Integrity & analytics

Integrity settings make an assessment harder to game while it’s being taken. Analytics tell you afterwards how it actually went — who took it, how they scored, and who needs help. The two live in different places: integrity is configured on each assessment, analytics is read on the assessment and organization-wide.

What this does

During an attempt Lurno can randomize what each learner sees, restrict how they move through the assessment, and record integrity events — switching tabs, losing window focus, attempting copy or paste, right-clicking. Those events feed an at-risk alert queue rather than a verdict.

Afterwards you get participation and score reporting at two levels: per assessment (funnel, headline numbers, score distribution, released attempts) and across the organization (story cards for what needs attention, plus a detailed data plane).

Before you start

You need permission to update the assessment to change shuffling; secure mode also needs the advanced-configuration permission.
Integrity settings apply to attempts started after you save. An attempt already in progress keeps the behaviour it began with.
Analytics only populate once attempts have been submitted.

Turn on integrity settings

Open the assessment and go to Settings.
In the Honesty card, turn on Shuffle question order so each learner sees questions in a random order, and Shuffle answer choices so choices within each question are randomized. Both save immediately.
Still in Honesty, set When learners see correct answers — Never, After they submit, or After everyone submits with a release datetime.
Expand Advanced and turn on Secure mode. One switch arms the whole lockdown: it monitors tab switches, blocks copy and paste, and prevents right-click during delivery. The same block holds a Confidence rating switch, which asks learners how sure they are of each answer. If you don’t see the Advanced block at all, you don’t hold the permission for it.

Picking Exam as the assessment type at creation already turns secure mode on, shuffles questions and choices, sets a 60-minute limit and one attempt, and holds results until you release them. Most people never need to touch these controls individually.

Per-section controls

Open a section’s Settings disclosure in the Content view to reach the controls that shape movement through the assessment:

Allow back navigation — whether learners can return to earlier questions in that section.
Pool: # to Select — draw a random subset of the section’s questions per attempt, so no two learners get the same set.
Time Limit (minutes) — a clock on that section, separate from the assessment-wide limit.

What gets recorded

When secure mode is armed, each attempt writes integrity events as they happen: Switched tabs, Window lost focus, Tried to copy, Tried to paste, and Right-clicked, each with a severity. Copy, paste, and the right-click menu are blocked as well as recorded. Lurno also tracks concurrent sessions, so the same attempt open in two places is visible.

Read the analytics

For one assessment

Open the assessment and use the Review stage of its rail:

Overview — the participation funnel, headline numbers, a score distribution chart, and the most recent submissions.
Results — a score-distribution histogram beside a roster of released attempts. Click a row to reopen that attempt read-only.
Analytics — reserved for deeper per-question reporting and not yet built; it shows a placeholder.

There is no per-question item analysis in the product today. For psychometric instruments, the Scoring & norms view carries the measurement model instead: scales and score bands are open to anyone who can edit the assessment, while norm groups, validity evidence, differential-item-functioning analyses, adverse-impact checks, and blueprints sit behind an Advanced disclosure that needs the advanced-configuration permission.

Across the organization

Go to Assessments → Analytics & Alerts. It has two views on a rail: Overview and At-risk learners. The pages inside still carry their older headings — Operational Dashboard and Intervention Queue — so don’t be thrown when the title doesn’t match the rail.

Overview opens with four story cards, each a plain question with one action:

Card	Answers
Who needs help right now?	At-risk learners flagged this week
What’s overdue to grade?	Submissions still waiting on a person
Where is integrity flagged?	Integrity events fired today
Which assessments aren’t pulling?	Published assessments to keep an eye on

Show detailed analytics expands the data plane underneath: summary stats, per-assessment participation, a learning curve, grading-quality stats, confidence calibration, outcome-alignment gaps against a competency framework, and a comparison of the current cohort against prior periods.

At-risk learners is the intervention queue. Each alert carries a reason and severity: Score Below Threshold, Attempt Abandoned, No Attempt Started, Multiple Integrity Flags, Speed Anomaly Detected, or Deadline Near — Not Started. Filter by assessment, switch between Unresolved and All, and act per row: Grant retake (for the reasons where a fresh attempt makes sense), Open the assessment, mark it Seen, or Resolve it. Opening an integrity-flag alert shows the learner’s recent integrity events with their severity.

