Mock exams are a sellable product and a public test of a centre's operations. The short answer on what to prepare: structurally valid papers, audio tested on the actual equipment, one clock, and a committed result date.
Before the session
- Fix the version: Academic or General Training, stated on the candidate notice so nobody prepares against the wrong one.
- Verify structure: 40 Listening questions across four parts, 40 Reading questions, two Writing tasks, three Speaking parts. Every objective item must have an answer key in the system before the session starts.
- Test the audio in the actual room, on the actual speakers or headsets, at real volume, with someone sitting in the furthest corner.
- Prepare fallbacks: a second set of speakers, a spare machine already signed in.
- For on-screen delivery: check connectivity and run one real candidate login end to end.
Timing each skill
Hold to real timings or the session loses its measuring value:
- Listening: about 30 minutes of audio. On paper add 10 minutes to transfer answers; on screen there is no such window.
- Reading: 60 minutes including answer entry.
- Writing: 60 minutes for both tasks.
- Speaking: 11 to 14 minutes per candidate, and it may be scheduled separately.
The first three run without breaks, as in the real test. If your centre allows a break, say so when returning results so candidates read their stamina correctly.
During the session
- One timekeeper, one clock, spoken announcements at ten and five minutes remaining.
- Never replay the audio except for a room-wide technical failure — and when that happens, record it and mark that room's results as non-comparable.
- For on-screen delivery, decide in advance how a frozen machine is handled: move the candidate and add back exactly the minutes lost.
After the session
Slow results are the most common reason candidates do not buy a second mock. Objective sections mark instantly; Writing and Speaking are the bottleneck.
The way to shorten it without lowering standards is to let the system produce a first-pass, criterion-by-criterion report and have teachers review rather than mark from scratch. On LinguaGrade, Writing and Speaking responses pass a confidence gate before release: anything with a warning sign stays in the teacher's queue, and only work clearing every condition is released automatically. Teacher time then concentrates on the few responses that genuinely need a human. See How far you can trust AI marking.
Results should include per-skill bands, the overall computed with the official rounding rule, and an error breakdown by question type — the part that tells candidates what to study next.
Frequently asked questions
How often should we run mocks?
Far enough apart for candidates to fix what the last one revealed — typically four to six weeks.
Can we use past official papers?
Do not use material you are not licensed to use. Use licensed papers, centre-authored papers, or papers assembled from your own lawful component bank.
Should mock marking be stricter than the real test?
No. Marking off-standard destroys trust in the number, in either direction.