How to handle a show date change registration comparison

The show had moved from late March to early June, which the commercial team had wanted for three years and had finally got. On the first of February the weekly pack went out with registrations down 22 per cent on the same date last year, and by lunchtime two people had asked whether the move had been a mistake.

The move had not been a mistake. The comparison had. A show date change registration comparison cannot be run on calendar dates, and on the first of February the previous edition had been 52 days from opening its doors while this one was 122 days from opening its. Those two numbers were measured at points in the campaign more than two months apart, and putting them side by side because the calendar happened to line up is a category error that a date change makes unavoidable and obvious.

Twenty two per cent that was never there

Write the two readings out properly.

The previous edition opened on 25 March 2025. This edition opens on 3 June 2026. The shift is 70 days, ten weeks to the day.

On 1 February 2025 the previous edition held 7,800 registrations, at day minus 52. On 1 February 2026 this edition held 6,084 registrations, at day minus 122. The ratio is 6,084 divided by 7,800, which is 0.78, so the pack reported minus 22 per cent.

Now rebase. Day minus 52 for this edition falls on 12 April 2026, and on that date the file held 8,190. Against the previous edition's 7,800 at its own day minus 52, that is 8,190 divided by 7,800, which is 1.05. The show is 5 per cent ahead at the same distance from doors.

The gap between minus 22 per cent and plus 5 per cent is entirely an axis choice. Nothing about the show changed between the two calculations.

The rebasing rule, in full

The rule is short and the second half of it is the part people skip.

First, align on days to show open. Every registration record carries a created timestamp. Subtract it from the opening morning of the edition it belongs to and you have a days out value for every row, which lets you build a cumulative curve for each edition on a common axis. Compare editions at equal days out, never at equal dates. Once a show has moved, calendar comparison compares two arbitrary points. It looks like a rougher version of the right answer and behaves like a different question.

Second, check whether the campaign window itself changed length. Days out alignment assumes both editions had been selling for a comparable stretch by the time you read them, and a date move is exactly the circumstance where that quietly stops being true, because registration opening is usually scheduled off internal readiness rather than off the show date.

On this show, the previous edition opened registration on 8 October 2024, which is 168 days before doors. This edition opened on 2 January 2026, which is 152 days before doors. The window is 16 days shorter. At day minus 52, the previous edition had been selling for 116 days and this one for 100.

That makes the plus 5 per cent a slightly stronger result than it looks, since this edition reached a higher count with a fortnight less campaign behind it. It also means any reading taken before about day minus 100 is worthless on this pair of editions, because one of them had not opened registration yet. Mark that cut off on the chart and refuse to report anything to the left of it.

What does rebasing fail to fix?

Aligning the axis fixes the measurement. It does not make the two editions comparable, and pretending otherwise is the second mistake people make after fixing the first one.

A March show and a June show sell into different quarters. Budget cycles, competing events, school holidays and the weather in the host city all changed when the date changed, and every one of them acts on the registration curve. The rebased comparison is now measuring the move rather than the campaign, which is a legitimate thing to measure as long as somebody says so out loud.

The shape of the tail is the piece most likely to shift. Maritz's Registration Insights Report 2024, built on more than 360,000 attendee registration records across 30 trade shows, found that 62 per cent of attendees living within driving distance of an event registered in the final four weeks, against 39 per cent of attendees who had to fly. Registration timing tracks travel commitment. If a June date pulls in more regional visitors and fewer long haul ones, the curve will run later than the March edition's did, and a rebased comparison that looks healthy at day minus 52 can still land somewhere different at doors.

The competitive calendar is the other one. UFI's Global Exhibition Barometer, 36th edition, published in January 2026 from 378 companies across 57 countries and regions, found competition from within the exhibition industry named in 12 per cent of answers as the fourth most pressing short term business issue, up two points on six months earlier. Shows move dates to get out of somebody's way and land in somebody else's window. Before you attribute a rebased difference to your own campaign, list the competing events inside the new window and inside the old one, and check whether that list changed more than your media plan did.

What goes in the pack

Report both numbers and label them. One line saying registrations are 5 per cent ahead at the same days out, one line saying they are 22 per cent behind on the same calendar date, and one line stating that the show has moved 70 days.

Putting the calendar figure in deliberately is worth the space. It is the number that will reach the board through some other route within a fortnight, usually from a system that reports on dates because that is what systems do, and it is far better arriving with its explanation attached. It also happens to be the number your cash collection follows, since the money arrives when the registrations arrive and the finance calendar did not move with the show.

Put the axis on every chart in words. Days to show open, this edition opens 3 June 2026, previous edition opened 25 March 2025. A chart with an unlabelled x axis in a year with a date move will be misread by somebody senior, and they will be right to misread it.

Which comparison survives a date move?

Final registrations against final registrations survives a date move, because it is measured at a point both editions actually reach.

That is a cold comfort mid campaign and it is worth saying plainly: for one cycle, you have no reliable pace read. The previous edition is a poor denominator, since its curve was shaped by a different season, and the edition before that is worse. You can rebase the axis, and you cannot rebase the year.

Two things partly fill the gap. The first is composition, which is more stable across a date move than volume and is the part of attendee analytics still worth reading in a year like this. If the returning share, the international share and the registration type mix at day minus 52 look like they always have, the audience is behaving normally even though the count cannot be judged. The second is the sales file, since exhibitor space booked for a moved edition is decided by a small number of people who tell you why, and a floor that sold on schedule into a new date is real evidence about the move.

Then let the moved edition become the new baseline. Once you have two editions in the new slot, the pacing comparison works again and the March years become history with an asterisk. Deciding how many prior editions to keep after a break of that size is O32's subject, and my own preference is to keep them, mark them, and let the reader see the discontinuity rather than deleting it.

Where this stops

The rebasing rule assumes the created timestamp on a registration record means what you think it means. Platform migrations, bulk imports and rebuilt files all corrupt it, and a date move often coincides with a system change because both come out of the same strategic review. Check the earliest created date in each edition against the day registration actually opened before you trust any of this, which is part of A8's territory.

The second limit is that days to show open is one axis and not the only defensible one. A show that moved from a Tuesday start to a Sunday start, or from three days to two, has changed more than its position in the year, and days to open will not capture that. When the format moves as well as the date, the honest answer is that this edition is a new series with one observation in it.

The third is arithmetic. A rebased index computed early in a shortened campaign window sits on a small base and moves violently. At day minus 122 this edition held 6,084 registrations and the previous one had not been selling long enough to have a comparable figure at all. Below a few thousand registrations in the denominator I would report counts and skip the percentage.

The first step takes an afternoon. Take the two editions either side of your move, compute days to show open for every registration record from its created date, and write down four numbers: each edition's show open date, each edition's registration open date, and the difference between them in days. If the two campaign windows differ by more than a fortnight, you have found the reason your comparison feels wrong before you have plotted anything.