Attendance data due diligence before you sign a trade show deal

The information memorandum says 14,200 attendees. Underneath the number is a footnote saying attendance is measured on a consistent basis year on year, which is the kind of sentence that gets written when nobody wants to define the basis.

Attendance data due diligence is the work of rebuilding that figure yourself, from rows, before it goes anywhere near a model. It is not an accusation of dishonesty. Most sellers genuinely do not know what their attendance number contains, because the definition was set by whoever configured the registration system four editions ago and has been inherited ever since.

UFI's Global Exhibition Barometer, in its 37th edition published in July 2026 from 466 companies across 59 countries, is the standing survey of what organisers expect and report. What no industry survey can tell you is what one specific show's number means, and that is the only question that matters when you are pricing that show.

Why the attendance number in the deck is unverifiable as printed

Attendance is a compound of at least four decisions, and a single figure conceals all four.

Whether the unit is a person or a badge. Whether attendance means registered, checked in, or scanned entering the hall. Whether exhibitor staff, contractors, press, speakers and organiser employees are in or out. Whether a person attending on all three days counts once or three times.

Pick the generous option at each of the four and you can roughly double the strict answer on the same underlying show. Neither figure is a lie. They are answers to different questions, and the deck never says which question was asked.

For scale, UFI's Global Exhibition Industry Statistics, released in May 2025 covering 2024, counted 318 million visitors across roughly 32,000 exhibitions worldwide. Every one of those visitor counts was produced by an organiser making the same four decisions with no common rule.

Rebuilding the figure from badge records

Ask for one thing: the row level badge export for each of the last five editions. Every row a badge, with badge type, registration source, registration date, the qualification questions and their answers, company, job title, country, and every entrance scan with a timestamp.

If that arrives, you can compute any definition you like and apply it identically to all five editions, which is worth more than any single number in the pack. If it does not arrive, that is itself a finding, and it belongs on the artefact list in the event business data room checklist with a note about what could not be produced.

Then work in this order. Deduplicate to unique people first, because every later subtraction gets distorted if the same person appears twice. Match on email, then on surname plus company plus country, and keep the merge log. On the show in this example, 16,940 badges resolved to 15,610 unique people, so 1,330 badges were duplicates, which is 7.8 per cent of the file.

Next, restrict to people who actually turned up. Of the 15,610, some 14,200 had at least one entrance scan. That is the number in the deck, and it is a defensible one as far as it goes.

One trap sits between those two steps. Some organisers report attendee days rather than people, so a buyer who came on Tuesday and Wednesday counts twice. On a three day show with a 1.7 day average visit, attendee days on this file would be about 24,100, and 24,100 is a number you can imagine appearing on a slide with the word attendance next to it. Check the ratio of scans to unique scanned people before you accept any figure. If the deck's number divided by your unique person count lands near the average visit length, the deck is quoting days.

The dedupe step deserves its own scepticism too. Matching only on email will leave duplicates behind, because the same person registers with a work address one year and a personal one the next. Matching too loosely will merge two people at the same company with the same surname. Keep both counts, the strict and the loose, and report the range: on this file, strict email matching alone left 15,940 unique people and the full rule left 15,610, so the dedupe decision moves the base by 330 people, which is small enough to stop worrying about and large enough to state.

The subtraction, line by line

Here is where the deck's number and the exhibitor's experience of the show come apart.

Exhibitor staff badges: 2,410. These are the people standing on the stands. They are attendees in a fire safety sense and they are the opposite of a buyer.

Contractors and build crew who scanned during show open: 385. Stand builders working through a live day, catering staff, audio visual crew.

Press: 240. Worth having and worth counting separately, because a press pass is a marketing cost.

Organiser staff, speakers and association officials: 190.

Comped badges that failed the show's own qualification rule: 1,125. Not every comp is unqualified, and the test has to be the show's published criteria applied to the answers on the form. Comps that qualify stay in.

Subtract those five lines from 14,200 and you get 9,850 qualified buyers. The subtractions total 4,350, so the qualified buyer count is 69.4 per cent of the reported attendance figure.

What does the reconciliation actually change?

It changes what an exhibitor is buying, and therefore what the floor can be priced at.

Exhibitor revenue at this show was 6.2 million pounds. Against the reported 14,200 attendees, that is 437 pounds per attendee. Against 9,850 qualified buyers, it is 629 pounds. An exhibitor evaluating cost per qualified conversation is running the second calculation, whether or not the organiser is, and the second number is 44 per cent higher.

That gap is the pricing headroom question in one figure. If the show has been raising rates on the strength of the 14,200, and exhibitors have been computing their own return on the 9,850, the rate increases have been landing on a base that already felt expensive to the customer. You would expect to see that as declining price realisation on a constant exhibitor cohort, and if you do see it, you have two independent measurements telling the same story.

Run the same waterfall on all five editions. A show where qualified buyers grew 2 per cent a year while reported attendance grew 7 per cent has been adding badges of a kind exhibitors do not value, and its exhibitor retention will be under pressure whatever the seller's retention slide says. Recomputing that retention properly is the companion exercise in exhibitor retention diligence.

Scan coverage cuts the other way, and it is the mistake buyers make

Everything above assumes the scan data is complete. Often it is not, and a buyer who treats a coverage failure as a demand failure will price the show wrong in the direction that loses the deal for no reason.

Check three things. Scans per entrance per hour, against the venue's actual entrances: an unscanned side door or a hotel link bridge can quietly remove a thousand people. Scan counts during the first ninety minutes of day one, when queues build and staff start waving people through. And the share of registered attendees with zero scans but a paid conference ticket, which is close to impossible and usually means the scanner at that door was offline.

On one file I would expect to find at least one of these. If verified attendance is 82 per cent of registrations on four editions and 61 per cent on the fifth, the fifth edition had an equipment problem, and normalising for it is legitimate as long as you write down what you did and why.

What should you ask for, and in what form?

Six artefacts, and the form matters as much as the content.

The badge export described above, as CSV per edition, not as a report. The show's own written qualification criteria for each edition, because the criteria change and the change is usually undocumented. The comp policy, including who could issue comps and any exhibitor guest pass allotment. The registration form as fielded each year, since a question that disappeared in 2024 explains a data gap that would otherwise look like a quality decline. The entrance map with scanner positions. And the reconciliation the seller performed for any third party audit, if one exists.

Ask for all six in the first request rather than iteratively. A seller who supplies four of six has answered a question about how the business is run, and the missing two shape what you can and cannot claim in the quality of earnings for events work later.

Where this stops

Rebuilding attendance gives you a defensible count of people who were in the building and met a definition. It says nothing about whether those people bought anything.

The honest limit is that qualified buyer count is a proxy for commercial value and a fairly crude one. Two hundred procurement leads from the four largest distributors in the sector are worth more than two thousand junior specifiers, and no badge field distinguishes them reliably. Job title fields are self reported, free text more often than not, and mapped to seniority bands by a rule nobody documented.

The second limit is that the strict number will be lower than the market's expectation, including the expectation of the exhibitors who currently pay for the show. Publishing a recomputed figure after completion is a commercial decision with real consequences, and it is not primarily a data decision. Plenty of acquirers compute the strict number for the model and keep reporting the old basis for a transition period, which is defensible only if the basis change is eventually disclosed rather than quietly never mentioned.

Start with one edition. Get the badge export, deduplicate it, and count how many unique people with an entrance scan hold an exhibitor staff badge. That single line is usually the largest of the five subtractions, and it takes an hour. Everything else in the synergy and integration diligence plan can wait until you know whether the headline figure is 70 per cent buyers or 90.