Proof and checking

A timetable,and the reasonsfor it.

Most tools hand you a timetable. Nexar Schedule will hand you the timetable, a proven floor under its number of days, and a verdict on every candidate from one checker.

The sample on this site, as judged

Valid. All 18 exams placed in 6 days. No timetable can use fewer: these 6 exams all share candidates.

Sample data: 18 exams, 96 candidates, 4 rooms. Invented for this site and worked out in your browser. This is not the product.

Two things you can check by eye.

The proof is a small table. The check is one candidate’s week. Both are worked out here on the sample, not typed in.

Proof of the fewest days, with the exams that set it.

In the sample, these 6 exams each share candidates with every other one. With one exam a day, no two of them can sit on the same day, so no timetable can use fewer than 6 days. That is the whole proof, and you can check it by eye.

The floor is a lower bound. If the best timetable found uses more days than the floor, you will be shown both numbers, never one dressed up as the other.

StrategyStats 1Econ 1Accts 1Econ 2MetricsBusiness StrategyStrategy30303044Statistics 1Stats 130303033Economics 1Econ 130303033Accounting 1Accts 130303044Economics 2Econ 2433439EconometricsMetrics433439

Sample data. Each number is how many candidates sit both exams.

Candidate 20402

25% extra time, separate room. Sample data.

Economics 2Mon 10 May, 09:30Room C3, 225 min
Accounting 2Wed 12 May, 09:30Room C3, 225 min
EconometricsThu 13 May, 09:30Room C3, 225 min
MicroeconomicsMon 17 May, 14:00Room C3, 225 min

25% extra time on a 3-hour paper: 225 minutes, separate room, same start.

Checked for this candidate: no conflict, one exam a day, no two exams inside 24 hours.

Every verdict checked candidate by candidate.

A timetable can have no conflicts and still be hard on the people sitting it. So the checker reads it the way a candidate does: each exam, each day, each room, each approved accommodation or access arrangement.

A missed arrangement is a broken rule, the same as a conflict.

One judge, one answer.

The solver proposes. The checker decides. The same checker judges everything below, so two screens can never disagree about whether a timetable is valid.

Drawing: a timetable sheet with one column marked off by a ruled line, a ruler lying across its corner and a rubber stamp with a tick resting on it.
A limit ruled on the sheet, and a check that signs it off.
The first timetable
Every exam, every candidate and every room, against every rule you set.
Every change by hand
Each slot answers before you put the exam down: clear, or the rule it breaks and for how many candidates.
The timetable you published
Checked the same way in a shadow run, with the candidates behind each finding listed by number.
The file you export
Checked once more as it leaves, so what you publish is what was judged.
When the exams cannot fit
Not “no solution found”. The exams and candidates that make it impossible are named, so you know which rule to discuss.
What sets a verdict
Plain code. AI reads your files and understands your requests, and you confirm what it read. No model decides whether a timetable is valid.

Questions about the proof.

All questions

What does “proof of the fewest days” mean?

If six exams each share at least one candidate with every other, and your rule is one exam a day, no timetable can use fewer than six days. Nexar Schedule will find that set and show it to you, with the candidates. It is a floor: if the best timetable found needs more days than the floor, you will see both numbers.

Will it always find the shortest timetable?

We do not promise that. It will prove a floor and show you the timetable it found against it. Where the two meet, as they do in the sample on this site, that timetable is the shortest possible. Where they do not, you will see the gap.

Is AI deciding our timetable?

No. AI reads your files and understands what you ask, in plain words or by voice, and AI models from different families cross-check what was read. The timetable is built by an optimiser, and whether it is valid is decided by one independent checker written in plain code. The AI never invents a timetable or a number.

Which rules will the checker know in the first pilot?

No candidate in two exams at once, a limit on exams per day, room capacity, extra time, and separate rooms. It will also count candidates with two exams inside 24 hours, so you can see them. Other rules come after the first pilot.

Start with last season.
It costs nothing.

Send the files from last season and the timetable you published. We check it candidate by candidate, rework it on the same days and rooms, and send one page back.

Free. No purchase, no switch, no installation. Being built now by Nexar Labs.

What a shadow run needs from you

  • Examscode, title, length
  • Enrolmentscandidate number, exam code
  • Roomsname, seats
  • The timetable you publishedexam, date, time, room

As your system exports them. Candidate numbers, not names. The adjustment, never the reason for it.