Excluding staff badges from attendance and disclosing what you took out

An exhibition sales director asks a fair question on the Monday after close. We printed 32,700 badges. Why is the attendance number on the front of the report 25,100, and who decided to remove 7,600 people from my show?

Nobody decided anything sinister. Somebody applied a filter, correctly, and then failed to write down what the filter was. That omission is the whole problem, because a number that shrinks by 23 per cent between two slides in the same deck will be argued about for the rest of the week, and the argument is never really about the arithmetic.

The fix for excluding staff badges from attendance is to classify every badge at the point it is created, report attendance by class, and publish the classes underneath the headline. It is cheap, it takes one field on the registration form, and it is the least glamorous piece of attendee analytics that will save you a week a year.

Who in the hall is actually a buyer?

Walk the floor at ten in the morning on day two and count the categories of human being wearing lanyards.

Exhibitor staff, who are there to sell. Stand builders and contractors finishing a repair. Organiser staff running the show. Venue and catering staff. Press. Speakers who came for their session and left. Association board members. Sponsors' agency teams. Accompanying persons brought along by a delegate. At a show with a consumer day, children.

Every one of those people crossed an entry scanner, so every one of them appears in the access control feed, whatever reconciling the scan totals across your systems later says about them. None of them is the audience an exhibitor paid to meet. An attendance figure that includes them is measuring building occupancy.

The reflex fix is to subtract exhibitor staff and stop. That is the largest single block and it feels like the honest correction. It is not enough, because the categories left behind are the ones that move most between editions. A show that adds a conference track adds several hundred speaker and delegate badges. A show that brings a build contractor in-house moves a few hundred badges from one bucket to another. Neither event changed the audience, and both move an uncorrected attendance figure.

The auditing standard already draws these lines for you

You do not have to invent the taxonomy. UFI's Calculation Standards and Definitions, which is Annex 2 of its auditing rules for UFI Approved Events in the June 2021 version, draws the boundaries from the auditing side, and they are stricter than most in-house ones.

Five categories are excluded from the calculation of visits outright: staff of the exhibition venue and organiser, staff working for service providers, staff of exhibiting companies and organisations, speakers during the event, and media representatives. A visitor, in the same document, is a person attending the exhibition who is counted once for the entire duration regardless of how many visits they make.

Total attendance is then built by addition, which is the half most organisers have never read. Unique visitors plus exhibitors' staff plus speakers plus media representatives equals total attendance. Venue staff, organiser staff and service provider staff sit outside both figures, so they belong in no published number at all.

That gives you two auditable populations from one file, with the difference between them itemised. Visitors is the smaller one. Total attendance is the larger one, and the gap is precisely your exhibitor staff, your speakers and your press, which is a far more useful thing to be able to say than a single corrected total.

What the standard will not do is decide the awkward middle. Accompanying persons and children appear in no exclusion list, and the only thing governing them is the requirement inside the visit definition that the person be representative of the market audience the organisers and exhibitors expected. That is a judgement, it belongs to you, and it has to be written down where somebody can find it.

Running the classification on one edition

Take the 32,700 badges from the opening scene and classify them.

Exhibitor staff, 4,900. Organiser staff and directly engaged contractors, 1,150. Media, 620. Speakers who did not also register as buyers, 310. Accompanying persons and children, 380. Third party service providers covering venue, catering, audiovisual and stand build, 240. That is 7,600.

Now build the ladder instead of running one subtraction.

Badges issued: 32,700.

Total attendance removes the two categories that belong in no published figure, organiser staff at 1,150 and service providers at 240. That leaves 31,310.

Visitors removes exhibitor staff, media and speakers from that, so it takes out a further 4,900, 620 and 310. That leaves 25,480.

Professional attendance, which is the base the industry series is built on, removes the 380 accompanying persons and children. That leaves 25,100, and it is the only step in the ladder the standard did not make for you.

