Exhibitor tier performance and whether a premium tier really delivers more leads
Exhibitor tier performance has to be measured on unique leads per square metre, computed inside product category and then recombined against one common set of category weights. Tiers bundle space and concentrate in particular categories, so a pooled comparison of totals can show a premium tier ahead while it trails inside every category on the floor.
The renewal deck had one chart on the tier page. Gold exhibitors averaged 412 unique leads, standard exhibitors averaged 168, and the gold package cost about two and a half times standard. The chart made the upgrade look like the cheapest line on the rate card.
Somebody asked how big the gold stands were, which is the first question exhibitor tier performance has to answer. The median gold stand held 90 square metres and the median standard stand held 18. Nobody in the room could say what survived of the gap once that was accounted for, and the deck went out unchanged.
A tier is a commercial promise, and where part of that promise is lead volume, the promise has to survive measurement. The measurement is harder than a group mean, because tier membership is correlated with almost everything else that moves lead volume.
Totals by tier are a chart of how much space you sold
Gold stands are bigger. They are bigger by design, because the tier usually bundles a space minimum, and the moment it does, any comparison of totals is measuring the bundle rather than the tier.
Use unique leads per square metre instead, computed inside product category. Why the contracted space figure is the defensible denominator, and where dividing by area misleads, is E25's argument and worth reading before you run this. What matters here is that the tier comparison has to happen on a size-neutral figure, and that the size-neutral figure is the one you already compute for everything else on the floor.
Use medians. Lead volume inside any group on a show floor is skewed hard enough that a mean of 40 gold stands can be set by two of them, and E21 covers why.
That change alone kills most tier charts. It also produces a new problem, which is the interesting one.
The gap that reverses when you split by category
Take a floor of 200 exhibiting companies in two product categories. Automation has 92 companies and software and services has 108. Gold holds 40 companies, standard holds 160.
The tiers sit very differently across the two categories. Of the 40 gold stands, 32 are in automation and 8 are in software, so 80 per cent of gold is automation. Of the 160 standard stands, 60 are in automation and 100 are in software, so 37.5 per cent of standard is automation.
Now the within-category medians of unique leads per square metre. In automation, gold returns 6.4 and standard returns 6.6. In software, gold returns 3.9 and standard returns 4.1. Gold is behind in both categories, by 0.2 either way.
Pool them anyway, which is what a tier chart does. Sort the 40 gold stands and the middle pair lands in automation, because automation is 32 of the 40, so the pooled gold median is about 6.4. Sort the 160 standard stands and the middle pair lands in software, because software is 100 of the 160 and those are the lower values, so the pooled standard median is about 4.1.
Gold 6.4, standard 4.1. Gold is ahead by 2.3, which is 56 per cent. On the same numbers, gold is behind in every category on the floor.
Simpson set this out in the Journal of the Royal Statistical Society, Series B, in 1951, in a four page paper on the interpretation of interaction in contingency tables. The association you observe in pooled data can run opposite to the association inside every subgroup, and the direction of the pooled result is decided by how the subgroups are weighted. A tier that concentrates in your highest-yielding category is exactly that weighting, applied by your sales process rather than by chance.
Standardising each tier to the floor's category mix
The fix is old and it is arithmetic rather than modelling. Compute each tier's median inside every category, then recombine those medians using one common set of category weights, so both tiers are scored against the same floor.
The floor mix is automation 92 of 200, which is 0.46, and software 108 of 200, which is 0.54.
Standardised gold: 0.46 times 6.4 is 2.944, plus 0.54 times 3.9 is 2.106. Total 5.05.
Standardised standard: 0.46 times 6.6 is 3.036, plus 0.54 times 4.1 is 2.214. Total 5.25.
Gold 5.05, standard 5.25. Standard is ahead by 0.20 leads per square metre, about 4 per cent. The headline moved from gold ahead by 56 per cent to gold behind by 4 per cent, and nothing changed except which category mix both tiers were scored against.
Space band deserves the same treatment for the same reason, because tiers bundle size. Cross category with space band, compute the tier median in every cell, and weight the cells by the floor's share. The cells go thin fast and E24 sets out the minimum you need before a cell can carry a percentile, which applies here too. Where a cell has fewer than a dozen stands in one tier, collapse it into its category and record that you did.
Is the surviving tier gap large enough to be anything?
A standardised gap of 0.20 on 40 gold stands is not obviously distinguishable from noise, and reporting it as a tier effect would be the same defect one layer down.
The cheapest honest test does not need a distributional assumption. Hold each stand in its category, shuffle the tier labels within that category, recompute the standardised gap, and repeat a couple of thousand times. You now have a distribution of gaps produced by a floor where tier means nothing. Read off how often a shuffled gap is at least as large in absolute value as the one you observed. If that happens in a third of shuffles, the tier gap is a shuffle.
This is worth building once, because you will run it every edition and because it is the only part of the analysis a sceptical sales director cannot argue with. The shuffle uses your own floor, your own categories and your own stands, so there is no external benchmark to dispute.
