Ask most event teams about last-minute registrations and you will hear the same thing: people who sign up the week of an event are the ones who do not turn up. It is a widely held belief, and it drives real decisions. Registration deadlines get set early. Late signups get discounted in headcount forecasts. Some teams close registration entirely to filter for commitment.

Report #05 established no-show benchmarks and Report #06 tested them against registration friction. In this report we test them against registration timing, using 216,282 registrations across 467 events where both registration and check-in timestamps are available.

The belief looks correct until you control for the event.

Executive Summary

• Pooled across all events, registration timing appears to predict attendance strongly. Same-week registrants no-show at 42.4% against 18.4% for those who registered six or more months out.
• Within individual events, the relationship disappears. Comparing early and late registrants at the same event, the median difference in attendance is 0.6 percentage points, and late registrants attend more reliably at 44% of events.
• The apparent effect is composition. Free events attract registrations a median 25 days later than paid events and attend far worse at every lead-time band. That mix, not individual timing, produces the gradient.

Dataset Overview

Dataset overview

• 467 live events with both registration and check-in data
• 216,282 completed registrations with a usable registration timestamp
• Registration lead time calculated as event date minus registration timestamp
• Median registration lead time: 35 days (25th percentile 13 days, 75th percentile 68 days)
• Events required to have 50 or more completed registrations and a check-in-to-registration link rate above 90%
• Registrations dated after their event date were excluded (11,582 records, 5.1%)
• Test, sandbox, and internal events were excluded
• Data aggregated and anonymised across live events

Metric definitions

Registration lead time is the number of days between a registration completing and the event start date.

Attendance is whether a completed registration produced at least one event check-in. No-show rate is its inverse.

What the Data Shows

Pooled, the Belief Looks Confirmed

Across all 467 events, attendance rises monotonically with lead time.

Attendance by registration lead time

A 24 percentage point spread, with no reversals. Read on its own, this is a strong endorsement of the conventional wisdom, and it is the number most likely to be quoted from this report.

It is also misleading.

Within the Same Event, the Effect Vanishes

The pooled table compares registrations across different events. The more useful comparison holds the event constant: at a single event, do the people who registered early attend more reliably than the people who registered late?

Across 374 events with at least 20 registrations on each side of a 30-day cutoff:

• Registered 30 or more days out: 84.1% median attendance
• Registered inside 30 days: 80.8% median attendance
• Median within-event difference: 0.6 percentage points

Late registrants attend more reliably at 44% of events. At 55% of events the gap is smaller than 5 percentage points in either direction. The distribution of the gap runs from minus 5.0 points at the 25th percentile to plus 3.4 at the 75th.

The finding holds across every variation we tested. Moving the cutoff to 14, 60, or 90 days, requiring 50 registrations per side, or restricting to ticketed registrations only all produce median gaps between minus 0.6 and plus 0.8 percentage points.

The Gradient Is Made of Free Events

Two facts explain the pooled result.

Free events register later. Median lead time at events with no paid registrations is 16 days, against 41 days at paid events. 69% of free-event registrations arrive inside 30 days, against 38% at paid events.

Free events attend worse at every lead time. Pooled within free events, attendance runs from 28.2% for same-week registrants to 59.3% at two to three months. Pooled within paid events, the same bands run 72.8% to 81.6%.

So the same-week band is disproportionately made of free-event registrations, which attend poorly regardless of when they arrive. The six-month band is almost entirely paid registrations, which attend well regardless of when they arrive. The gradient is the mix changing, not behaviour changing.

The residual gradient inside paid events is small and not even monotonic: same-week registrants attend at 72.8%, those registering two to four weeks out attend at 67.9%, and the six-month band reaches 81.6%. When that group is compared within events rather than pooled, the difference is minus 0.7 percentage points.

Key insight: Registration timing does not predict attendance. It predicts what kind of event someone is registering for. Late registrations cluster at free events, and free events have a no-show problem that has nothing to do with when people signed up.

Practical Implications for Event Teams

• Do not discount late registrations in your headcount forecast. At your event, someone who registers the week of is about as likely to attend as someone who registered three months ago.
• Closing registration early will not improve your show rate. It will reduce your registration count. The commitment filter people expect from an early deadline is not visible in this data.
• If you run free events, the lever is pricing or commitment mechanisms, not timing. Free events lose 40% to 70% of registrants across every lead-time band. A registration deadline does not address that.
• Use lead-time benchmarks for capacity planning, not for quality scoring. Knowing that 14% of registrations arrive in the final week is genuinely useful for staffing a check-in desk. Treating those registrants as lower quality is not supported.
• When you see a strong pattern in your own registration data, check whether it survives holding the event constant. Patterns that look dramatic across a portfolio frequently dissolve inside a single event.

A Note on Method

This is the third time in this series that a strong relationship has disappeared under event-level control. Report #06 found that registration completion appeared to predict no-shows until pricing was accounted for. A follow-up analysis found the same for ticket class, where a large between-event gap shrank to a few percentage points within events. Registration timing now makes three.

The pattern is consistent enough to be worth stating as a general caution. Event portfolios contain very different kinds of events, and any variable that correlates with event type will appear to predict outcomes that it does not cause. Free versus paid is the dominant such variable in registration and attendance data, and it has confounded every relationship we have examined so far.

Download the Full Report

Download the full Event Data Lab report

Get the complete lead-time distributions, the within-event comparison across all cutoff variants, the pooled decomposition by event pricing, and detailed methodology notes.

More from the Event Data Lab

• Report #05: One in five registered attendees won't show up. No-show benchmarks by event size and pricing model
• Report #06: Does easier registration lead to more no-shows?. Registration friction tested against attendance, plus a correction note
• Report #08: Attendees attend 5.5 sessions, and one in four attend none. Session attendance, programme scope, and what doesn't move attendance

This report is part of the Event Data Lab, an ongoing research initiative analysing real-world event performance across registration, onsite operations, engagement, and ROI.