Concepts
How automatic generation actually works
Without the marketing. What the engine is doing, why the problem is hard, and why the quality of your constraints matters more than the cleverness of the algorithm.
The problem, stated plainly
You have a set of teaching sessions. Each needs a time slot, a room and a lecturer. Some resources are shared, some sessions must not overlap, and a long list of rules constrains which combinations are legal. Find an assignment that breaks no hard rule and violates as few preferences as possible.
This is a constraint satisfaction problem, and university timetabling is a well-studied hard case of it. The number of possible assignments grows explosively with the number of sessions: exhaustive search is not merely slow, it is astronomically infeasible for even a modest faculty.
So no practical system searches exhaustively. They use heuristics — informed strategies for deciding what to place next and what to do when stuck — to find a good solution quickly rather than a provably optimal one eventually.
What the engine is actually doing
- Ordering. Deciding which sessions to place first. Highly constrained sessions — a large cohort needing a specific laboratory with one qualified technician — are usually placed early, because placing them last means unpicking everything else.
- Placing. Assigning a session to a legal slot, room and lecturer combination that violates no hard constraint.
- Backtracking. Undoing earlier decisions when a session turns out to have no legal placement left. This is where most of the running time goes.
- Scoring. Evaluating how well the current partial solution satisfies soft preferences, so the search can prefer better regions of the space.
- Improving. Once a complete legal solution exists, making local changes that improve the score without breaking anything.
Why your constraints matter more than the algorithm
This is the part vendors are least keen to say. Given a well-specified constraint set, most competent engines produce comparable results. Given a badly specified one, no engine produces anything useful.
The most common failure is not algorithmic. It is a constraint set where preferences have been declared inviolable, so no legal solution exists and the engine correctly reports a large number of unplaceable sessions. The second most common is stale or wrong input data — room capacities nobody has checked since a refurbishment, availability collected six weeks ago.
Which is why the useful question to ask a vendor is not "what algorithm do you use" but "how do you help me get the constraints and the availability right".
What good looks like in practice
A usable system generates fast enough that regeneration is routine rather than an overnight job — Wise Timetable regenerates even large timetables in under a minute, which means you experiment rather than committing to one run.
It reports what it could not place, and why, rather than silently breaking a rule to produce something that looks finished. It respects sessions you have locked, so automatic and manual work can be mixed. And it produces multiple variants so you can compare rather than accept.
Want to see this working?
Book a 45-minute online presentation and we will walk through it against your institution's own scheduling problem.