Conference Registration Forecasting: Today's Count to a Signed Contract

Conference registration forecasting turns the registrations you hold today into the number you expect on the day the doors open.

The method that survives a real event is narrow: take the share of final registrations your own past editions had in hand at the same number of weeks out, divide today's count by that share, and report a range. The rest of this page is what you do with that range, because the forecast exists to be signed against.

Why are registrations arriving later?

A large share of registrations now arrives in the final weeks before the event, and the share that has arrived by any given week keeps moving between editions. The first fact makes the middle of a campaign look worse than it is. The second is what stops an older pacing model from correcting for it.

The Maritz Registration Insights Report analysed more than 360,000 registration records across 30 trade shows and found that in 2023, 45% of registrants signed up in the final 4 weeks before the event, more than a quarter in the final 2 weeks, and 9% on site (PCMA Convene). Read the sample before you read the numbers: those are registration records from trade shows, not a survey, and not association conferences. Use the shape as a warning and keep your own figures. Kyle Jordan, director of meetings at INFORMS, put the operating problem to Skift Meetings this way: "Our old registration pacing models are not as reliable as they used to be."

Less reliable still leaves you a curve to work with. What has changed is that the curve moves between editions, so a model anchored to one year's shape drifts further out of true every year it goes unchecked.

The specific failure is the calendar-date comparison. You hold 1,040 registrations on 3 October; last year on 3 October you held 1,180; the weekly report says you are 12% behind. If last year's registrants simply arrived a fortnight earlier, you are not behind at all, and the report has just handed your team a reason to spend money it does not need to spend. The same comparison indexed by weeks before the event removes the problem, because both years are then measured from the only date the registrant actually cares about.

Two things follow. Rebuild your pacing history as shares of a final total at a fixed number of weeks out, and drop calendar dates from the weekly review entirely. Then accept that a forecast taken at 8 weeks is wider than one taken at three, and report that width rather than rounding it away.

How do you forecast final registrations from today's count?

Start by turning each past edition into a set of shares. For every week you care about, divide the registrations that edition had in hand at that point by the count it finished with. Three editions give you 3 shares for each week, and the spread between them is the honest width of your forecast.

The largest share produces the lowest projection, because the share sits in the denominator. That inversion is the part teams get wrong in their first week, and it is the part that explains the late-registration problem: as registrants arrive later, the share in hand at any given week falls, so the same count today implies a larger final total.

Here is an association annual meeting, 8 weeks out, holding 1,040 registrations. Its last 3 editions had these shares of their final count in hand at the same point.

The three shares run 62%, 55%, 51%: the same drift the Maritz shape describes, visible in one organisation's own records. When the drift runs one way 3 editions in a row, the median is a conservative base and the top of the range is the more likely landing point, so say that in the report rather than leaving finance to infer it.

Now put the business plan from your event budget next to it. This event is budgeted at 2,100 registrations and covers its unavoidable costs at 1,750. The base forecast of 1,891 sits about 10% below plan, which is a revenue conversation. The bottom of the range, 1,677, sits below the point where the event pays for itself, which is a different and more serious conversation, and it is only visible because the forecast was reported as a range.

Three rules keep this arithmetic usable. Recompute every week, because the range narrows as the share in hand grows. Use at least 3 editions, since 2 shares give you a width and no sense of whether it is typical. And run the calculation separately for registration types that behave differently, such as member versus non-member or group bookings placed by one administrator, then add the projections rather than blending the rates. A first edition with no history gets no curve at all: plan against the widest range your contracts can survive and treat the year as the measurement.

Registration is only half of the headcount question, because a registration is not a person in a chair. The separate guide to event attendance rate and no-show rate covers the second multiplication, and its numbers are the ones to apply when a contract asks for bodies rather than sign-ups.

What signals belong in the forecast?

A signal belongs in the forecast when it meets 3 tests: it leads registration rather than following it, you can produce its value for the same week in past editions, and you can say in one sentence what a change in it means. Most numbers on a weekly marketing dashboard fail the second test, so the honest answer to "what else should we watch" is often "nothing yet, but start recording it now".

These usually pass, given a year of records:

• Net new registrations each week. A week that adds 40 when the curve expects 95 is the earliest reliable sign that the range needs re-reading.
• Room block pickup. People book a room once they have decided to travel, and with long approval cycles that often happens before they register.
• Registration starts that never finish. Abandoned registrations point at price, form length, or an approval the registrant cannot get, and each of those has a different fix.
• Group and delegation commitments still unplaced. Member organisations that hold seats every year are a pipeline you can name and call.
• Speaker, abstract, and exhibitor confirmations. A committee waits for a confirmed programme before it releases budget, and confirmations carry dates.

