How to audit a trade show marketing agency: the attendee qualification framework
The short version. Registrations are the one number that does not persuade an exhibitor. Every agency reports them because they are the easiest number to move and the hardest to dispute. Six tests separate the agencies that can attribute a registration from the agencies that can only count one — and the sixth test is the one nobody expects. |
Most organisers audit a trade show marketing agency by asking what it can do. That is the wrong question, and it is the question every agency is built to answer well.
The right question is narrower and much harder to perform: can you tell me where a registration came from, and will you tell me when the answer is bad. An agency that can answer both will usually be competent at everything else. An agency that cannot will produce a number every month that nobody can check, and the checking is the entire job.
This is the audit we would want run on us. It is published in that spirit, which means it includes the tests we come out of badly.
Why is this worth an afternoon of your time?
Because the cost of getting it wrong is not the fee. It is an edition.
A visitor acquisition engagement runs across a twelve-week window in which somewhere between forty and eighty optimisation decisions get made. Each one is a small bet placed on what the reported numbers say is working. If the measurement underneath those numbers is wrong — and in our experience of inheriting other people's measurement, it usually is wrong somewhere — then most of those bets are wrong, and the campaign's final number is a coincidence rather than an outcome.
You find out eleven months later, in the rebooking conversation, when the exhibitor who took a stand on the strength of last year's attendance figure asks what those visitors actually bought.
What does the industry data say about how much room for error there is?
Very little, and less than there was two years ago.
Freeman's 2025 research puts blended year-on-year attendee retention at 30 per cent, which means an organiser has to replace close to 70 per cent of the audience every year just to stand still. The same research finds that attendees who meet their objectives at a show are 85 per cent more likely to return — so the churn is not random, it is a consequence of who was in the hall and whether the show was built for them.
CEIR's Q2 2025 Index has the industry sitting 8.4 per cent below its Q2 2019 level, with exhibitors down 8.8 per cent and real revenues down 15.6 per cent against the same benchmark. The share of events surpassing their pre-pandemic performance fell from 39.6 per cent to 32.7 per cent in a single year. Recovery did not stall; it reversed.
Against that, the Gulf is the exception. UFI's industry statistics for 2024 make the Middle East the only region on earth where exhibition space rented has grown since 2019, at +0.9 per cent a year, while Europe fell 1.4 per cent and Asia-Pacific 0.9 per cent. UFI's July 2026 Barometer simultaneously finds 55 per cent of companies based inside the GCC reporting strong negative impact from the regional situation, against 14 per cent globally.
So the market you are buying marketing for is the one growing market in a shrinking industry, and it is also the most anxious one. That combination is exactly the condition under which an organiser overpays for volume and underbuys measurement.
The agency that cannot tell you where a registration came from is not a cheaper agency. It is a more expensive one, and the invoice arrives at rebooking.
What should you look for in a trade show marketing agency?
It is the principle that a registration's value is a function of who registered, not how many did — and therefore that every claim an agency makes about volume has to be reducible to a claim about composition.
An agency operating on that principle can tell you, for its last three shows: which buyer segments it set out to reach, which it deliberately excluded, what excluding them cost in headline volume, how many of the people who registered actually walked through the door, and what the exhibitors received afterwards that they could act on.
An agency not operating on that principle can tell you the registration total went up. The six tests below are how you find out which one is sitting across the table. Run them in this order. Each has a question, the shape of a good answer, the shape of a bad one, and — the part that matters — how to verify the answer rather than accept it.
Test one — the attribution test
“For your last three shows, what proportion of registrations could you trace to a named source at show close? Not modelled. Traced.”
What a good answer sounds like
A number, immediately, without a preamble. Then an unprompted account of where the untraceable share came from — direct traffic that was really mistagged email, a registration platform on a separate domain, walk-ins with no digital origin — and what they did about it.
A very good answer includes a figure lower than you expected and an explanation of why the lower figure is the honest one.
What a bad answer sounds like
“We use full-funnel attribution across all channels.” Any sentence in which a modelling vocabulary substitutes for a number. Any answer that begins by explaining why attribution is difficult before telling you what theirs was.
Also bad: a number close to 100 per cent, offered confidently. Nobody attributes 100 per cent of trade show registrations. An agency claiming it is either counting platform-reported conversions as traced, or has not looked.
