practice arrives untimed with unlimited penalty-free attempts. Questions come from 19 types. Everything a person has to mark lands in one queue spanning assessments, lessons and programmes, scored against analytic, holistic or single-point rubrics.
- Single-point rubric
- A rubric with one column instead of a grid. You describe only what proficient work looks like for each criterion, then note what fell short and what went beyond. Markers have less to click, learners have more to read. Lurno supports all three — analytic (criteria by levels, weighted), holistic (one scale for the whole piece), and single-point.
None of those defaults are fixed. Change any setting on one assessment, or set your organisation's own defaults for a type. Grading policies are built from ordered passes — automatic, manual, peer or AI-assisted, each required or blind — and an acceptance rule picks the grade: first, last, highest, average, unanimous, or a moderator's call.
A type is a set of decisions, not a label
Each one is a different pedagogical shape, so each one ships different defaults.
| Capability | Type | Time limit | Attempts | Secure mode | Back navigation | Correct answers |
|---|---|---|---|---|---|---|
| QuizAuto-marked; the highest of three attempts counts. | None | 3 | No | Yes | On submit | |
| ExamClipboard blocked, tab switches recorded as integrity events. | 60 minutes | 1 | Yes | No | After release | |
| PracticePenalty-free; the answer key waits until you release it. | None | Unlimited | No | Yes | On submit | |
| DiagnosticFinds out what someone knows already. The answer key never shows. | None | 1 | No | No | After release | |
| SurveyNo scores and no answer key — feedback, intake, pulse checks. | None | 1 | No | Yes | Never | |
| Portfolio taskArtefacts and milestone gates, submitted for assessor review. | None | Unlimited | No | Yes | After release | |
| Psychometric instrumentScored against scales and norm groups, not a mark out of ten. | None | 1 | No | No | Never |
Defaults, not limits. Every setting is editable on one assessment, and an organisation can override the whole matrix for a type.
The marking pile, sorted the way a teacher works
Assessments, lessons and programme work in one list
A marking pile is not sorted by which part of the product produced it, so this queue is not either. Filter by what the work belongs to, by pass, by grader or by lifecycle state, and take the next thing off the top.
- Score item by item, with written feedback and a voice note per item and overall.
- Two graders open the same submission and the second gets a clear conflict, not a silent overwrite.
- Assign a pass to a reviewer pool; reviewers claim, grade and submit. A blind pass hides other scores until yours is in.
A second pair of eyes where the stakes need one
Set a moderation threshold and anything above it is held for a moderator before release. A released grade is appealed inside the product rather than over email, and every result moves through a fixed twelve-state lifecycle.
- Appeals kept as a trail — released, appealed, resolved, finalised — plus a feedback thread on each result.
- Questions are library items, so one question can sit in several assessments. See authoring and the content library.
Portfolios, peer review, calibration and group submissions
Work that takes weeks gets proper handling, not a file upload field.
- Portfolios: artefacts added over time, milestones with gates to clear, and an assessor workbench moving each through review, revision requested, accepted or rejected.
- Peer review: anonymised reviewer pools with due dates, peer scores, feedback-quality ratings, and reconciliation when reviewers disagree.
- Calibration on anchor responses gives inter-rater reliability before real grading starts. Group submissions carry contribution weights per member.
Adaptive psychometric testing, and the evidence it was fair
Lurno runs psychometric instruments as measured tests rather than quizzes. You define scales, subscales and composites over the items, reverse-key and weight them, import norm groups and score-conversion tables, and draw interpretation bands with the narrative a candidate reads. Scoring is deterministic and versioned, so re-running an old session returns the same answer. An instrument in adaptive mode selects each next item by item response theory and stops once the measurement is precise enough. That is adaptive psychometric testing, not an adaptive course path.
- Item response theory (IRT)
- A way of modelling a test where each question carries its own numbers — how hard it is, how sharply it separates strong from weak, how likely a guess is — while the candidate carries a single ability estimate. Both sit on the same scale, so the test can pick the question that tells it the most next, and stop when the estimate is precise enough.
Fairness work is kept as a record, not a report you export and lose. Differential item functioning (Mantel-Haenszel, with effect size and flag level), validity evidence, adverse-impact checks against the four-fifths rule, and Angoff or Bookmark standard-setting each write an audit entry that cannot be quietly edited.
One honest limit: IRT parameters are imported, not estimated here. Lurno is a runtime engine, not a calibration studio.
Checks proportionate to the stakes
On where they matter, off where they do not. Practice does not need a lockdown.
Secure delivery
Full screen, clipboard blocked, tab switches recorded, everything shuffled per attempt. The answer key is stripped on the server, so it never reaches the browser.
Integrity events
Tab switches, copy attempts, a second concurrent session, suspicious timing — each recorded against the attempt, with time spent per item.
At-risk alerts
Flags a submission that came back too fast, several concurrent sessions, or low engagement. Each alert is acknowledged.
AI-writing check
Run a scan on a written submission and read it beside the work. It informs a person's decision; it does not make one.
Accommodations
Extra time as a percentage, more attempts, an exemption, high contrast, larger type, screen reader, reduced motion — per learner, applied automatically.
AI grading drafts
AI drafts rubric scores, feedback and a proposed grade over anonymised context. Nothing is applied until a person confirms it.
Before you put an exam on it
Bring an exam you already run.
A rubric you already use, a portfolio brief, a psychometric instrument — we will set it up on the call rather than show you a sample course.