Attendee retention is the share of one edition's attendees who attend the next one. Freeman's
end-of-year trends report puts blended industry retention barely above 30% year over year, which means a
conference drawing 1,000 people turns over most of the room each time it runs.

The number worth
planning against is your own, and it comes from joining the attendee lists you already hold,
one edition to the next, on exact email.

## What share of attendees come back?

Retention measures the audience you had, one edition later.

**Attendee retention (%) = Attendees this edition who also attended the previous one ÷ Attendees at the previous edition × 100**

The denominator is the earlier edition's audience, and that is the half teams most often get
wrong. A second question, what portion of the people in front of you have been here before, has
its own calculation and its own name.

**Repeat share (%) = Attendees this edition who attended any earlier edition ÷ Attendees this edition × 100**

An event that grows can raise its repeat share and lose retention in the same year, because a
larger intake of first-timers changes one denominator and not the other. Report both, and label
which is which every time.

Freeman's End-of-Year Trends Recap puts blended year-over-year attendee retention "barely above
30%", and the most generous data Freeman found in the low 40s
([reported by Trade Show Executive, January 2026](https://tradeshowexecutive.com/freemans-end-of-year-trends-recap-emphasizes-the-importance-of-retention-in-2026/)).
Freeman is a services supplier to the market it measures, and the methodology behind those figures
is not published: no sample size, no sector split, no field dates. Treat it as the order of
magnitude to expect, not as a benchmark your board should hold you to.

Your own rate moves with things that average cannot see. A rotating host city changes who can
travel. A biennial event breaks the year-over-year comparison entirely, and an association
conference selling to a member base behaves differently from an open trade show.

Retention is also not the attendance rate. The attendance rate asks how many of the people who
registered for this event walked in; retention asks how many of the people who walked in last
time came back. An event can hold 85% attendance every year and still turn over two thirds of
its audience, and the two numbers get fixed by different work. The guide to
[event attendance rate and no-shows](https://eventiq.io/md/blog/event-attendance-rate) covers the first of those.

## How do you build a return cohort?

A cohort is the set of people who attended one edition, followed forward. Building it needs no
tooling beyond the exports you can already pull.

1. Export attendees, not registrants, for each of the last 3 editions. Keep email, name,
   organization, registration type, registration date, and the check-in result. If the oldest
   edition only survives as a registration list, keep it as a separate series and say so on the
   chart, because mixing the two inflates every rate below.
2. Normalize the key and nothing else. Lowercase the email, trim the spaces, match exactly. Do
   not repair addresses by hand or by pattern.
3. Publish the count of attendees with no email address next to the rate, and leave them out of
   the cohort. They are a known hole in the denominator, and guessing fills it wrongly.
4. Label every person in every edition as first time, returned from the previous edition, or
   returned after a gap.
5. Count before you interpret. The tables come out of step 4 mechanically.

Exact email is a strict key and it carries one known bias: a person who changed employers
arrives with a new address and counts as a first-timer. That pushes the measured rate below the
real one by an amount the records cannot show. Report retention as a floor with that caveat
attached, in the same wording every year, so the series stays comparable to itself.

Counted this way, a conference across 3 editions looks like the table below. The numbers
are illustrative, chosen so the arithmetic closes.
| Edition | Attendees | Returned from the previous edition | Returned after skipping an edition | First time |
| --- | --- | --- | --- | --- |
| 2024 | 1,000 | baseline | baseline | 1,000 |
| 2025 | 1,100 | 340 | not yet possible | 760 |
| 2026 | 1,150 | 385 | 60 | 705 |
Retention from 2024 to 2025 is 340 ÷ 1,000 = 34.0%. From 2025 to 2026 it is 385 ÷ 1,100 =
35.0%. The repeat share in 2026 is (385 + 60) ÷ 1,150 = 38.7%, and the 2026 room splits 33.5%
returning from 2025, 5.2% back after a gap, and 61.3% first time.

Following the same 1,000 people from 2024 forward gives the view an edition-by-edition table
misses.
| What the 1,000 attendees of 2024 did next | People | Share of the cohort |
| --- | --- | --- |
| Attended 2025 and 2026 | 170 | 17.0% |
| Attended 2025, not 2026 | 170 | 17.0% |
| Skipped 2025, came back in 2026 | 60 | 6.0% |
| Attended neither | 600 | 60.0% |
| Total | 1,000 | 100% |
Two editions later, 230 of the original 1,000 are still in the room: 23.0% of the cohort. That
is the figure to put in front of a board, because it says what a year's audience is worth over
time.

