Publishing an exam in StudyDrome

Publish once, and the paper stops moving

StudyDrome Exam Manager publishes an exam as an immutable snapshot. One transaction copies the whole delivery graph into 40 published record types, from items and answers to the resolved roster. Six settings stay editable on the live exam. Candidates already writing keep the version they started on.

Book a pilot

What happens when you publish an exam?

  1. 1
    Publishing runs seventeen numbered steps inside a single transaction. It starts by loading the test read-only.
  2. 2
    It validates the configuration and warms the question-bank cache, so a large paper does not pay for the same bank twice.
  3. 3
    Any publication that was already Active is superseded and pointed at the new one.
  4. 4
    The delivery graph is then copied into published tables: items, banks, questions, answers, vignettes, roster, grading scales, OSCE stations, images and access codes.
  5. 5
    The roster is resolved here. Groups, course clusters, individuals and exclusions collapse into one final list of who may sit the paper.
  6. 6
    The transaction commits. The database snapshot is durable before the response returns, and file copies finish in the background.

What does the snapshot contain?

What is frozen

What that covers

The paper

Test items, question banks, questions, answers and shared case vignettes

The roster

Groups, individual students, exclusions, course groups and clusters, resolved to one final access list

The marking rules

The grading scale and every level, with translations

OSCE material

Stations, station questions, examiners and the global rating scale

Media and fields

Intro and completion images, custom field values

Entry

Access codes, generated fresh or carried over

Forty published record types in all. The snapshot is the reason a score survives a challenge a year later: the paper a candidate saw is still readable, exactly as it was served.

Which settings can you change on a live exam?

Six settings live on the test itself and are deliberately kept out of the snapshot. Change any of them and the change takes effect at once, with no republish:

  • Grader anonymity: whether graders see candidate names while they mark.
  • Results detail: how much a candidate sees on their own results page.
  • Teacher attempt visibility: which attempts teaching staff can open.
  • Teacher grading permission: whether teaching staff may grade at all.
  • Grading deadline type: which date starts the grading clock.
  • Grading day limit: how many days that grading window runs.

Where do you change them on screen?

From the Grading & Results drawer, for one exam or for many at once. The drawer states the rule on itself: toggle on the fields you want to change, and untouched fields keep their existing values.

  • Everything else — items, roster, randomisation, time limit, integrity controls, access control — belongs to the snapshot and moves only by republishing.
  • That split is the point. The things an exam office adjusts mid-cycle are live. The things a candidate could be graded against are frozen.

What does StudyDrome check before it publishes?

A validation endpoint runs the same checks the publish path runs, so an author sees blockers before committing. The wizard shows them under Configuration Summary.

  • Bank shares: when questions are drawn by percentage, the shares across banks must total 100%. Publishing is blocked until they do.
  • Global rating scale: an OSCE scale must still exist, and it must carry exactly one borderline grade.
  • Paper size: a written exam must ask for more than zero questions. An OSCE derives its count from its stations.
  • All or nothing: validation runs again inside the transaction. A failure returns structured errors rather than a half-published exam.

How do you fix a flawed question after delivery?

Two operations change scoring after an exam is over. Neither edits the snapshot.

Adjusting the points on one question. A reviewer re-weights a single published question. The original point value is never touched. The new value is stored beside it, with the time and the person who set it. Every attempt on that publication is then recalculated. Clear the adjustment and the original value applies again.

Voiding a question. A flawed item can be excluded from scoring across every attempt with one toggle. The toggle is idempotent, and the question's point values survive it. Restoring the question returns the exact prior score. The item stays in the snapshot; only its contribution to the maths changes.

Both actions are permissioned, and both leave the delivered paper intact. That matters at appeal: you can show what the candidate answered and what it was later worth, side by side.

How do exam versions behave over time?

  • Only one publication of a test is Active at a time. Each one carries its own time-ordered identifier.
  • Publishing again supersedes the previous publication and links it to its replacement. A publication is Active, Superseded or Archived.
  • Candidates who already started keep working against the version they began. New attempts bind to the Active version.
  • Attempt caps are counted across every publication of the test, so republishing never hands a candidate a fresh allowance.
  • Access is withdrawn by archiving a publication, which keeps the record and the results intact.
  • Publication History lists every version with its status, who published it, how many students it resolved to, and when it was published and superseded.
  • Republishing is the same operation as publishing. There is one code path, so a corrected paper is snapshotted exactly like the first one.

Who needs this most?

Publishing is plumbing until someone challenges a result. Then it is the whole case.

  • Exam offices running the same paper across several sittings, who need each sitting to stand on its own record.
  • Assessment leads who publish a correction mid-cycle and must show which candidates saw which version.
  • Quality and appeals panels, who ask what the candidate actually saw rather than what the bank holds today.
  • IT and information-governance reviewers, who want the durability boundary stated plainly before they sign off.

Frequently asked questions

What exactly does publishing freeze?

The whole delivery graph, copied into 40 published record types. That covers test items, question banks, questions, answers, shared case vignettes, the resolved student roster, grading scales and their translations, OSCE stations and examiners, images, custom field values and access codes. The snapshot is what a candidate sat, kept readable long after the bank has moved on.

Can we change anything on a live exam without republishing?

Yes, six settings. Grader anonymity, how much detail candidates see on their results, which attempts teaching staff can open, whether teaching staff may grade, which date starts the grading clock, and how many days it runs. You change them in the Grading & Results drawer and they take effect immediately.

What happens to candidates who are mid-attempt when we republish?

They finish on the version they started. Publishing supersedes the old publication rather than deleting it, so an in-flight attempt keeps its own paper, its own roster and its own marking rules. New attempts bind to the Active version. Attempt caps still count across every publication of the test.

How do we handle a flawed question found after the exam?

Two ways, both without touching the snapshot. Re-weight the question, and its original points stay recorded beside the adjusted value with the time and the person who set it. Or void it, which excludes it from scoring on every attempt. Both recalculate affected attempts, and both are reversible.

How do we withdraw access to a published exam?

Archive the publication. Archiving keeps the snapshot, the attempts and the results exactly as they are, so the record survives the withdrawal. That is deliberate: an exam you can no longer sit is still an exam you may have to account for at appeal or audit.

Does republishing create a second set of results?

No. Results belong to the publication a candidate sat, and each publication keeps its own. Publication History shows every version side by side with its student count and its dates. Reporting can therefore separate the cohort that sat the original paper from the cohort that sat the correction.

Book a pilot

Bring one paper. We will publish it, republish a correction, and show you both versions side by side in Publication History.

Book a pilot

Related