Skip to content

Returning exhibitor lead trends across editions and how to separate show effects

Exhibitor analyticsUpdated 2026-08-189 min read

In short

Returning exhibitor lead trends carry three things at once: what the stand did differently, what the organiser did differently, and what happened to the audience. Separate them by indexing each stand against the change in unique scanning attendees inside its own product category, then reporting the residual as a ratio rather than a difference in percentage points.

The email said the stand was down 18 per cent and asked what had gone wrong with the show. The rebooking call was in nine days and the account manager wanted an answer that was true.

The stand had captured 312 unique leads at the previous edition and 256 at this one. That part was not in dispute. What nobody in the room could say was whether 256 was a bad result, because the floor those 256 came from was a different floor.

Returning exhibitor lead trends are two numbers doing three jobs at once. A year on year change for one stand carries whatever that stand did differently, whatever the organiser did differently, and whatever happened to the audience. Separating the first from the other two is a small piece of arithmetic and it changes the conversation completely.

Building a denominator the exhibitor cannot argue with

You need a measure of how much the show moved, and the choice of measure decides whether the exhibitor accepts the result.

Registrations are the wrong denominator, because registrations include people who never turned up and the no show rate moves between editions. Verified attendance is closer and still wrong on shows where a large share of the badge population is there for the conference and never walks the hall.

Use unique scanning attendees: the count of distinct badge identities recorded by at least one exhibitor scanner anywhere on the floor during the edition. That is the population which was genuinely available to be scanned, and it has one property nothing else has, which is that it is measured with the same instrument as the numerator. Both sides of the ratio come out of the same scanner estimate, so a change in the instrument mostly cancels.

Mostly, not entirely. If lead retrieval activation moved between editions, the denominator moves for reasons that have nothing to do with the audience, and you should publish the activation coverage figure next to the index every time. E3 and E5 cover what activation is and how to measure it.

Should you subtract the percentages or index them?

Show-wide unique scanning attendees fell from 9,840 to 8,659. Divide: 8,659 over 9,840 is 0.880, so the floor was down 12.0 per cent.

The stand went from 312 to 256. Divide: 256 over 312 is 0.821, so the stand was down 17.9 per cent.

The first way to take the difference is to subtract the percentages. Minus 17.9 against minus 12.0 gives a stand specific gap of 5.9 percentage points. That is the version most people reach for and it is the one I would not report.

The second way is an index. Apply the show's change to the stand's own starting point to get what the stand should have captured, then compare. Expected is 312 times 0.880, which is 274.6. Actual is 256. The residual is 256 over 274.6, which is 0.932, so the stand came in 6.8 per cent below expectation.

The two answers differ, 5.9 against 6.8, and they differ more as the show's own change gets larger. The index is the one to use, for two reasons. It is a ratio, so it is comparable between a stand that captured 312 and a stand that captured 3,120, and percentage point differences are not. It also composes, which the subtraction version does not, and that matters as soon as you have more than two editions.

Card and Krueger set out the general shape of this argument in the American Economic Review in 1994, comparing fast food employment in New Jersey against restaurants immediately across the border in eastern Pennsylvania after New Jersey raised its minimum wage in April 1992. The Pennsylvania restaurants were exposed to the same economy and the same season and not to the wage change, which is what made them a usable control. A show floor gives you the same structure for free. Every other stand at your event was exposed to the same dates, the same venue, the same marketing spend and the same weather.

Choosing the comparator, and what happens when you narrow it

The whole floor is the crudest control available and usually the wrong one, because audience change is not spread evenly across a hall.

Compute the same denominator inside the exhibitor's own product category: distinct badge identities scanned by at least one stand in that category. On this floor it fell from 3,410 to 2,762. Divide: 2,762 over 3,410 is 0.810, so the software category was down 19.0 per cent against a floor down 12.0.

Rerun the index against that. Expected is 312 times 0.810, which is 252.7. Actual is 256. The residual is 256 over 252.7, which is 1.013, so the stand came in 1.3 per cent above expectation.

The sign flipped. Against the floor the stand underperformed by 6.8 per cent and against its own category it slightly outperformed. Both are correct arithmetic and only one of them answers the question the exhibitor asked, which is whether their stand was the problem.

Narrow the comparator as far as your stand counts allow. Category is usually the right level. Hall works where a show is laid out by hall and the categories are mixed. Category crossed with space band is better still where the cells hold up, and E24 sets the minimum size a cell needs before anything computed on it is stable. Whatever level you pick, print how many stands are in the comparator, because a category comparator built from nine returning stands is a number with a lot of luck in it.

Pick the level once, before you look at any individual result. The temptation to choose the comparator that produces the nicer answer for a large account is real and it destroys the method the first time you give in to it.

Chaining residuals across more than two editions

The reason to keep the index in ratio form arrives at the third edition.

A stand with residuals of 0.932, 1.041 and 0.978 across three consecutive transitions has a cumulative residual of 0.932 times 1.041 times 0.978. Work it: 0.932 times 1.041 is 0.970, times 0.978 is 0.949. The stand is 5.1 per cent below where the category would have carried it over three editions.

