Australian Gym Churn Case Study: 37 Days From Last Door Scan to Recorded Cancellation

By Zack Leaudais · Founder, Zalio & GymFactories · 11 August 2026

I operate the gym studied here. This case study asks a simple question: by the time a membership cancellation is recorded, how long has the member already been gone? We measured it with approximately 18 months of door-scan and contract data from one ~700-direct-member Australian gym — a case study of a single club (n = 1), published with full methods and limits because we could not identify a directly comparable public Australian estimate of this interval as of August 2026.

Key findings (10 August 2026, GymFactories Research)

In this one ~700-direct-member Australian gym, the median gap between a member's last recorded door scan and their recorded cancellation was 37 days (36.8 days; middle 50% of cases: 13–75 days; n = 625 termination events with an observed prior visit, end-date window March 2025 – July 2026).
62.5% of the 682 qualifying termination events (426/682) had no recorded visit in the 28 days before the recorded cancellation; 75.2% (513/682) had none in the prior 14 days.
Only 17% (107/625) recorded a visit in the final week before the recorded cancellation.
Among terminations with at least six months' tenure, mean attendance fell from 2.04 visits per week 26 weeks out to 0.47 in the final week, first dropping below one visit per week five weeks before the recorded cancellation (n = 277–283 events per week).
7.3% of qualifying terminations (50/682) recorded no door scan at all during the membership episode (memberships that began within the observed period).

682 termination events correspond to 664 people; recalculating on each person's latest event gives a 37.3-day median — the unit of analysis barely moves the answer.

How long before recorded cancellation was the last gym visit?

In this club: a median of 37 days, measured across 625 termination events with an observed prior visit. A quarter of terminations show a last recorded visit 75+ days before the recorded cancellation; one in ten, 139+ days.

Mean visits per week by weeks before recorded cancellation — 26-week tenure-eligible panel, n≈280 termination events per week.

Mean attendance declined overall across the six months before cancellation, with week-to-week fluctuation. The weekly median first reached zero 18 weeks before the recorded cancellation and was zero in 16 of the final 18 weekly bins — a positive mean alongside a zero median means at least half of eventual cancellers recorded no visit in a given week while a minority still attended.

As a non-headline reference stratum: prepaid fixed-term memberships (excluded from the main cohort) had a 14.3-day median gap to recorded cancellation among 41 events with an observed visit. Those contracts are excluded because fixed-term expiries are structurally different from recurring direct-membership terminations; the comparison does not establish why either group ended.

What share of termination events had no visit in the prior 28 days?

In this one ~700-direct-member Australian gym, 62.5% (426/682) of qualifying termination events had no recorded visit in the 28 days before the recorded cancellation, and 75.2% (513/682) had none in the prior 14 days. Restricting to events with a complete observation window barely changes this: 63.1% (411/651) and 75.4% (505/670) respectively.

Can gym attendance predict which members will cancel?

This study cannot answer that, and does not claim to. It looks backwards from known cancellations only. Saying how many currently absent members will go on to cancel requires comparing them against members whose attendance lapses and who nonetheless stay — a matched-control design this case study deliberately does not include. What it does establish is narrower: at this club, most recorded cancellations were preceded by weeks of absence that was visible in the door-scan data.

How we measured it

Data: door-scan entry logs (~147,000 visits) and membership contracts from the club's management system, analysed in a single database snapshot on 10 August 2026 under a prespecified protocol frozen before analysis. Aggregates only — no member-level data left the analysis environment.
Who counts as a “direct member”: adult recurring gym/boxing memberships only. All 59 membership plans referenced by primary contracts were classified before the analysis ran. Excluded: third-party schemes such as Fitness Passport (not direct recurring memberships), free and $1 trials, prepaid upfront terms, under-16 memberships, staff and complimentary access.
What counts as a termination: a membership episode (chained contracts, so a plan change or immediate re-sign is not churn) whose final contract ended within the March 2025 – July 2026 end-date window, with no continuation within 30 days. 682 events across 664 people qualified.
The measurement: days between the last recorded door scan attributable to the membership episode and the date the cancellation was recorded in the club system.
Stability: across the prespecified sensitivity analyses (alternative anchor date, per-person weighting, episode-gap definitions), median gaps ranged from 33.8 to 38.2 days; leave-one-month-out estimates ranged from 35.3 to 37.8 days. Monthly medians varied more widely — from 11.9 to 59.2 days — and drifted upward during 2026; the cause is unknown and is flagged for the planned re-run.

What this is not

Not a benchmark. One club, one region, one format mix. Neither the 37-day estimate nor the shape of the attendance curve should be treated as an industry figure; whether either replicates elsewhere is unknown.
Not a prediction model (see above) and not a causal claim — nothing here shows that acting on absence prevents cancellation.
The cancellation field is an operator-recorded timestamp, not a validated member-decision or notification time. Lagging, backdating, or administrative entry practices could change the measured interval; this study does not assert the direction of that bias.
Voluntary cancellations and payment-failure cancellations cannot be separated in this dataset.
Attendance is measured by door scans at one site, read from a sync copy of the club system; visit history begins 31 January 2025, and seven zero-visit events predating that floor were excluded from the gap analysis.

About the author & disclosure

Zack Leaudais operates the gym studied here and is the founder of Zalio, gym-management software, and GymFactories. Zalio is piloted at the club. The club is not named on this page and approved publication; readers familiar with the author's gym interests may nevertheless infer its identity. Published results are aggregate only and contain no member names, identifiers, row-level records, or plan-level cells. Self-funded and not peer-reviewed. This page publishes the aggregate findings and a methods summary; the prespecified protocol and analysis queries are available on request. Methodology questions and corrections: hello@zalio.ai.

Suggested citation: Leaudais, Z. / GymFactories Research (2026). Australian gym churn case study: last door scan to recorded cancellation at a ~700-direct-member gym. http://localhost:3064/gym-member-churn-study

More gym-business data: Australian fitness industry statistics · what it costs to open a gym in Australia · how we label sources: /methodology.