How to verify it
Ask for the analytics property, read-only, for one past show. You are not auditing their work; you are checking four things that take ten minutes:
Does the conversion event fire on form submission, or on page load? On page load, every visitor to the confirmation page counts as a registration, including refreshes. This single fault inflates reported registrations and is the most common one we find.
Is there cross-domain measurement between the show site and the registration platform? If registration sits on a different domain and this is not configured, every registration is attributed to the show site as a referral and the true source is gone.
Do paid and organic collapse into a single channel anywhere? An untagged redirect is enough to do it, and it makes the paid report meaningless.
Was the property reconfigured mid-cycle? If so, the year-on-year comparison in the agency's deck is not a comparison.
If the agency declines read-only access to a completed show's analytics, that is the finding. There is no commercially sensitive information in a conversion-event configuration.
Test two — the segment test
“Which buyer segments did you deliberately exclude from the last campaign, and what did excluding them cost you in headline registration volume?”
What a good answer sounds like
Named exclusions and a quantified cost. “We cut students and general-public interest, which took roughly a fifth off the registration number and raised the procurement-authority share.”The agency should be slightly uncomfortable telling you this, because the number they gave the previous organiser was smaller as a result.
What a bad answer sounds like
“We target all relevant audiences.” An agency that has never excluded anyone has never done segment definition — it has done reach buying, which is a different service at a different price.
How to verify it
Ask to see the campaign structure for one past edition: the audience definitions, the exclusion lists, and the spend split across them. Then ask which segment had the highest cost per registration and why they kept funding it. The answer to that second question tells you whether anyone was actually reading the data during the campaign or only at the end of it.
This is where the original version of this framework used the phrase intent context binding, and the idea is sound: a search query or a content engagement reveals the buyer's state, and a campaign built on those signals reaches people in declared states rather than unknown ones. But the phrase is not the test. The test is whether they can show you who they turned away.
Test three — the show-up test
“What was registration-to-attendance on your last three editions?”
What a good answer sounds like
Three numbers, and an admission that one of them was poor. Followed by what they changed between editions and whether it worked.
What a bad answer sounds like
Confusion about why you are asking. Or the claim that attendance is the organiser's responsibility and registration is the agency's — which is the answer that tells you most, because it describes an agency that has drawn its own boundary at the exact point where its number stops being checkable.
How to verify it
Compare the registration figure in the agency's report against the badge-scan count from the registration provider for the same edition. These two numbers are held by different parties, which is what makes the comparison worth running.
The gap between them is the most recoverable number in the entire funnel and the industry quietly accepts a large one. An agency that has never measured the gap has been optimising toward a metric that does not describe anyone entering a building.
This is also the test to run on yourself before you run it on anyone else. If nobody at the organisation can produce registration-to-attendance for the last edition, the agency is not the only thing that needs auditing.
Test four — the exhibitor test
“What did the exhibitors receive within fourteen days of the last show closing, and what was in it?”
What a good answer sounds like
A per-exhibitor artefact, not a show-level one. Which buyer categories visited that stand, how many, what the sector composition was, and how it compared to the previous edition. Delivered inside two weeks, because that is the window in which an exhibitor's memory of the show is still doing the work.
What a bad answer sounds like
A show-level recap deck sent to everyone. Total attendance, some photographs, a thank-you, a save-the-date. This is the category default and it is why rebooking conversations start from zero every year.
How to verify it
Ask to see one. Redacted is fine. You are looking for whether the numbers in it are per-stand or per-show, and whether an exhibitor could take it to their own CFO as evidence.
Then ask the harder question: how many exhibitors opened it. An agency that produced the packs but cannot tell you the open rate produced them for the organiser's comfort rather than the exhibitor's decision.
Why this test carries more weight than it looks
Because this is where the organiser's revenue is. CEIR's 2026 Marketing Spend Decision Report finds exhibitions account for 40.8 per cent of overall marketing spend among participating exhibitors, with 28 per cent adding shows in 2026 and 75 per cent investing in digital channels alongside exhibiting. That is a buyer with budget, options, and an increasingly quantified alternative.
An exhibitor who receives a stand-level outcome pack has evidence for the renewal. An exhibitor who receives a recap deck has a memory, and memory loses to a spreadsheet every time.
Test five — the failure test
“Tell me about an edition that went badly. What happened, what did it cost, and what did you change afterwards?”