Three publishable figures from one file: 32,700, 31,310 and 25,100. The spread between the top and the bottom is 7,600, which is 23.2 per cent of badges issued, and the headline audience figure is 76.8 per cent of the badges you printed.

That ratio is worth tracking on its own. If it is 76.8 per cent this year and 71.4 per cent next year, either your exhibitor staff allocation changed or your classification did, and you want to know which before somebody reads it as a fall in audience.

Where should the badge class be set?

The reason most organisers cannot produce that ladder is that the class is inferred afterwards, from the badge type name, by a person reading a list of forty badge type strings created by four different people across three editions.

Do it at the source instead. Every badge type in the registration system gets one mandatory attribute with a small controlled vocabulary. I would use six values: professional visitor, exhibitor staff, media, speaker, organiser or service provider, and accompanying person. Six is enough to reproduce every population the standard names, and small enough that the person setting up a new badge type in March can pick the right one without a meeting.

Two rules make it hold up. The attribute is required, so no badge type can be created without one, which stops the annual appearance of a VIP Guest type that nobody can classify in October. And the mapping is versioned, so when somebody reclassifies a type mid-cycle the change has a date and an author and the prior series can be rebuilt.

The awkward cases are worth deciding in advance, well before the reconciliation meeting. A speaker who works for a buying organisation and spends the rest of the day on the floor is genuinely doing two things, and I would classify on the badge that got scanned at the hall entrance. Somebody who registers as a visitor and later joins an exhibitor's stand team is exhibitor staff on the day, and the class should follow the badge that was scanned rather than the one first requested. A distributor exhibiting a manufacturer's products is exhibitor staff, and the manufacturer, if it has nobody of its own on site, is what UFI's standard calls a represented company, excluded from the exhibitor tally altogether.

Disclose the subtraction on the same page

The number that gets argued about is always the one whose derivation lives in a different document.

So put the ladder on the page. Not in an appendix, not in a definitions annexe, on the same page as the headline. Four lines of a table is enough: badges issued, total attendance, professional attendance, and the count removed at each step. Anybody who wants to know where the 7,600 went can see it without asking, and the sales director in the opening scene never sends that email. The wider evidence trail that makes a published figure a verified attendance figure is D10's, and this ladder is one row of it.

This also protects you in the other direction. A show that publishes only its professional attendance figure looks smaller than a competitor publishing badges issued, and there is nothing you can do about that except make the comparison legible. The CEIR Index, released by IAEE on 4 May 2026, measures the industry on professional attendance as its audience metric alongside net square feet, exhibiting company count and gross revenue. If you want to sit inside that series, professional attendance is the figure, and a show quoting badges issued against it is comparing itself to nothing.

Where this stops

The classification is only as good as the registration form, and registration forms are filled in by people who want to get to the confirmation screen.

The category that breaks first is exhibitor staff. Exhibitors routinely register team members as visitors, either to get around a badge allocation, which is where the ratio of exhibitor staff badges to booth size earns its keep as a pre-show check, or because whoever did the admin used the wrong link. Those badges will sit in your professional visitor population looking exactly like buyers, and the only reliable way to find them is to match the registration email domain against the exhibitor account list. That match is imperfect for the same reasons every company match is imperfect, and it will miss the agency staff and the freelancers entirely.

The second limit is that class is a property of a person on a day, and your data model probably treats it as a property of a badge for all time. Somebody who visits as a buyer on day one and works a partner stand on day three is genuinely both. If that population is small, pick the class of the first scan and note it. If it is large, which happens at association shows where members exhibit, your model needs the class on the scan rather than on the badge, and that is a schema change rather than a reporting fix. It also sits on top of the older question of what makes two rows one unique attendee, which is D6's.

The third is that no classification makes a badge into a buyer. A correctly excluded staff population still leaves you with a professional visitor count containing job seekers, students and competitors doing reconnaissance. The standard has no category for those and neither do you.

Export your badge type list from the last edition, put one classification column next to it, and count how many types you cannot classify without asking somebody. That count is the size of the disclosure problem, and it is usually between five and fifteen types at a large show.