Release notes¶
Everything new, improved, and fixed in Janison Insights. Each entry below summarises the headlines — open the full release notes for the complete detail, including ticket references.
Test versions and results by year level and language¶
Each version of a test component can now be scoped to conditions such as year level and language, and reporting dimensions can produce scaled scores and performance bands for each of those cuts. Incident reports gain search and a test centre filter, and a long list of fixes steadies delivery and Quick Marking.
Author New
Test versions scoped by year level and language — a test component now declares the conditions its versions can use, drawn from the matching user attributes in Assessment Event Settings. Each row in the Assessment Versions grid picks its conditions from that pool, and a new Conditions column shows how many combinations the row covers. An Allowed Conditions summary above the grid lists what the versions must satisfy. Where the assessment uses a mixed delivery mode, each version can also be paired with a paper test in the new Paper Test column. Components created before this release load and save as before, with no conditions set.
Deliver Improvements
- Invigilation actions appear in Recent Activities as they happen on NoSQL delivery. Pausing or resuming a test-taker, adding or removing extra time, and postponing, abandoning or submitting an attempt now show on the test attempt modal within moments. Pausing or resuming a session shows in the session's Recent Activities too. Previously these entries appeared only once the session was finalised.
- Incident reports are easier to search and read. Search now finds an incident by its ID or the test-taker's username, as well as by text in the Details column. A multi-select Test Centre filter narrows the list to the centres you're monitoring, and works alongside a search term. Severity Rating, Outcome, Outcome rationale and Outcome category no longer appear in the table, but they're still in the Excel export.
- Signing in a second time before the session starts no longer locks a test-taker out where the event's settings allow re-entry without an unlock. That covers switching browser, using a private window or opening a second tab. The padlock on the session dashboard clears once the new sign-in takes over.
- The calculator's percentage now works against the preceding value. 100 + 10% returns 110 rather than 100.1, and 100 − 10% returns 90. On its own, 50% still converts to 0.5.
- Fewer test-takers are submitted automatically just after launching a test. If the server reports no time left in the first minute, the player works out the remaining time from the test duration and the time already used.
- Enrolling by group type creates one enrolment per test-taker, however many groups of that type they belong to. Previously a test-taker in several groups got an enrolment for each.
People Improvements
- Deleting a user now removes their sign-in account too. Previously the user left Insights but their sign-in account stayed in the connected directory. Users who sign in through your own identity provider aren't affected.
Report Improvements
- Reporting dimensions can produce separate data cuts. Scaled scores, standardisation and performance bands on an assessment program can now be calculated for a test-taker or user property such as year level, cohort or region. Previously they ran across the whole population for a dimension. A dimension's conditions are limited to those allowed by the test components of its linked events. The scaling table, standardisation and performance band templates can be downloaded and validated for a chosen set of conditions.
Platform Improvements
- Large events put less load on the database. Several frequently run queries can now be reused by the database rather than treated as new each time, removing short spikes in database CPU.
Deliver Bug fixes
- Fixed: an assessment event that used Quick Marking with Allow Group Type filtering couldn't be deleted
- Fixed: an attempt reconciled from a device with its clock set wrongly could show a completion date earlier than its start date. The server now corrects the timestamps when the device's clock is out by more than a minute.
- Fixed: a "get failed" error could appear when the test player looked up a saved answer for an item that hadn't been attempted. It was most likely when an attempt was replayed in the test player.
- Fixed: test-takers occasionally saw an "unexpected error" message while an answer was saving, even though the answer had saved
- Fixed: on tests combining reliable delivery with branching, ad hoc preparation could fail when a PIN code was applied, and moving to the next section sent the test-taker back to their last completed item
- Fixed: on some sites, test-takers in sessionless delivery could be left on a loading screen after entering their one-time code and confirming their name
- Fixed: an example test with Require access token ticked could still be opened from its link without a token
- Fixed: on NoSQL delivery, finishing an exit survey while the device was offline recorded an error
- Fixed: selecting a column header on the PNP details page didn't sort the records
- Fixed: a test-taker could try to submit while offline even with Enable Offline Submit turned off. The attempt status on the test session after the device reconnects is also corrected.
People Bug fixes
- Fixed: importing a user spreadsheet with two or more rows sharing a username showed a Microsoft Graph API error in the Import Log. The import now completes, keeping the username as written in the first row and taking every other detail from the last.
Mark Bug fixes
- Fixed: on the Quick Marking Reviewed tab, responses with a score of 0 couldn't be selected for bulk submission. The Pending Review tab is unchanged.
- Fixed: when two markers worked on the same response at once, the second saw an unclear error that came back on every try. Submitting, one response or many, now asks for confirmation first, then summarises which responses went through and which didn't.
- Fixed: reviewing, saving or submitting a response raised an error when the AI marker's feedback contained angle brackets
- Fixed: scanned multiple choice and keyword answers could be marked incorrectly when the options were printed in a different order from the authored one
- Fixed: composite items could be marked incorrectly when attempts were completed in bulk rather than submitted by the test-taker. Re-mark attempts affected before this release to correct their scores.
- Fixed: match, hotspot, gap match and other items scored by the QTI evaluator could be marked incorrect with no score recorded
- Fixed: attempts could stay on Marking after their auto-marked responses were submitted. They now move to Marked once both marked and auto-marked responses are submitted.
Platform Bug fixes
- Fixed: a user import on a site without the proctoring integration could log proctoring synchronisation errors in the Import Log
- Fixed: switching role or impersonating another user sometimes left you signed in under your original role. Recovery from a brief cache fault now takes 5 seconds rather than 60.
Older releases¶
API release notes¶
The APIs are versioned and released separately from the platform, so they have their own notes. Update your environments to the newest API version available to pick up the latest features.
The most recent is Events API v11 (February 2026), which exposes student break data to integrated systems — BreakTimeUsed and StartBreakTime on the /attempts endpoints and in the related customisable reports, and the break configuration properties on the /pnp endpoints.
See API framework for the Swagger references and how to authenticate.