How do you build a multi-track agenda that does not collide?
By Donghui Qian (Milo)Updated 19 September 20264 min read
Short answer
Check for three kinds of collision, not one: rooms; people, including chairs, hosts and moderators; and audience, meaning two sessions for the same attendees in the same slot. Size rooms by ranking expected demand against seated capacity, leave the changeover time you measured by walking it, and record attendance per session to build next year's grid.
A single-track day is a list. The moment two things happen at once, the programme becomes a scheduling problem with three different ways to fail — and only one of them is visible in the spreadsheet.
This is a field guide to building and checking the grid. It applies to any agenda tool, ours included.
Three collisions, only one of them obvious
| Collision | What it means | Found by |
|---|---|---|
| Room | Two sessions in the same room at overlapping times | Any tool. This is the one everybody checks |
| Person | A speaker, chair or host is in two places at once — including the moderator nobody listed as a speaker | Checking people rather than rooms, which most grids do not do |
| Audience | Two sessions the same people came for, opposite each other | Nothing automatic. It is a judgement about who your attendees are |
The third is the expensive one. A room clash is embarrassing and fixable on the day; an audience clash quietly halves the room for both sessions and shows up only in the attendance report afterwards.
Build the constraints before the grid
Most collisions are created in the hour when sessions are dropped onto slots. Write these down first and the grid mostly builds itself:
- Room capacity, per room. Not the venue's headline number — the seated number with the layout you are actually using.
- Who is unavailable when. Speaker travel, the chair who is also on a panel, the sponsor who can only be there in the morning.
- Fixed points. Keynotes, lunch, the sponsored slot in the contract, the room you only have until three.
- Walking time between rooms. Measure it once, including stairs and the lift queue. If sessions run in two buildings, measure that too.
- Which sessions serve the same audience. Tag them. This is the input that prevents the collision nothing else catches.
Sizing rooms against demand
The common failure is a popular session in a small room and a specialist session in the ballroom. You cannot know demand exactly, but you can rank it, and ranking is enough.
expected demand ≈ registrations × share of attendees this topic serves
× share who will be on site at that hour
room fit = expected demand ÷ seated capacity of the assigned room
Anything over 1 needs a bigger room, a repeat, or an overflow plan. Fill the shares from what you know about your own audience — a registration question that asks which topics someone came for turns this from a guess into a number.
If the tool lets attendees pre-select sessions, use it as a demand signal rather than as a commitment. The ranking is reliable; the absolute numbers are not.
Sequencing rules that hold up
- Never put your two strongest sessions opposite each other. You are not choosing between them; your attendees are, and half of them will resent it.
- Spread each track's highlights. If everything good in a track happens before lunch, the afternoon room is empty and the speakers notice.
- Protect the first slot after lunch. Put something participatory there, not the densest technical talk of the day.
- Leave the changeover you measured, not the one that fits the grid. Sessions that end late cascade through every following slot.
- Put repeats far apart. If you repeat a popular session, the second run belongs on the other day or the other end of the day.
The agenda will change after you publish
Speakers cancel, rooms move, a session splits. What separates a calm change from a messy one is deciding in advance how changes travel.
- One source
- The published agenda and the on-site signage must read from the same record. Two sources diverge within a day.
- Who is told, and when
- A room change the night before is an email; a cancelled session is an email and a sign on the door. Decide the threshold before you need it.
- What the attendee keeps
- If people saved a personal schedule, a moved session must move in it too, or the personal schedule becomes the thing that is wrong.
Check the grid before anyone else does
Four passes, in this order. Each one catches things the previous one cannot.
- By room — overlaps and changeover gaps.
- By person — every speaker, chair, host and interpreter, not just the headline names.
- By audience tag — any two sessions with the same tag in the same slot.
- By walking — trace one attendee's plausible day end to end, including getting lunch and reaching the far room.
Afterwards, measure what actually happened per session rather than per day. Attendance against room capacity is the only input that makes next year's grid better, and it has to be captured while people walk in.
Frequently asked questions
Check three things rather than one. Rooms, which any tool will flag; people, including chairs, hosts and moderators who are often not listed as speakers; and audience, meaning two sessions that serve the same attendees scheduled against each other. The third has no automatic check — tag sessions by the audience they serve and review the grid by tag.
Measure it rather than adopt a number. Walk the longest route an attendee could need between two consecutive rooms, including stairs, lifts and the door of a room that empties slowly, and use that. Venues differ enough that a figure from someone else's conference is not evidence about yours.
Rank rather than predict. Estimate demand as registrations times the share of attendees the topic serves times the share on site at that hour, then divide by the seated capacity of the assigned room; anything over 1 needs a bigger room, a repeat or an overflow plan. If attendees can pre-select sessions, treat that as a ranking signal rather than a headcount.
Attendance per session against the capacity of the room it was in, plus which sessions the same people attended together. The first tells you where the grid was wrong; the second tells you which sessions must never be scheduled opposite each other next year.