Exhibitor benchmarking within a show beats comparing against published industry averages

An exhibitor arrived at the sales office on the last morning with a printout. Somebody's blog post, a figure for average leads per exhibitor at B2B trade shows, and a question: we got 143, so why is your show half as good as the industry?

The number on the printout had no show attached to it, no year, no country, no product category and no definition of a lead. Exhibitor benchmarking within a show holds all five of those fixed, which is the whole reason to prefer it. The figure she had brought was, as far as anyone could tell, a mean of means from a survey of exhibiting companies who had self-reported whatever their own systems produced. And she was entirely reasonable to bring it, because we had given her nothing better to compare against.

What is a published industry average actually measuring?

Published exhibition research is worth reading and worth citing. It is not worth using as a per-stand yardstick, and the reason is visible in the research itself.

UFI's Global Exhibition Barometer, 36th edition, published January 2026 and concluded on data collected in December 2025, surveyed 378 companies across 57 countries. On activity in 2025, 47 per cent of respondents reported an increase of more than 5 per cent in their domestic market, 42 per cent reported stability within plus or minus 5 per cent, and 10 per cent reported a decrease of more than 5 per cent. The report breaks its findings out across 19 focus countries and regions and five aggregated regional zones precisely because the aggregate conceals so much.

Hold that shape in your head. Nearly half the industry grew strongly, a tenth shrank, and the average of those two groups describes neither. Any single global figure for what an exhibitor typically achieves is built on the same underlying spread, plus differences in product category, plus differences in what each respondent counts as a lead.

Costs move the same way. The Exhibitor Advocate, in its 2025 annual survey of exhibition rates, independently audited by EVOLIO Marketing, reported installation labour rates rising between 12.7 and 16.7 per cent from 2022 to 2025 depending on whether the rate was straight, overtime or double time, with electrical overtime up 41.2 per cent and material handling secondary rates up 26.4 per cent. Those increases land differently by venue and by city. An exhibitor comparing a cost per lead against a national average is comparing across a spread that the average was never built to represent.

The comparator that holds everything constant

There is exactly one population that shared your audience, your open hours, your marketing spend, your venue, your aisle layout, your weather and your calendar collision with a competing event. It is the other stands at your own show.

That is the entire argument, and it is stronger than it looks, because the confounders you are eliminating are the large ones. Audience composition explains more variance in stand-level lead volume than almost anything an exhibitor controls. A show with 4,000 verified buyers produces different lead counts from one with 22,000, and no amount of stand craft closes that gap. Comparing within the show removes it by construction.

It also removes the definition problem. Your lead count means one thing, applied identically to every exhibitor, because your system produced all of it. Two exhibitors reporting 400 leads from two different shows may have counted two entirely different things. Two exhibitors reporting 400 leads from your show counted the same thing.

Ranking inside the show

Take a food ingredients exhibition with 412 exhibiting companies. The exhibitor with the printout captured 143 unique leads.

Show-wide, the median is 118 and the upper quartile is 267. So 143 sits above the middle of the floor. Ranked against all 412, she is 197th from the top, which puts her at the 52nd percentile.

Now rank within her category. Flavours and extracts has 38 exhibiting companies, and their median is 156. Against those 38 she ranks 22nd, at the 43rd percentile. The category is a strong one and she is slightly behind the middle of it.

Add the space band. Among the 14 stands in flavours and extracts between 18 and 54 square metres, the median is 121 and she ranks 5th, at the 69th percentile.

Three defensible numbers, all from the same underlying count. Which one you report is a decision you should make once and apply to everybody, and I would report the category and space band figure with the peer count printed beside it, falling back to category alone when the cell is thin. Fourteen is small, and E24 covers where the floor on cell size should sit and what to do when you drop below it.

The point is that all three of these tell the exhibitor something she can act on, and the industry average told her nothing except that she had been cheated. The 69th percentile figure also happens to be the honest one, because it compares her against stands that bought comparable space in the same product area.

Notice how much the answer moved. Fifty-second percentile against the whole floor, 43rd against her category, 69th against her category and space band. All three are computed from the same 143 unique leads, and the spread between them is 26 percentile points. That spread is not noise. It is the size of the effect that category and space have on lead volume, measured on your own data, and it is the reason a single show-wide ranking will make small stands in strong categories look worse than they are and large stands in weak categories look better.

