Skip to content

Waitlist backfill for meetings turns a cancellation into a second chance

MatchmakingUpdated 2026-08-188 min read

In short

Waitlist backfill for meetings is the practice of holding a ranked standby list of near-miss proposals for each time slot and offering a freed table to the top candidate whose participants are both available. Forty cancellations at a 55 per cent backfill rate recover 22 meetings that would otherwise be empty tables.

Two hundred and thirty meetings were placed. Sixty-nine accepted proposals never got a table because the hall ran out of capacity. Then 40 cancellations arrive across the two days, and 40 tables sit empty at times when 69 pairs of people would both have liked to be at them.

Waitlist backfill for meetings closes that gap, and it is the cheapest capacity you will ever add. A new table costs floor space you could have sold. A standby queue costs an index, a ranking rule and a notification, and it runs on demand you have already generated and already paid for.

The queue is per slot, not per programme

The single design decision that determines whether this works is what the queue is keyed on.

A single global standby list sorted by score looks tidy and fails in practice, because when slot 9 at table 22 frees up, the top of a global list is a pair who are both busy in slot 9. You then walk down the list checking availability, and by the time you find a candidate the slot has started.

Key the queue on the slot. For each of your 24 slots, hold the ranked list of unplaced proposals where both participants are free in that slot. A proposal appears in several queues, once for every slot both parties could attend, and gets removed from all of them the moment it is placed anywhere.

This costs more memory and no meaningful compute. With 69 unplaced proposals and an average of 6 mutually free slots per pair, you are holding about 414 queue entries. That is nothing, and the lookup on a cancellation becomes a single read of the head of one list.

Two refinements earn their keep. Hold the exhibitor's identity in the key as well, because a freed table at 14:20 usually frees a specific exhibitor rather than a generic one, and the candidates who want that exhibitor are a much shorter list than the candidates who want any meeting at 14:20. If the exhibitor cancelled rather than the buyer, the freed resource is a table and a buyer, so the queue you read is the one keyed on that buyer instead. Building both directions is the difference between recovering half your cancellations and recovering a quarter of them.

Who belongs on the standby list?

Only proposals where both parties already said yes.

The temptation is to fill the queue from the full proposal set, including pairs who never responded and pairs where one side declined. Both are mistakes with different costs. Non-responders convert poorly, so your fill rate falls and the measurement gets muddy. Declined pairs are worse: offering a buyer a meeting with an exhibitor they already turned down reads as the system ignoring them, and it is the single fastest way to lose trust in the matching itself.

So the queue is the accepted-but-unplaced set, ranked by match score, and if that set is small then your capacity problem is smaller than you thought and your acceptance problem is larger.

There is a useful second tier, held separately and used only when the first tier is exhausted: proposals from buyers who are currently under-served, meaning fewer meetings than their quota, regardless of score. Working the first tier maximises match quality. Working the second tier fixes the buyer who is going home with two meetings. Decide which one you are optimising before the show, because during it somebody will decide for you.

The arithmetic on 40 cancellations

Take 40 cancellations across a programme, split by notice.

Twenty-four arrive more than 24 hours ahead. Both parties can be contacted, confirmed and given time to prepare, so the fill rate here is high. At 75 per cent, 18 are filled.

Ten arrive between 24 hours and 2 hours ahead. The replacement buyer is probably already on site or travelling. At 40 per cent, 4 are filled.

Six arrive inside 2 hours. Filling these means finding somebody in the building who can walk to table 22 in the next few minutes. At 0 per cent through a queue, though a floor team with a tablet sometimes catches one, call it 0 filled.

Total filled is 18 plus 4, which is 22 out of 40. The blended backfill rate is 55 per cent, and that blended number is the one everybody quotes and the least useful of the four figures above. The three component rates tell you where to work. The 24 hour bucket is already performing; the middle bucket at 40 per cent is where an hour of engineering pays.

Put a value on the 22. On a programme where the hosted cost per meeting works out at 116.67, recovering 22 meetings is 2,567 of hosting spend that produced something instead of nothing. If your exhibitors pay for the meeting programme separately, it is also 22 meetings you promised and delivered.

Why does the notice split matter so much?

Because backfill and cancellation policy are the same lever pulled from two ends, and confusing them wastes effort.

A cancellation that arrives with 3 days of notice is nearly free, and the fill rate proves it. A cancellation at 08:50 is nearly unrecoverable no matter how good your queue is. So the value of improving your queue is bounded by the notice distribution you receive, and the value of improving the notice distribution is bounded by how well your queue converts.