What a good answer sounds like
A specific edition, a specific mistake, a specific cost, and a specific change. The agency should get slightly less comfortable during this answer, and the answer should contain something that reflects poorly on them.
What a bad answer sounds like
A reframed success. “We had a client with unrealistic expectations.” “The show was in a difficult market.” Anything where the failure belongs to someone else. An agency with no owned failures has either not worked long enough or is not going to tell you when the current engagement starts going wrong — and the second is the risk you are actually buying.
How to verify it
You mostly cannot, and that is fine. This test works on the register of the answer rather than its content. But there is one check: ask for a reference from a client they no longer work with. An agency that can produce one has ended a relationship well, which is more informative than a reference from a client currently paying them.
The rule underneath this test
The reason we ask it is the rule we run internally: if a number cannot be attributed, we do not report it as ours. That rule costs us credit for work we have genuinely done, several times a year. An agency without an equivalent rule has no mechanism for telling you bad news, because every number is available to be claimed.
Test six — the handover test
“If we do not renew, what do we keep?”
What a good answer sounds like
A named list. The measurement configuration, documented. The audience definitions and exclusion lists. The channel performance history in a form that survives the account being closed. The campaign architecture written down. A playbook rather than a final report.
What a bad answer sounds like
Silence, followed by improvisation. Or a genuine offer to hand over assets that turn out to be a folder of creative files, which is the least valuable thing they hold.
Why this is the test nobody expects, and the one we would weight highest
Because it is the only test whose answer the agency has no incentive to have prepared.
An organiser who has run eight editions with three different agencies, and kept nothing transferable from any of them, has not run eight editions. They have run the first edition eight times. Each new agency rebuilds the measurement, rediscovers the segments, relearns which channels work for this sector in this market, and bills for the learning that the last agency already paid for out of the same organiser's budget.
The handover answer tells you whether you are buying a service or buying an asset. It is also, conveniently, the answer that predicts the other five, because an agency that documents for handover is an agency that documented in the first place.
How to run this on the agency you already have
Most people reading this are not choosing an agency. They have one, and something is not right.
The tests work unchanged, with three adjustments.
Run test one first and run it cold.
Do not announce an audit. Ask for read-only analytics access for a completed show as a routine request. What comes back, and how fast, is data.
Run test three against a number they have already given you.
Take last edition's reported registration figure from their own deck, put it next to the badge-scan count from the registration provider, and ask them to explain the gap. You are not looking for a small gap. You are looking for whether they already knew.
Replace test six with its harder form:
“Send me everything we would keep if we stopped tomorrow.” Give a deadline. An agency that has been documenting will send a folder within a week. An agency that has not will send a proposal.
TEST | AUDITING A PITCH | AUDITING THE INCUMBENT |
1 — Attribution | Ask for the number, then verify on a past show | Request read-only access without framing it as an audit |
2 — Segments | Ask what they excluded and what it cost | Ask which segment has the worst cost per registration and why it is still funded |
3 — Show-up | Ask for three editions | Put their reported figure next to the badge-scan count |
4 — Exhibitor packs | Ask to see one, redacted | Ask your own exhibitors what they received |
5 — Failure | Ask for an owned failure and a lapsed reference | Ask what they have told you that you did not want to hear |
6 — Handover | Ask what you keep | Ask them to send it, with a deadline |
What this audit will not tell you
It will not tell you whether the work will be any good creatively. Nothing in these six tests examines a film, a stand design, a delegate campaign or a piece of copy, and those things matter — a show's brand is the reason a visitor recognises it in year four.
It will not tell you whether you will enjoy working with them, which is a real variable across a twelve-week window in which someone has to be reachable at eleven at night in the week before doors.
It will not price the work. An honest range is a separate conversation and an agency that will not give you one before the proposal stage is telling you something, but this framework does not test for it.
And it will not tell you whether you need an agency at all — which is the next section, because it is the question the framework is most often used to avoid.
When you should not hire an agency
Three conditions, honestly stated, in which the answer is no.
When the show is under about eight thousand visitors and the sector is narrow.
At that scale the audience is often reachable through the association list, the trade press and the exhibitors' own databases. The marginal registration a paid campaign buys is expensive relative to a well-run partner and email programme. Fix the list before you buy the media.
When the measurement is broken and nobody internally owns it.
An agency can rebuild your measurement — we do it, and we do it before spending anything. But if no one inside the organisation will own it afterwards, it decays inside two editions and you will pay to rebuild it again. In that specific case, hiring an analytics contractor and an in-house marketer beats hiring us, and we have said so to prospects.