Run that comparison across the whole floor once and you can put a number on it. Compute every exhibitor's show-wide percentile and their category-and-space-band percentile, then take the absolute difference for each. If the median difference across 412 exhibitors is 4 points, your floor is homogeneous enough that a single ranking is defensible and you have saved yourself a great deal of work. If it is 18 points, the show-wide ranking is systematically misdescribing most of the floor and the peer group work is unavoidable.

What goes in the exhibitor report

A rank statement needs four parts to survive contact with a stand manager: the raw value, the peer group definition, the peer count, and the position.

Your stand captured 143 unique leads. Among the 14 exhibitors in flavours and extracts holding between 18 and 54 square metres, that places you 5th, at the 69th percentile. The peer group median is 121.

Every part of that sentence is doing work. The raw value lets her reconcile against her own scanner export. The peer group definition tells her who she is being compared to, and she will immediately picture roughly the right set of companies, which is the point. The count tells her how much weight to put on the percentile, and 14 is small enough that she should treat 69th percentile as a band rather than a precise position. The median gives her the gap in the units she actually cares about, 22 leads, which is a more useful thing to think about than 19 percentile points.

What does not go in is any other exhibitor's raw number. You will be asked. The answer is that peer values are aggregated and never itemised, and that the same rule protects her figures from the stand next door. Say it once in the front of the pack and the question mostly stops coming.

What if your show really is the problem?

Sometimes the exhibitor is right and the show is the problem. Within-show benchmarking cannot detect that, by construction, because everything is measured against the same floor.

The way to catch it is at the show level, on a different metric. Track your own lead capture per verified attendee across editions, and across the shows in your portfolio. If one show returns 0.9 captured leads per attendee and the rest of the portfolio sits between 1.6 and 2.1, that is a show-level finding worth investigating, and it will point at scanner adoption, hall layout or audience quality rather than at any individual stand. If instead one product area is dragging while the rest of the floor holds up, diagnosing a whole underperforming category is E26's method and a different piece of arithmetic.

That comparison works because you own both sides of it and you defined the metric. Comparing your figure to a published average for the sector does not work, because you do not know how the published figure was constructed, and in most cases neither does the person who published it.

What to do with published research instead

Use it for direction and context, which is what it is good at.

The CEIR 2026 Marketing Spend Decision Report is useful for telling your sales team that exhibitions hold 40.8 per cent of participant marketing spend, that 47 per cent of exhibitors expected to attend the same number of exhibitions in 2026 and 28 per cent expected to add shows, and that 83 per cent expected to hold their booth size steady. Those are facts about the market you are selling into. None of them is a target for a stand in your fluid power aisle.

Cite the research in the front of your exhibitor pack, then benchmark inside your own floor for the rest of it. That split is the shape most exhibitor analytics reporting should take, with published work setting context and your own file carrying every per-stand number. If an exhibitor brings you a published figure, the answer is not that the figure is wrong. The answer is that it was measured on a different population under a different definition, and here is the comparison that was not.

The limit

Within-show benchmarking makes a floor internally comparable and tells you nothing about whether the floor as a whole is good.

If your scanning coverage is poor across the entire show, every percentile you publish is a rank within a badly measured population, and the exhibitor at the 80th percentile is the best of a group that mostly failed to capture. Percentiles are perfectly happy to describe a floor where nothing worked. Count the stands that captured nothing before you publish any ranking at all, which is E3's territory and the first thing to look at.

There is also a fairness problem in small categories that no amount of method solves. A category with six exhibitors produces percentiles that jump by twenty points per rank position, and the stand that came fourth of six is at the 40th percentile whether it captured 200 leads or 20. Where the cell is that thin, report the rank and the count in words, say fourth of six, and leave the percentile out.

Pull your last edition's lead export and, for a single exhibitor who complained, compute all three figures: their percentile across the whole floor, within their product category, and within their category and space band. If the three numbers disagree by more than fifteen points, decide now which one you will publish next edition, and publish the same one for everybody.