Ask a staffroom what makes a good timetable and you will get two answers, delivered with equal conviction and usually by different people.
Teachers want compact days. Arriving for period 1, teaching four lessons, leaving after period 5 is a good day. Arriving for period 1, free until period 4, teaching two, free again, teaching period 8 is a day that consumes eleven hours and pays for six.
Students — and the people responsible for them — want continuity. A free period in the middle of a fourteen-year-old's day is not a break; it is a supervision problem and, for a student who travels, an hour in a corridor.
Both are reasonable. The uncomfortable arithmetic is that they compete for the same resource: where in the week a lesson sits. Compacting one person's day pushes lessons into slots that stretch someone else's. The scarcer your rooms and the tighter your teacher availability, the sharper the competition.
Why this is not a software problem
It is tempting to expect an optimiser to eliminate the trade-off. It cannot, because the trade-off is not an inefficiency — it is a property of the problem. In most real schools there is no arrangement that is simultaneously best for teachers and best for students. There is a frontier of arrangements where improving one necessarily costs the other, and the school has to choose a point on it.
What software can do is make the choice explicit, consistent and reversible — which is what a human planner, working card by card at midnight, cannot.
How weights actually work
In Bildena — and in the tradition of timetabling generally — hard rules and preferences are different species. Hard rules are inviolable: a teacher cannot be in two rooms, a lab lesson needs a lab. Preferences are weighted, and the solver maximises their weighted sum.
That last phrase is where most misunderstanding lives, so it is worth being precise. A weight is not a priority queue. The solver is not told "fix teacher gaps first, then student gaps." It is told how much a teacher gap costs relative to a student gap, and it finds the arrangement with the lowest total cost across the whole school at once.
The consequence is often counter-intuitive and always correct: the solver will accept a worse outcome in one place to gain more elsewhere. If tolerating one extra teacher gap removes four student gaps, and your weights say a student gap costs more, it will do that trade every time — including in the corner of the timetable nobody was looking at.
Three profiles, and what they actually mean
| Profile | Favours | Fits |
|---|---|---|
| Balanced | Neither strongly; spreads the pain | Most schools, first run |
| Compact days | Student continuity; fewer free periods in the middle | Younger year groups, long commutes, supervision duties |
| Teacher-friendly | Teacher gaps and split days | Many part-time staff, recruitment pressure, teachers travelling between sites |
These are starting points, not answers. The right calibration for your school emerges from comparing two or three runs on your own data — which is fast when a run takes seconds and impossible when it takes a weekend.
Read the distribution, not the average
This is the most useful practical advice in this article.
Every timetabling tool reports aggregate quality — an average, a score, a percentage. Averages hide the thing that generates complaints. A plan with an excellent mean can contain one teacher with three split days and one student with four free periods on a Tuesday, and those two people will define how the whole school talks about the timetable in September.
So look at the tail:
- The worst-off teacher. Not the average gap count — the maximum. Who has the most fragmented week, and is that person part-time, commuting, or already carrying something else?
- The worst-off student group. Same question. A single class with a bad Wednesday is a manageable problem if you know about it in July.
- The spread. Two plans with identical averages, one where everyone is slightly inconvenienced and one where three people carry all of it, are very different plans. The second usually feels unfair, because it is.
A solver optimises what you measure. If you only ever measure the mean, you will get plans that are excellent on average and locally miserable.
A method that works
The schools that get this right seem to follow roughly the same path.
Decide the ordering before you run anything. Not the numbers — the ranking. "Student continuity matters more than teacher compactness this year, because we have a new intake with long commutes." One sentence, agreed by people who will have to defend it.
Generate two or three plans with different weightings. Not variations of one plan — the same data under genuinely different priorities.
Compare the tails, not the scores. Worst-off teacher, worst-off class, spread.
Then decide, and write down why. Next August, the person calibrating the weights may not be you, and "we chose this because of the new intake" is the most valuable thing you can leave them.
Bildena lets you keep multiple distributions side by side — same data, different weights — so the comparison above takes an afternoon rather than a summer.