When what you actually need is a registration platform, not a marketing partner.
If your problem is that the registration flow leaks, or that badge data never reaches the sales team, the right purchase is technology and integration work. An acquisition agency pointed at a leaking registration path spends your money faster, not better.
The in-house case, made fairly: an internal team knows the exhibitors, sits in the rebooking conversations, and does not have to be re-briefed every year. What it usually lacks is the paid-media depth to run a five-stage campaign across four channels in eight markets, and the outside view that notices the measurement has been wrong for three editions. Portfolio organisers generally need both. A single-show organiser with a strong marketing manager often does not need us at all.
Running the six tests on us
It would be a poor instrument if we did not. Here are our answers, in the same order, including the two we would rather not publish.
Test one — attribution. We do not report a number we cannot attribute. That means our reported figures will often be lower than the platform’s headline numbers, but they are figures we can stand behind.
Test two — segments. We define segments before any spend and publish the exclusions to the client at the start. On a recent Riyadh industrial edition, the exclusions materially reduced the headline volume while improving the quality of the audience reaching the campaign.
Test three — show up. We measure registration to attendance against the registration provider’s scan count rather than our own reporting, because our own reporting is not independent of us. The number that matters is not how many people registered. It is how many actually showed up.
Test four — exhibitor packs. We produce per exhibitor outcome packs within fourteen days of close as a standard deliverable in stage three. The purpose is not to produce another presentation. It is to give each exhibitor a record of what the campaign and the show actually delivered.
Test five — failure. On the Automechanika Riyadh engagement we opened by refusing to run the campaign and spent the first three weeks rebuilding attribution before touching the media budget. That was correct. What we got wrong on that engagement is written up in the case study, at length, including what it cost.
Test six — handover. Stage five of our engagement is a documented playbook rather than a final report, and it is the stage clients are most likely to decline because it is the one with no visible deliverable during the show. We have not always insisted on it, and engagements that skipped it left the client exactly where this article says they should not be. We now price it into the edition rather than offering it separately.
Two more we will say without being asked. We are biased toward attribution as the primary axis because it is the axis we are strongest on. An organiser whose measurement is already clean should weight tests four and six higher and discount test one accordingly.
And we are a small bench across Bangalore and Dubai. That means we take a limited number of editions per season and are the wrong call for an organiser who needs a large number of shows covered simultaneously.
The one-page version
Print this. Take it into the meeting.
# | TEST | THE QUESTION | THE DISQUALIFIER |
1 | Attribution | What share of registrations was traced to a named source at close? | Modelling vocabulary instead of a number, or a figure near 100% |
2 | Segments | Who did you exclude, and what did it cost in volume? | “We target all relevant audiences” |
3 | Show-up | What was registration-to-attendance, last three editions? | Attendance framed as not their responsibility |
4 | Exhibitor packs | What did exhibitors get within 14 days, and who opened it? | A show-level recap deck sent to everyone |
5 | Failure | Which edition went badly, what did it cost, what changed? | A failure that belongs to someone else |
6 | Handover | If we don't renew, what do we keep? | Improvisation, or a folder of creative files |
Any agency that passes four of six is a serious candidate. An agency that fails test one and test six is selling you a service you will have to buy again from someone else in two years.
Start where we start
Before we take an engagement we run a paid diagnostic on the last edition — the same six tests, applied to your own numbers rather than an agency's, plus what your reported figures can and cannot support. You keep the report whether or not we work together, and if the finding is that you do not need an agency, that is what it will say.
Harry Aloysius is the founder of Kreate for Events, the trade show marketing practice of Kreative Clan Private Limited, and works on organiser-side visitor acquisition and exhibitor renewal across the GCC, Africa and South Asia. An earlier version of this framework was published by Swathi Nair in July 2026.
Sources. Freeman 2025 trends research, presented by Ken Holsinger, reported by Skift Meetings, 15 January 2026 · CEIR Q2 2025 Index Report, via IAEE · CEIR 2026 Marketing Spend Decision Report · UFI Global Exhibition Industry Statistics, data year 2024, published 20 May 2025 · UFI Global Exhibition Barometer, 37th edition, July 2026. Figures marked as ours are Kreate's own and are stated with their basis; where a figure is not yet auditable it is not stated.
Comments