The reason sits in the next table, which splits the 2025 audience by its own history and
follows each part into 2026.
| Group within the 2025 audience | People in 2025 | Returned in 2026 | Return rate |
| --- | --- | --- | --- |
| Had attended 2024 as well | 340 | 170 | 50.0% |
| First time in 2025 | 760 | 215 | 28.3% |
| All 2025 attendees | 1,100 | 385 | 35.0% |
Someone who has attended twice comes back at roughly double the rate of someone who has
attended once. The blended 35.0% hides that split, and the split is the part you can act on:
the audience is lost between the first edition and the second, which puts the money that moves
retention on first-timers in their first year. Separately, of the 660 people from the 2024
cohort who did not come in 2025, 60 returned in 2026, a 9.1% recovery rate. A lapsed attendee
is not gone, but they come back at a quarter of the rate of someone still in the habit.

## Who did not come back, and can you tell why?

The lapsed list is the cohort minus the returners, and the same query that produced the rate
produced it: 660 people after 2024, 715 after 2025. Teams that never build it end up discussing
retention as a percentage when it could be a list of names somebody can call.

Segment it with the fields already on the record, in this order:

- **By organization.** An account that sent 12 people and now sends none made one decision, not
  twelve. This is the fastest read in the exercise and the only one that routes to a named
  person for a phone call.
- **By first-time or repeat.** The table above says these are different populations. Reporting
  them together hides the group that is actually leaking.
- **By registration type.** Member, non-member, comp, exhibitor staff, and speaker each renew
  on different logic, and a comp who does not return costs nothing to have lost.
- **By region or distance.** If the host city moved, this column usually explains more of the
  drop than the program did.
- **By registration lead time.** People who registered in the last 2 weeks behave less like
  committed attendees the following year.

The records tell you who and when, never why. They will show a drop
concentrated among non-members after a price change, which is a correlation you can put in a
sentence. They will not show the travel freeze, the reorganization, or the person who found the
program thin and said nothing. The only route to why is asking the people who did not come, and
a survey aimed at non-attendees is a different instrument from the one that reaches the room
afterwards. The guide to [post-event survey questions](https://eventiq.io/md/blog/post-event-survey-questions)
covers how that one is built.

Some of the loss is invisible by construction. A person who left the industry, retired, or
changed employers and email address is counted as lapsed whether they chose to leave or not.
Say that in the same breath as the number, every year, so the rate is not read as a verdict on
the program.

## What can registration records show about non-attendees?

"Non-attendee" covers 4 groups that share nothing except absence, and the work starts by
keeping them apart.
| Group | How the records identify them | What the records can say |
| --- | --- | --- |
| Registered, never checked in | A registration with no check-in result | The gap between the list and the room this year. This is the attendance rate question, not retention |
| Cancelled or refunded | Registration status, with its date | Demand that existed and was withdrawn, and how late the withdrawal came |
| Lapsed attendees | Attended an earlier edition, no registration this year | The largest reachable group, and the only one with a history to segment |
| Never attended, already on your lists | Present in your own membership or contact records, no registration across 3 editions | The size of the audience you hold and have not converted once |
The first row causes the most trouble. Counting someone who registered but never walked in as
retained raises the rate without a single extra person in the room, and two teams reporting
different retention for the same event usually differ exactly there. Use the check-in result as
the cohort key wherever you hold it, and where you do not, label the series as
registration-based and keep it separate.

The third row does 2 jobs. As a retention diagnostic it tells you what you lost and among
whom. As a demand list it is a named source of registrations for the next edition, and it
belongs on its own line in the registration forecast, where a general marketing assumption
would otherwise absorb it. [Conference registration forecasting](https://eventiq.io/md/blog/conference-registration-forecasting)
is its own method, and the lapsed list is one of its named inputs.

What the records cannot say matters as much. They do not show that the person attended a
different event that month, that a budget was cut, that they are still in the field at all, or
whether the absence was even their decision. None of that should be inferred from a row that is
simply missing.

Put retention, repeat share, and the lapsed count into the same
[post-event report](https://eventiq.io/md/blog/post-event-report-template) as the attendance figures, with the
definitions written out in full next to them. A retention series is only worth having if the
rule that produced year one still produces year four, and the fastest way to lose 4 years of
comparability is to change the denominator quietly. When the series is stable, it feeds the
[event ROI](https://eventiq.io/md/event-roi) case directly: the cost of holding an attendee and the cost of finding
a new one are different numbers, and until retention is measured, only one of them is visible.

## 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, which is what lets the cohorts in this
article stand on check-ins instead of sign-ups. Contacts are matched by exact email, and
every record keeps its result: matched, unmatched, or no email. 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.

Ask any vendor, including us, which of these it holds in the product rather than on a slide:
one person counted once across sources, the full cost of the event, and a total in a single
currency. The cohort method on this page is the one you run on top of those records. See how
editions join on a sample attendee list in a 20-minute demo.

---

HTML version: https://eventiq.io/blog/attendee-retention