Work the numbers in that order. If 60 per cent of your cancellations already arrive with more than a day's notice and you fill 75 per cent of them, the ceiling on queue improvements is small and the work belongs in the policy that moves cancellations earlier. If your notice distribution is good and your fill rate is 30 per cent, the queue is the problem.

Revenue management has the same structure and a longer history. Talluri and van Ryzin give overbooking a full chapter in their 2004 book The Theory and Practice of Revenue Management, published by Springer, working through static and dynamic models of how much to sell above capacity when some share of bookings will cancel or fail to appear. LaGanga and Lawrence made the clinic version concrete in Decision Sciences in 2007, showing that overbooking improves patient access and provider productivity while increasing both waiting time and overtime, so whether it is worth doing depends on the relative value a clinic places on those outcomes.

Backfill is the conservative cousin of overbooking. You are not selling capacity you do not have. You are reselling capacity that has just been returned, which carries none of the denied-service risk and captures a decent share of the same value.

Making it run during the show

Three things have to be true or the queue sits there looking useful and filling nothing.

The cancellation has to reach the queue automatically. If a buyer cancels by emailing their concierge, the queue never hears about it. Every cancellation route has to write to the same booking record, which is the same discipline the availability table needs and the same reason the placement pass has to own the booking state rather than a spreadsheet.

The offer has to be one tap. A notification saying a table has become free at 14:20 with an accept button gets a response in minutes. An email asking the buyer to log in and search does not.

There has to be a timeout. Offer to the top candidate, wait 8 minutes, move on. Without a timeout, one unresponsive buyer holds a table until the slot starts, and you have converted a cancellation into two empty tables.

The timeout number is worth tuning against your own data. Too short and you skip past buyers who were in a meeting. Too long and you run out of clock. Eight minutes on a 20 minute slot grid leaves room for two offers before the slot before it ends.

A fourth thing helps and is often skipped: tell the exhibitor. An exhibitor who learns at 14:05 that their 14:20 has been replaced by a different buyer needs the new buyer's profile in the same message, or they will run the meeting cold and rate it poorly. Sending the standby buyer's company, category and stated interest with the confirmation costs one extra field in the template and materially changes how the recovered meeting goes.

Where this stops

Backfill measures well and can quietly make the programme worse, in one specific way that is worth watching.

Every backfilled meeting is a lower scoring meeting than the one it replaced, by construction, because it came from the unplaced set. If you report meetings held without reporting mean match score alongside, a rising backfill rate looks like pure improvement while average meeting quality falls. Report both.

The second limit is that a queue cannot create demand. On a floor where capacity was never the binding constraint and every accepted proposal was placed, the standby list is empty and 40 cancellations become 40 empty tables no matter what you build. Check the size of your accepted-but-unplaced set before you build any of this. If it is under about 20 proposals, the work is not worth doing this year.

The third is participant fatigue. A buyer who receives six standby offers in two days stops reading them, and the fill rate decays across the show in a way that a single blended figure hides completely. Cap offers per buyer per day, and plot fill rate by hour of the show rather than trusting the average.

Start by counting your accepted-but-unplaced proposals from last edition and your cancellations bucketed by notice. Multiply the two through the fill rates you can defend, and if the answer is more than about 15 recovered meetings, build the per-slot queue before you buy another table for the matchmaking zone.

Questions people ask about waitlist backfill for meetings

What is waitlist backfill in a meeting programme?
It is a ranked standby queue of proposals that were accepted or near-accepted but never placed, held per time slot. When a scheduled meeting is cancelled, the freed table and the freed exhibitor are offered to the top candidate on that slot's queue whose buyer is also free, so the capacity gets reused rather than lost.
What backfill rate should a standby queue achieve?
It depends almost entirely on notice. Cancellations arriving days ahead can be filled at high rates because both parties can be reached and confirmed. Cancellations arriving minutes before the slot fill rarely, because the replacement buyer has to be physically near the table. Measure the two separately or the blended figure will mislead you.
Who should be on the standby list for a meeting slot?
Proposals that both parties already accepted but that missed placement on capacity or a clash, ranked by match score. Adding proposals that were never accepted turns backfill into cold outreach, which converts far worse and annoys buyers who already declined that exhibitor once.

Related reading

All matchmaking articles