Engagement-based at-risk alerts

Alongside the assessment-derived reasons above, Lurno now detects learners at risk from their engagement pattern, and fills the “Who needs help right now?” card that used to sit empty.

Three reasons, checked in descending order of confidence so an alert carries the strongest true one:

Reason	What it means
Failing while progressing	Working through the course and failing more than 45% of graded attempts, with at least a quarter of the course done. First, because it’s the case a progress bar scores as healthy — and the most actionable thing in the queue.
Inactive before a deadline	No activity for 14 days with a deadline within about 10 days, or already passed.
Fading	Activity has halved or worse across the observed weeks and is now genuinely low, with a deadline still within reach.

Each alert states a plain-language cause — “Working through the course but failing 62% of graded attempts.” — never a score or a risk rating.

This is arithmetic, not a model. Deliberately: a model deciding who is at risk would be unexplainable to the person acting on it, unauditable when someone asks why a learner was flagged, and would score the same learner differently twice. Every rule above can be read aloud to the learner it’s about.

What it deliberately does not flag

Precision is chosen over recall, because a queue with three bad alerts a week stops being read — including the alerts that were right. These learners are suppressed before any rule runs:

Anyone who has completed the course. Silence after finishing isn’t a signal.
Anyone on approved leave. Alerting here tells you something you authorized.
Anyone enrolled less than 7 days. “No progress yet” is a Tuesday, not a signal.
Burst learners — people who do everything on one day a week. A six-day gap is their normal, and a naive gap rule flags them every single week.
A drop from very high activity to merely high. Halving from 40 to 19 is not fading.

Learners are never shown any of this. At-risk is a staff-side signal; “you seem to be struggling” is a verdict about a person, and the learner-side equivalent is an offer about a question instead — see Help & wellbeing.

Granting a retake writes an override that lets that learner take the assessment one more time beyond the normal limit, and records your name and the reason in the audit history. See Audit log.

Reference

Setting / signal	Where	What it does
Shuffle question order	Settings → Honesty	Randomizes question order per attempt
Shuffle answer choices	Settings → Honesty	Randomizes choice order within each question
When learners see correct answers	Settings → Honesty	Never / after they submit / after everyone submits
Secure mode	Settings → Advanced	Monitors tab switches, blocks copy-paste, prevents right-click
Allow back navigation	Content → section settings	Whether learners can revisit earlier questions
Pool: # to Select	Content → section settings	Random subset of the section’s questions per attempt
Integrity events	Recorded during the attempt	Tab switch, window blur, copy, paste, right-click — each with a severity
At-risk alert	Analytics & Alerts → At-risk learners	A reason plus severity, with per-row actions
Engagement at-risk reasons	Analytics & Alerts → At-risk learners	Failing while progressing · Inactive before a deadline · Fading — all deterministic
Score distribution	Assessment → Overview and Results	Spread of scores across attempts

Troubleshooting

There’s no Advanced block in Settings. Secure mode is behind the advanced-configuration permission, which organization admins hold by default. Ask an admin. See Roles & permissions.

I can’t find a tab-switch toggle. There isn’t a separate one. Tab-switch monitoring, copy-paste blocking, and right-click prevention are all armed by the single Secure mode switch.

Shuffling didn’t change the order I saw. Settings apply to attempts started after you save. Use Preview as learner or start a fresh attempt; an attempt already running keeps the order it began with.

Analytics are empty. Statistics need submitted attempts. If nobody has finished, or every attempt is still in progress, there is nothing to compute.

The Analytics view on an assessment is blank. That view is a placeholder. Use Overview and Results in the same Review stage for participation and scores, and Analytics & Alerts for the organization-wide picture.

An integrity flag appeared but I can’t act on it. Flags are signals, not verdicts. Review them alongside the score and timing in the alert’s detail panel, and set your own policy for what a flag means. If a fresh attempt is the right answer, use Grant retake.