Weighting benchmark data by size before comparing it to your show

The sector figure on the slide says exhibition space grew 10.4 per cent last year. Your show grew 1 per cent. The show director wants to know what is wrong with your show.

Possibly nothing. Weighting benchmark data by size is the first check to run on a published average before treating it as a target, because a mean built by giving every show one vote describes a different world from one built by giving every square metre one vote, and almost no published figure tells you which you are looking at.

What an unweighted average actually measures

Take five shows in one sector, measured on net exhibition space sold.

A 500 square metre show grew 40 per cent, to 700. A 1,200 square metre show grew 12 per cent, to 1,344. A 4,000 square metre show shrank 3 per cent, to 3,880. A 9,800 square metre show grew 2 per cent, to 9,996. A 50,000 square metre show grew 1 per cent, to 50,500.

Average the five percentages. Forty plus twelve minus three plus two plus one is 52, divided by five is 10.4 per cent. That is the number on the slide.

Now add the space. The five shows held 65,500 square metres between them and now hold 66,420. The gain is 920 square metres, and 920 over 65,500 is 1.4 per cent.

The sector grew 1.4 per cent. The average show in the sector grew 10.4 per cent. Both computations are correct, and they differ by a factor of seven and a half because the 500 square metre show that grew 40 per cent added 200 square metres, while the 50,000 square metre show that grew 1 per cent added 500. In the unweighted mean the smaller one carries the same vote as the larger. In the weighted mean it carries two fifths of the larger one's contribution.

The median is a third answer. Sorted, the five growth rates are minus three, one, two, twelve and forty, so the median is 2 per cent. Three defensible averages of the same five shows: 10.4, 2.0 and 1.4.

Why does a size weighted mean give a different answer?

Because a percentage discards the base it was computed on, and averaging percentages then treats those discarded bases as if they were equal.

The correct weighted figure never averages percentages at all. It adds the numerators, adds the denominators, and divides once. Growth of 920 square metres over a base of 65,500 square metres. That single division is the market rate, and it is the one your own portfolio roll-up should use if the question is how much space the sector absorbed.

The instinct to average the percentages survives because it is what a spreadsheet does when you drag a formula down a column of growth rates and put AVERAGE at the bottom. Nothing in the arithmetic warns you, and the answer looks plausible, which is worse than an answer that looks wrong.

There is a second weighting choice hiding underneath, which is what to weight by. Space, revenue, exhibitor count and visitor count give four different weighted means from the same five shows, because a show that is large in square metres is not necessarily large in exhibitors. Pick the one that matches the decision. Space for hall planning. Revenue for a board pack.

The industry's own averages say they are unweighted

The most widely quoted standing survey in this sector is explicit about the problem, which is more than most reports manage. The 37th UFI Global Exhibition Barometer, published in July 2026 from a survey concluded that June, carries this in its opening remarks:

Also, since weighting of answers based on the size of the responding company or its national market is not possible, it shall be noted that some regional or global consolidations, especially quantitative results, are not as accurate as the ones for the detailed markets.

That is a research team telling you, in the document, that its global lines are one company one vote. UFI (2026) collected 466 replies from 59 countries and regions and publishes the full per-country appendix, which is what makes the consequence measurable rather than theoretical.

Read that appendix against market size and the skew shows up. Europe supplied 178 of the 466 replies and Asia-Pacific supplied 113. Set those against indoor venue capacity from the UFI and jwc World Map of Exhibition Venues, reported in UFI's Global Exhibition Industry Statistics of April 2026 as 16.1 million square metres for Europe and 16.9 million for Asia-Pacific. That is 11.1 European replies per million square metres of capacity against 6.7 for Asia-Pacific, so a European respondent carries roughly 1.7 times the weight of an Asia-Pacific one in any global consolidation. North America supplied 62 replies against 7.6 million square metres, or 8.2 per million.

None of that makes the barometer wrong. It makes the global line a different statistic from the market lines, exactly as UFI says. The mistake is downstream, in the pack that quotes the global figure as though it were the market. Reading that report on its own terms, including which of its lines are sentiment and which are quantities, is worth doing once properly before it becomes a standing input to your planning.

There is a cheap defence against all of this that does not require the publisher to do anything. Whenever a report gives you both a per-market breakdown and a global roll-up, rebuild the roll-up yourself from the market lines using a weight you can defend, then compare it against the published global figure. If the two land within a point of each other, the weighting question is moot for that statistic and you can stop. If they diverge, you have learned something specific about which markets are driving the headline, and that is usually the more interesting finding anyway.

How large is the average show in a certified benchmark sample?

The other half of the problem is that benchmark samples are size-biased before any averaging happens, because being in the sample costs money and effort that small shows do not spend.

UFI's Global Exhibition Industry Statistics, April 2026, puts the global market at 32,000 exhibitions with 138 million square metres of booth space. Divide and the average exhibition worldwide sells 4,313 square metres. That is a show of a few hundred stands, and 4.7 million exhibiting companies across 32,000 exhibitions gives an average of 147 exhibiting companies per event.

Now take a certified sample. UFI's Euro Fair Statistics for 2022, published in November 2023, contains the certified figures for 1,826 exhibitions from 14 countries, of which 1,783 were physical events accounting for 18.4 million square metres of rented space. That is an average of 10,320 square metres per event, about 2.4 times the global mean.

So the certified European sample is made of shows more than twice the size of the world average show, which is unsurprising, because certification is a thing larger organisers buy. If you run a 3,000 square metre show and you benchmark its growth against that sample, the sample is not your peer group whether it is weighted or not, and reading a benchmark whose sample excludes your show is a separate skill from reweighting one that includes it.

UFI (2023) also states that those certified events represented around 47 per cent of the European exhibition market by net space rented, which is a useful honest ceiling on what the report can claim.

Which weight belongs on which question

Four rules that survive contact with a real pre-show pack.

• Space or revenue weighted for market statements. Anything phrased as what the sector did belongs on a ratio of sums.
• Unweighted for operational statements. How many shows grew, what the typical show experienced, whether small shows behave differently from large ones. Here every show genuinely is one observation.
• Report both when they diverge by more than a couple of points. The divergence itself carries information, because it tells you the growth is concentrated in one size band.
• Band before you average. Splitting the sample into size bands and reporting a mean per band beats any single weighted figure, and it is the version an exhibition sales director can use.

The last one matters most. Once you have three or four size bands with a mean each, the reader can find the band containing their show and stop arguing about methodology.

Where this stops

Weighting fixes the arithmetic. It does nothing about a sample that never contained a show like yours, and it can make a benchmark less useful rather than more.

A space-weighted sector figure from a sample where two shows hold 60 per cent of the space is, in practice, a report on those two shows. You have replaced one show one vote with something closer to two shows all votes, and the confidence interval around it is narrower than it should be because the effective sample size has collapsed. There is a standard diagnostic here worth running: compute the share of total weight held by the largest three contributors, and if it is above about half, say so next to the figure.

There is also a case where the unweighted mean is the honest one and gets rejected for looking bad. If forty small shows grew and two large ones shrank, the weighted figure is negative and the unweighted figure is positive, and both belong in the report. Choosing the one that supports the argument you already had is the failure this whole exercise exists to prevent, and how a study defines the population it drew from is where that choice usually gets buried.

For a first step, open the last benchmark figure you quoted, and find out whether the publisher averaged percentages or divided sums. If the report does not say, treat it as an unweighted per-show average until told otherwise, and recompute your own portfolio the same way so the comparison is at least like for like. The rest of the measurement standards work follows from getting that one division right.