Every event ends the same way. Someone opens last year's slide template, drops in the attendance number, pulls a revenue figure from the CRM, divides one by the other, and sends it to whoever approved the budget.

The number almost always looks acceptable. That's the problem.

This used to be a low-stakes exercise. It isn't anymore. ASAE's first State of Associations report found nearly 39% of association CEOs reporting a financial decline against about 10% reporting improvement, and named meetings as the revenue stream taking the hardest hit, driven by falling attendance including reduced international participation. When a line item is under that kind of pressure, the report you write about it stops being a formality. It becomes the document that decides whether the event runs again.

So it's worth asking what those reports actually measure. Event ROI reports rarely fail because the arithmetic is wrong. They fail because the inputs get chosen after the conclusion is already settled.

Here's what breaks, and what's worth tracking instead.

Four ways event ROI reports go wrong

1. The conclusion arrives before the math. The most common failure isn't a bad formula. It's a report assembled in reverse: the team knows the event needs to clear a bar, so they choose the revenue definition, the attribution model, and the cost line items that get them there. Nobody involved thinks of this as dishonest. It feels like judgment. But a metric chosen because of the answer it produces isn't a measurement, and the report loses the ability to tell you anything you didn't already believe.

2. Registrations get counted as attendance. Our Event Data Lab analysis of check-in data across 1,070+ live events put the median no-show rate at about 20%. Every per-attendee figure calculated off the registration count inherits that error. What makes it worse is that the error isn't constant: free events ran a median no-show rate near 28% against about 17% for paid events, and events with 10 to 49 attendees came in around 32%. So if your portfolio mix shifts between free and paid, or between small and large, your year-over-year cost-per-attendee trend moves for reasons that have nothing to do with performance.

3. The attribution window moves between reports. A 30-day window for the deck due next week. A 180-day window when the first number disappointed. Both windows are defensible in isolation. Switching between them retroactively means you no longer have a trend line, you have a series of unrelated snapshots. Pick the window before the event, write it down, and keep it for at least four event cycles.

4. The report answers a question the event wasn't run to answer. This is the one that hits associations hardest. Most event ROI templates in circulation were built for corporate demand generation, where the output is pipeline. If your annual conference exists to deliver certification hours, retain members, and generate non-dues revenue, a pipeline-shaped report will make it look like a failure regardless of how it performed. Given the funding pressure on meetings described above, measuring them with a borrowed framework isn't just imprecise. It's a risk to whether they get funded at all.

Five event ROI measurement inputs that hold up

1. Verified attendance, plus the gap. Track scanned check-ins, not registrations, and report the difference between the two as its own line. The gap is the most useful diagnostic in the whole report. A widening gap tells you your registration promise and your event experience are drifting apart, which is a fixable problem you'd never see in a blended ROI figure. Platforms that handle registration and on-site check-in against the same attendee record, PheedLoop included, produce this comparison without a reconciliation exercise, because both records live against the same attendee.

2. Engagement depth per attendee, not engagement totals. Total session scans and total app logins rise when attendance rises, which makes them close to useless as performance signals. You can post record engagement totals on a worse event. What you want is the distribution: how many attendees hit three or more meaningful touchpoints, against how many registered, walked in, and left. Pick the threshold yourself, write it down, and hold it constant across events. The share of attendees clearing it is a number that only moves for real reasons, and it's a better leading indicator of next year than any revenue figure you can produce in week two.

3. Meetings held, logged before the event ends. For corporate teams, this is the input that fails most often, and it fails operationally rather than analytically. Booth conversations that never make it into the CRM are invisible by Monday. Set the expectation that leads and meetings get logged on-site, with lead retrieval tooling that works off a badge scan rather than a laptop, and the attribution problem shrinks considerably. Retrospective CRM archaeology two weeks later is not measurement.

4. First-time attendee return rate. Take everyone who attended for the first time this year and check how many come back next year. It's a lagging metric, which is why almost nobody tracks it, and it's also the closest single number to "did this event actually work." It applies cleanly to associations, trade shows, and corporate user conferences alike, and it can't be manipulated by changing an attribution window.

5. Cost per verified engaged attendee. Divide fully loaded event cost by the number of attendees who cleared your engagement threshold from point two. The number will be considerably worse than your cost per registration, and it's the one worth putting in front of finance. It reflects what you actually paid to produce the outcomes you care about, and it gives you a lever: you can improve it by spending less or by engaging more of the people who showed up, and both of those are real strategies.

What an honest report looks like

Shorter, less flattering, and more useful than the version most teams send.

Three inputs, fixed before the event: the attribution window, the engagement threshold, and the cost lines that count. Then verified attendance against registrations, the engagement distribution, and cost per engaged attendee. Revenue attribution still belongs in there, but as one input among several rather than the headline.

The reason to do it this way isn't rigour for its own sake. It's that a report built to justify the event can only ever tell you the event was justified. A report built to measure it will occasionally tell you something you can act on, which is the only reason to write one.

If you want the underlying numbers behind the show rate figure above, Event Data Lab Report #05 breaks the no-show data out by event size and pricing model.

‍