Scheduling to a division
Pick the date, the window and the duration, publish to 9-A, and every enrolled student is notified. A clash with another exam on that division is refused, not warned about.
Publish a paper to a division, set the window, and every student is notified. The exam runs timed and proctored in a browser that already exists on their phone — and every answer is saved the moment it is typed, not when they press submit.
Setting a paper live is the easy part. Everything below exists because forty students on forty different devices and connections will find every weakness you left in.
Pick the date, the window and the duration, publish to 9-A, and every enrolled student is notified. A clash with another exam on that division is refused, not warned about.
The clock starts when a student opens the paper, not when the window does. A late arrival still gets their full duration, capped at the window close.
Every answer is written locally as it is typed and synced when there is a connection. Closing the tab, a flat battery or a dead signal costs a student nothing.
Tab-switch counts, paste detection and a per-student question order. Signals a teacher can read, not a webcam feed nobody has time to watch.
No app to install and no minimum device. A full three-hour paper is 318 KB, so it opens on a shared handset over a patchy connection in a village.
If something genuinely goes wrong, an invigilator can grant a student more time or reopen their paper. Every such action is logged against the exam.
A navigator showing what is answered and what is flagged, the time remaining, and an autosave indicator that is always telling the truth. This is the whole interface.
Explain why a gas fills its container completely while a liquid does not, with reference to intermolecular forces.
One student's three hours, as the teacher sees it. Every signal is a moment on a timeline rather than a number in a table — because when something happened is usually the whole explanation.
Five signals, and five things we will not collect at any price. Nothing here penalises a student automatically — signals sit beside their script when a teacher opens it. That second list is not a plan tier and not a setting: there is no version of this where a camera turns on in a child's bedroom. Schools needing invigilated conditions still put students in a room, which works and always has.
What the teacher watches for three hours. Every tile is a student, its colour is their state, and the two who never signed in are visible from the first minute rather than discovered at the end.
Roll 24 resumed on a second device at 11:47 after two minutes offline. Answers synced intact. Most likely a flat battery — grant time, or leave it and let the log stand.
Grant 10 min Reopen paper Message DismissEvery action here is logged against the exam with the invigilator's name and a timestamp, and appears on the result sheet. Granting time is meant to be easy and never invisible.
An exam that punishes a bad signal is not testing the syllabus. Answers save on the device and sync when the line returns, so a dropped connection costs a student the seconds it lasted and nothing else.
Run automatically the moment a teacher hits publish. Four of these will stop an exam outright. The other two are judgement calls, so they say so and let a person decide.
No draft or unreviewed question can reach a student. The publish button is disabled until the last one is signed off.
The approval is per question and per teacher, with a name and timestamp against each. A head of department can see who signed off what, which matters the one time a question turns out to be wrong.
Every objective question has a key and every written one has a rubric, or the paper cannot be marked and will not publish.
For written questions the rubric is the key: the expected points and the marks each carries. Without it the AI examiner has nothing to score against, so the paper is blocked rather than marked badly.
Section totals add to the paper total. A 78-mark paper labelled 80 is caught here rather than by a parent on results day.
Checked at section level and paper level. The most common cause is a question edited after the blueprint was set — the check catches it in the second it happens rather than at results.
The division is not already sitting something else, and the window does not run past the school day it was scheduled in.
Also checks the division is not already sitting a paper in an overlapping window, and that a three-hour exam scheduled at 15:00 does not run past the end of the school day.
Students on the division roll who have never signed in are listed by name, because they will not be sitting anything on Friday.
Listed by name, not as a count, because a count is easy to dismiss. Warns rather than blocks: sometimes a student genuinely has left, and the exam should still run for the other forty-one.
Three marks a minute is flagged as probably wrong. Sometimes it is deliberate, which is why this warns instead of blocking.
Compares total marks against the window and flags anything above roughly three marks a minute. Speed tests are a legitimate thing to set, so this never blocks — it only makes you look once.
The honest version of “it worked”. Three students needed intervention and two never turned up — but no paper was lost, and that is the only number in this section we would put in a contract.
Five of forty-two needed something to go right that wasn't automatic. Every one of those five still submitted a complete paper. We report this shape rather than a nine-nines uptime figure, because a school does not care whether the platform was up — it cares whether Ravi's paper survived.
The four that come up almost every time.
Nothing they will notice beyond a banner. Answers are written to the device as they are typed, so the paper keeps working with no network at all — the student can carry on answering. When the connection returns, everything syncs in the background and the banner clears. The clock keeps running, because pausing it on a dropped signal would be trivially easy to abuse. If a student genuinely lost significant time, an invigilator can grant more, and that action is logged against the exam.
We do not claim to stop it — we make it visible and expensive. Question order is shuffled per student, so a screenshot passed around is worth little. Tab switches, paste events and unusual answer timing are counted and shown to the teacher next to that student's script. What we will not do is stream video of children to a server, and schools that need invigilated conditions still put students in a room. Proctoring signals inform a teacher's judgement rather than replacing it.
Yes. There is no app, no minimum Android version worth naming, and a full three-hour paper transfers about 318 KB. It runs in whatever browser is already on the handset. On a shared device, a student signs in, sits their paper and signs out, and their answers are tied to their account rather than the phone — so two siblings can use the same handset for two different exams.
Yes, and plenty of schools do for board preparation specifically. Generate the paper, export it as a print-ready PDF, and conduct it in a hall as normal. Objective answers can then be entered from an answer sheet in a few minutes per division to get the chapter analytics, or you can skip that and treat it as a purely offline exam. The platform does not insist on being in the room.
Book a free 30-minute demo. We'll put you in a live exam window and pull the plug on your connection halfway through.