That single number is worth more than any of the three that produced it. One edition's residual moves on staffing, weather, a competitor launch and who happened to walk the aisle, and a stand can easily be 7 per cent off for a reason nobody will ever identify. Three editions chained together filter most of that out and leave something you can act on.

It also gives you a useful way to sort a renewal list. Rank returning exhibitors by cumulative residual over the last three editions instead of by last year's change, and the accounts that surface are the ones drifting rather than the ones that had a bad Tuesday. A drift that runs across many editions is a different shape again, and plotting lead volume against editions exhibited is E30's job, survivorship problem included.

When does the index stop working?

The index assumes the only thing that changed about the stand is how it was run. Four things break that assumption and all four are recorded in your own systems.

The stand changed size. A company that went from 36 to 54 square metres should have captured more, and comparing raw counts across a size change measures the purchase. Compute the index on unique leads per square metre instead of unique leads, using the contracted space denominator E25 sets out.

The stand moved. A relocation from a main aisle to the back of hall three is something the organiser did, and attributing the resulting fall to the exhibitor is both wrong and easy to disprove. Join the floorplan coordinates for both editions before you send anything.

The show changed shape. A new venue, a date move, a co-location, a change from three days to two. Any of those makes the previous edition a weak control and the honest response is to say so in the methodology note rather than to publish an index that looks precise.

The instrument changed. A new lead retrieval supplier, a different deduplication window, a change in what counts as a unique lead. The denominator and the numerator both shift and the residual is measuring your own configuration change. E11 and E12 cover the definitional side of that.

Reporting the residual before the rebooking call

Give three numbers and the name of the comparator. The stand's own change, the comparator's change, and the residual, with the count of stands in the comparator underneath.

Your stand captured 256 unique leads against 312 last edition, down 17.9 per cent. Unique scanning attendees across the 47 returning stands in your product category fell 19.0 per cent over the same period. Against that, your stand came in 1.3 per cent above expectation.

Three sentences, no adjectives, every figure traceable. An account manager can read that out loud, and an exhibitor who was about to argue about the show now has a specific thing to check on their own side. Very little else in an exhibitor analytics pack survives a rebooking call that well.

UFI's Global Exhibition Barometer, 37th edition, published in July 2026 on a survey of 466 companies across 59 countries concluded that June, asked respondents how rented space in their home market would move in 2026 against 2025. Twenty-six per cent expected an increase of more than 5 per cent, 38 per cent expected stability within plus or minus 5 per cent, 21 per cent expected a decrease of more than 5 per cent, and 14 per cent could not say. Six months earlier the same question had returned 44 per cent expecting an increase and 8 per cent expecting a decrease. Shows move by large amounts in both directions and they do not move together. That spread is the reason a stand's raw year on year figure tells you so little on its own. IAEE's June 2026 piece on winning and keeping exhibitors makes the related operational point, that organisers should be raising booth placement with exhibitors before the exhibitor raises it with them.

The limit

One residual is a single difference with no error bar attached, and 6.8 per cent below expectation sounds meaningful in a way it has not earned.

The fix is to compute the residual for every returning stand in the comparator and look at the distribution. On most floors that distribution is wide, with a middle half spanning something like 15 or 20 points, and a stand at minus 6.8 per cent sits comfortably inside the ordinary range. Report the exhibitor's residual as a position in that distribution and the conversation calibrates itself.

The control has to be a fair proxy for what the stand would have done, and it is not always. If the exhibitor is the dominant company in a small category, they are a large part of their own comparator, and the index will drag their result towards zero by construction. Exclude the stand from its own comparator in that case, and say that you did.

The last one matters most in a renewal conversation. A negative residual is not the same as the exhibitor's fault. It captures everything that happened to that stand and not to its category, including a relocation, a build delay, a neighbouring stand pulling out and leaving a dead frontage, and the aisle closure on the second afternoon. Check what you did to that stand before you conclude anything about what they did.

Pull unique leads and unique scanning attendees for two consecutive editions, at floor level and at category level, and compute the residual for every returning stand. Sort the list and look at the bottom twenty. If most of them share a hall or a category, you have found something about your own floor rather than about twenty exhibitors.

Questions people ask about returning exhibitor lead trends

What denominator should you use to adjust an exhibitor's year on year change?
Unique scanning attendees, meaning distinct badge identities recorded by at least one exhibitor scanner anywhere on the floor. Registrations include people who never turned up, and verified attendance overstates the pool at shows where much of the badge population never walks the hall. Unique scanning attendees is measured with the same instrument as the numerator.
Should you subtract the percentage changes or use an index?
Use the index. Apply the comparator's change to the stand's own starting point, then divide the actual by the expected. In the worked example, subtraction gives a gap of 5.9 percentage points and the index gives 6.8 per cent below expectation. The index is a ratio, so it compares stands of different sizes and it composes across editions.
How narrow should the comparator group be?
As narrow as your stand counts allow, and chosen once before you look at any individual result. Product category is usually right. In the example, the same stand comes out 6.8 per cent below expectation against the whole floor and 1.3 per cent above it against its own category, and only one of those answers the exhibitor's question.

Related reading

All exhibitor performance articles