Set the decision rule before you look. A tier claim that survives standardisation and survives the shuffle is a finding. Anything else goes in the methodology note and not in the deck.
What is a premium tier really selling once the lead gap is gone?
Most premium tiers, tested this way, turn out to be selling position. Gold stands sit near entrances, on main aisles, beside catering, or in the hall that opens first. That is a real thing to sell and exhibitors want it. CEIR's 2024 report on exhibitor future outlook, part of its series on maximising engagement on the exhibition floor across the North American market, found 33 per cent of exhibitors asking organisers for better attendee traffic flow and 14 per cent asking for booth location flexibility. Position is what they are buying.
The consequence for measurement is that you should test position directly instead of testing the label wrapped around it. Join every stand to its floorplan coordinates and ask whether distance from the nearest entrance, or aisle rank, predicts leads per square metre once category and space band are held fixed. If it does, you have a location effect, and what that location is worth in money is a pricing question. Corner and island premiums are F5 and F6, and fitting a hedonic model to your own floorplan is F7, none of which is a post-show exhibitor analytics job.
Gopalakrishna and Lilien, in Marketing Science in 1995, separated trade show performance into attraction, contact and conversion, and modelled booth space, pre-show promotion, attention-getting techniques, competition, and the number and training of booth staff against those stages. The separation is useful here. A tier that buys traffic acts on attraction and should move the count of people who pass and stop. A tier that bundles a speaking slot or a hosted meeting programme acts somewhere else entirely, and a lead-per-square-metre comparison will never see it.
Tiers also sell things that no scan records. Explori and UFI's 2025 Channel Insights Report, drawing on more than 3,000 events between 2017 and 2025, tracked exhibitor sentiment measures that moved sharply after the pandemic, with average exhibitor net promoter score up 27 points and traditional trade shows averaging 8.5 against 27.9 for B2B festivals. Sentiment is measured by survey and it moves on its own schedule. A tier can be worth buying on sentiment while returning nothing measurable in leads per square metre, and the honest report says both.
Reporting tier performance to the sales team that sold the tier
Give the standardised figure and the raw one together, with the category mix printed beside them. Gold is 80 per cent automation and standard is 37.5 per cent automation. That single line explains the whole reversal to a non-analyst in about ten seconds, and without it the standardised number reads as an analyst having taken a good result away.
Never publish the tier comparison without the per-tier stand counts. Forty against 160 is a real asymmetry and it is why the shuffle exists.
Say what the tier is now claimed to deliver. If the lead gap does not survive, the tier is selling position, visibility and a bundle of services, and the renewal conversation is better for being about the thing that is actually true. An exhibitor who was sold a lead uplift and did not get one works it out from their own CRM eventually, and they work it out in the week before they decide whether to rebook.
The limit
The standardisation only removes the confounders you put into it. Category and space band are the two that matter most and they are the two you have recorded cleanly. Stand format is a third and it is usually missing: an island unit with four open sides and an inline unit of identical area do not perform alike, and if gold stands are disproportionately islands then part of what you just attributed to the tier is frontage.
There is a deeper problem that no amount of stratification fixes. Exhibitors choose their tier. The companies that buy gold are the ones with budget, a launch to run, a bigger team to staff the stand and a reason to be at the front of the hall. Any tier comparison is comparing self-selected groups, so a gap that survives standardisation still cannot be read as what gold would have done to an average exhibitor. The only clean answer would be assigning tiers at random, which no organiser is going to do.
The nearest practical approximation is watching stands that switch. Take the companies that moved from standard to gold between two editions and compare each one against itself, then do the same for the companies that moved down. That comparison has its own problems, including that switching is not random either, and it is still closer to a causal reading than any cross-section of one edition.
Take last edition's export, add the tier column, and compute the median unique leads per square metre for each tier inside each product category rather than across the floor. Put the two sets of numbers side by side. If the within-category gaps disagree in direction with the pooled gap, stop using the pooled figure this week, before the renewal deck is built on it again.
Questions people ask about exhibitor tier performance
- Why do gold exhibitors appear to capture more leads than standard ones?
- Usually because the tier bundles a space minimum, so the comparison is measuring the bundle. In the worked example here the median gold stand held 90 square metres against 18 for standard. Once you switch to unique leads per square metre and compute inside product category, most of the apparent gap disappears.
- How do you correct a tier comparison for category mix?
- Compute each tier's median inside every category, then recombine those medians using the floor's own category shares so both tiers are scored against the same mix. In the example, gold is 80 per cent automation and standard 37.5 per cent, and standardising moves the headline from gold ahead by 56 per cent to gold behind by 4 per cent.
- How do you know whether a surviving tier gap is real?
- Shuffle the tier labels within each category, recompute the standardised gap, and repeat a couple of thousand times. That gives you the distribution of gaps a floor produces when the tier means nothing. If a shuffled gap matches or beats the one you observed in a third of runs, the gap you reported is a shuffle.
Related reading
- Building exhibitor peer group percentiles that a small stand can actually trust
- Leads per square metre normalises exhibitor performance without touching booth pricing
- Building a hedonic pricing model for booth location on your own floorplan