What a late registration surge does to your show planning

The catering guarantee for a food and beverage show went to the venue on a Tuesday, fourteen days before doors. The registration file held 9,660 people. The show opened with 11,640 registered and finished with 12,000 once the on site desk closed on day two.

Nobody did anything wrong. The guarantee was submitted on the date the contract required, using the number that existed on that date. That number was 19.5 per cent below the number that walked in, because a late registration surge is what the last fortnight of a B2B campaign produces, and the operational commitments that depend on the file are locked before it lands.

The final month, week by week

Take a 12,000 registration target with 30 per cent of the file arriving in the last four weeks. That is 3,600 registrations, and they do not arrive evenly. On this show they came in like this.

Four weeks out, in the seven days from day minus 28 to day minus 22, 560 registrations. Three weeks out, 700. Two weeks out, 840. The final seven days before doors, 1,140. Then 360 more registered at the desk across the two show days.

Read that as cumulative and the problem is obvious. At day minus 28 the show held 8,400. At day minus 21 it held 8,960. At day minus 14 it held 9,660. At day minus 7 it held 10,500. It opened on 11,640 and closed on 12,000.

The final week alone delivered 1,140 registrations, which is 9.5 per cent of the whole file, arriving after every commercial commitment of any size had been signed.

Maritz's Registration Insights Report 2024, drawn from more than 360,000 attendee registration records across 30 trade shows over three years, found 45 per cent of attendee registrations landing inside the final four weeks and 22 per cent in the week of the show, a figure that includes the nine per cent who did not register until they arrived on site. It also found the effect is heaviest in exactly the sectors that feel it most, with 54 per cent of food and restaurant show registrations arriving in the final four weeks against 29 per cent for medical and healthcare conferences. A 30 per cent tail is a moderate case.

What is the catering guarantee actually set from?

The catering guarantee comes off expected attendance, which is registrations multiplied by a turnout rate. That multiplier is a separate estimate with its own error, and it belongs to A10.

Run it through anyway, because the arithmetic is the point. At a 70 per cent turnout, the 9,660 registrations on the lock date imply 6,762 people in the building. The 12,000 the show finished with imply 8,400. The gap is 1,638 people.

Venue catering contracts usually oblige the caterer to serve some tolerance above the guarantee, commonly around five per cent, and to bill you for the guarantee or the actual consumption, whichever is higher. Guaranteeing 6,762 buys you service for 7,100. You are then 1,300 lunches short on a floor with 8,400 people on it, and the recovery options at that point are all bad and all expensive.

The reverse failure is quieter and more common on shows that have grown used to being burned. A guarantee set from a paranoid forecast of 9,000 attendees against an actual 8,400 costs you 600 covers you have to pay for and nobody ate. At an illustrative 22 per head that is 13,200 gone, with no line item anyone will ever query because catering came in under contract.

The first morning at the registration desk

The surge shows up hardest as throughput, and throughput is where the arithmetic gets uncomfortable.

A pre-registered visitor with a QR code collects a badge in roughly twenty seconds at a kiosk. Somebody registering from scratch at the desk takes closer to three minutes once you allow for form completion, payment where there is one, and the questions. That is a nine to one difference in staffed time per person.

This show had 360 on site registrations across two days. Assume 70 per cent of them arrived in the ninety minute window either side of the opening keynote, which is 252 people. At three minutes each that is 756 person minutes of desk time inside a 90 minute window, so 756 divided by 90 is 8.4, and you need nine staffed positions doing nothing but on site registration during the busiest ninety minutes of the show.

Now suppose your show behaves like the Maritz average and nine per cent of the file registers on site. On 12,000 that is 1,080 people, of whom 756 arrive in the same window. That is 2,268 person minutes in 90 minutes, which is 25.2 positions. Between a three per cent on site share and a nine per cent one sits sixteen extra staffed desks, hired weeks earlier from a labour supplier working to its own lead time.

Nobody sizes the desk from that calculation. They size it from last year's desk, which is a reasonable heuristic right up to the year the mix moves.

The staffing plan is thin before the surge arrives

The organiser side has less slack than the venue side. UFI's Global Exhibition Barometer, 35th edition, published in July 2025 from 386 companies across 58 countries and regions, found 40 per cent of businesses planning to increase staff numbers over the following six months, which means the majority were planning to run the next cycle with the team they had.

The people who absorb a late surge are the same people running the campaign that caused it. Registration support tickets, badge corrections, exhibitor guest code chases, and the visa letters that arrive with three weeks to go all land in the final fortnight on a team that is also loading in. Sizing that fortnight in advance is one of the few things attendee analytics can hand an operations plan before the surge arrives.

Forecast the tail instead of counting it

The fix is to stop feeding operational commitments a live count and start feeding them a forecast made at the same moment.

At the lock date, day minus 14, the show held 9,660. If your last five editions held a median 80.5 per cent of their final registration number at day minus 14, then the implied final is 9,660 divided by 0.805, which is 12,000. That is the number the guarantee should have been built from, and it was available on the day the guarantee was due. The method behind that division, and what it assumes, is A9's subject.

Two conditions make it usable. The historic share has to come from editions whose registration open date and price tier boundaries sat at comparable distances from doors, because otherwise you are dividing by a share from a differently shaped campaign. And the estimate has to travel with a range. If the five editions held between 78 and 84 per cent at day minus 14, the implied final runs from 11,500 to 12,385, and the guarantee decision should be made against the low end while the desk staffing decision is made against the high end. Those are asymmetric costs and they deserve different points on the same interval.

Why is pulling the surge forward the wrong instinct?

Every show director's first response to a heavy tail is to discount harder and earlier. The evidence for that working is thin, and the evidence against it is specific.

The Maritz 2024 analysis measured what attendees spent above and beyond the registration fee, on items such as added sessions, on demand content and social events. Registrants arriving in the final four weeks spent 59 dollars more per head on those items than registrants who booked more than 90 days out. The heaviest spending sat in the 31 to 60 day window at 370 dollars a head, against 278 for the earliest registrants at 120 or more days out, a gap of 92 dollars. The people who register the moment the site goes live are largely repeat attendees who were coming regardless, so an early bird discount aimed at them mostly discounts revenue you already had.

There is a real reason to want registrations earlier, and it is operational rather than commercial. Every week a registration arrives earlier is a week of certainty for the guarantee, the desk and the shuttle contract. That case is worth making honestly, and it is a weaker case than the revenue one people usually reach for.

Where this stops

The tail share is not stable across editions of the same show, and treating it as a constant is the failure mode of everything above. A date move, a change in international mix, a co-located event, or a competitor moving into your window will all shift it, and the shift will be invisible until the tail either arrives or does not. Plotting the whole campaign curve across several editions is where that movement shows up first, and reading its shape is A2's subject.

The turnout multiplier compounds the problem. A late registration surge and a lower turnout rate among late registrants can cancel each other out in the attendance total while wrecking your desk plan, because those people still queue whether or not they were always coming. Registrations and attendance move together loosely and the operational load tracks registrations, so plan the desk on registrations and plan the catering on attendance.

The honest limit is that on site registration is the one number you can never forecast well, because the people producing it made their decision after your last measurement.

Take the last three editions of one show and compute, for each, the registration count at day minus 14 and the final count. Divide one by the other and you have three observations of the share you had at the moment your catering guarantee was due. Put those three numbers next to the guarantee you actually submitted each year, and you will know within an hour whether your show has been over-ordering or under-ordering, and by how much.