These do not belong in the number, however they are moving: social followers, email list size, impressions, last year's attendee satisfaction score, and the number of sessions on the agenda.

The rule for using them matters more than the list. Signals move where you plan inside the range; they do not create a number outside it. If room pickup and net new registrations both run ahead of last year at the same week, plan against the upper half of the range and write down that you did. If they run behind, plan against the lower half. A signal that pushes your commitment past either end of the range is telling you the pacing history is wrong, which is a reason to rebuild the curve rather than override it for one week.

Record the reading and the reason each week. By the third edition you will know which signals actually moved ahead of the count and which ones you were watching out of habit, and the post-event report is where that verdict belongs.

When do you have to decide on food, beverage, and room blocks?

You decide when the contract says you decide, which is why this section starts with the contracts. Read each agreement and list every date that converts a number into money: the hotel cutoff after which unsold rooms return to the hotel, the attrition threshold that charges you for the block you did not fill, the food and beverage guarantee that sets the minimum you pay whether or not the meals are eaten, and the print, badge, and staffing orders that follow the same headcount.

Carry the example through. At an 84% attendance rate, the forecast of 1,891 with a range of 1,677 to 2,039 becomes an expected 1,588 people on site, with a range of 1,409 to 1,713. Lunch take-up in this organisation runs at 80% of the people on site, so the guarantee is 1,270 covers at the base and 1,127 at the low end. At $95 a cover, the two answers are $13,585 apart, and the difference is paid for food nobody eats.

The room block works the same way. The contract blocks 900 room nights with attrition at 80%, so 720 nights have to be picked up. History says this audience books about 0.45 room nights per person on site, which gives 715 nights at the base forecast and 634 at the low end. Even the base lands 5 nights short of the threshold. The low end misses it by 86, and at $85 a night that is $7,310. Release 100 nights before the cutoff date and the threshold falls to 640, which the base clears and which turns the same low-end outcome into a $510 miss.

Guaranteeing at the base and leaving the block at 900 therefore costs roughly $20,900 across those two contracts if the low end arrives, and the forecast had the low end on the page from week eight. Sign against the bottom of the range and add back later: hotels will usually take rooms into the block again, and a caterer will take a guarantee up more readily than down.

When the forecast comes in below plan, work in this order. First check the curve before you check the marketing, because a shortfall measured against last year's calendar dates is the most common false alarm here and it costs nothing to rule out. Second, reduce exposure at the next contractual date: release rooms, lower the guarantee, take the smaller room, and confirm in writing what it costs to add capacity back. Third, go after demand you can name, which means lapsed registrants from the last 2 editions, listed by an attendee retention cohort, group holds that were never placed, and the committee members whose organisations send delegations every year. Fourth, tell sponsors and exhibitors while the number is still a forecast, because the audience figure in their agreement is a commitment you made.

Discounting sits last on that list deliberately. It lowers revenue per registrant at the same moment it raises the count, so an event 10% short on registrations can finish further from its budget after a discount than before it. Model it as 2 numbers, registrations and revenue, and check that the pair still clears the threshold your event ROI case is built on.

Give finance the same 4 items every week: the base, the range, the date the range narrows, and the next contractual decision it feeds. Then the weekly review is about which commitment to move, and the argument over whether the number is right happens once, when you build the curve.

Where EventIQ fits

EventIQ replaces nothing. It connects on top of the platforms you already run: registration and ticketing (Cvent, Zoom, Swapcard, StubHub), CRM (Salesforce, HubSpot, GoHighLevel), and marketing (Google Ads, Meta Ads, LinkedIn Ads, Mailchimp, Google Analytics). Platforms with an API outside that list are connected on request.

Registrations and attendance stay separate records. Contacts are matched by exact email, and every record keeps its result: matched, unmatched, or no email. Marketing spend sits by event and channel, with the author of every change. Records sync on a schedule: Swapcard every 15 minutes, Cvent every 30, Zoom every 2 hours, Salesforce every 4. The Event Dashboard and the Portfolio Dashboard show them in one view.

Registrations and check-ins arrive as separate records, so the pacing shares in this article are a division you can run on them. The forecast is visible before the event; ask any vendor, including us, what it is built on. See a pacing curve against contract dates on sample data in a 20-minute demo.