Post Proctoring

Post Proctoring is the bulk review surface: instead of opening candidates' sessions one at a time, you load a whole batch and work through their proctoring together — watch everyone's footage side by side, set verdicts, edit detections, and finalize — all from one screen. A progress counter in the header tracks how many of the loaded candidates are finalized.

Choosing the candidates

Start Post Proctoring in either of two ways: select sessions in the session list first and carry them in, or use the built-in audience picker to fetch a batch — the whole exam, or a slice of it. You can change the audience at any time without losing your place elsewhere.

The candidate worklist

A sidebar lists every loaded candidate with live verdict and review-stage chips that stay in sync no matter where you make an edit. A stage filter lets you triage — for example, show only candidates not yet finalized. Checkboxes select candidates for two bulk actions:

• Bulk verdict — set the manual flag (red / yellow / green) for all selected candidates in one action.
• Bulk finalize — lock the proctoring stage for all selected candidates. Only candidates who already have a red, yellow, or green verdict can be finalized; the rest are skipped and reported in the summary.

Playback view — watch everyone at once

Playback turns the batch into a synchronized screening: one card per candidate, all sharing a single playhead. When you press play, every candidate's session advances together — each card steps through its own captured material in lockstep, so sessions of different lengths still start and move as one. Adjust the playback speed, filter to webcam or screen material, and page through large batches.

Playback is deliberately read-only — it is built for spotting, not editing. When something catches your eye, click the candidate to open their review.

Wall view — review and edit in place

The wall shows one card per candidate with their captured media and detection markers. This is where the editing happens, without opening any session:

• Quick verdict — set a candidate's manual flag directly on their card; it saves immediately.
• Approve / decline findings — confirm real detections (raises their weight in the risk score) or decline false positives (removes their points permanently; declined findings never return on future AI re-runs).
• Add manual detections — flag anything the AI missed, especially on screen recordings, which are not AI-analyzed.

Detection edits across candidates are batched: Apply persists everything pending in one pass (with a progress indicator), Discard reverts it all. If you try to leave with unapplied edits, an exit guard warns you first — nothing is lost silently.

Detail mode — the full review, without leaving

Clicking a candidate anywhere — worklist, playback, or wall — opens their complete single-session review embedded in place: the same surface described on the Session Proctoring page, with the verdict selector, AI analysis dialog, all review views, and finalization. Previous / next arrows walk the roster candidate by candidate, and a back button returns to the view you came from. If the candidate has unapplied wall edits, you are prompted to apply them first so the detail view never diverges from your work.

A practical rhythm for a large exam: load the batch, filter the worklist to non-finalized candidates, run Playback once at high speed to spot visual anomalies, work the wall — decline noise, approve real findings, set quick verdicts — then bulk-finalize the clean majority and open detail mode only for the candidates that deserve a closer look.

Suggested workflow