Cancellations are not sudden
Members almost never decide to quit on the day they tell you. The decision is made over weeks, in behaviour, and it shows up in your data long before it shows up in your inbox. Two to four weeks of warning is typical, if anyone is looking.
The trick is watching for changes in behaviour rather than waiting for a request.
1. Attendance velocity
Not attendance — the rate at which it is falling. A member who went from four visits a week to two is showing you something, even though they are still walking through the door.
| Pattern | Risk |
|---|---|
| Consistent attendance | Low |
| 25% decline in visits | Medium |
| 50%+ decline in visits | High |
| No visits in 14+ days | Critical |
Reach out within a week of noticing the pattern. Do not wait for them to disappear completely.
2. Missed class bookings
The member books and does not show. That is intention without follow-through — they wanted to come and something stopped them. That something tends to become permanent. Two missed bookings in a row is the line worth watching.
Keep the follow-up casual: "Missed you Thursday — everything all right?" is a check-in. "We noticed your engagement has declined" is a form letter.
3. Payment retry patterns
Sometimes a decline is just an expired card. Repeated failures — or a member who does not bother updating their details — usually means they have already checked out mentally.
| Scenario | Risk |
|---|---|
| First decline, updated within 48 hours | Low |
| First decline, no update in 7 days | Medium |
| Multiple declines, no response | High |
4. Contract expiry timing
Members who intend to leave often reduce engagement 30–60 days before their contract ends. They are mentally preparing. Have the renewal conversation at 45 days, not at seven.
5. The engagement gap
Compare a member against their own history rather than against the average. Someone who used to open every email, book every class and buy from the shop — and now does none of it — has disengaged, even if their raw numbers still look acceptable.
- Email open rate versus their own baseline
- Class booking frequency
- App or kiosk usage
- Additional purchases
6. Tenure risk windows
| Period | Risk | Why |
|---|---|---|
| Month 1–3 | Highest | Habit not formed yet |
| Six weeks after January | High | Resolution motivation fades |
| Summer | Medium | Holidays and outdoor training |
| Contract anniversary | High | A natural decision point |
Increase touchpoints inside the high-risk windows, and do not assume long-tenure members are immune at renewal.
Turning signals into a queue
Six separate signals are six separate things to remember. Weighted into one score they become a single ordered list of who to call today.
| Signal | Suggested weight |
|---|---|
| Attendance decline | 30% |
| Days since last visit | 25% |
| Payment issues | 20% |
| Contract status | 15% |
| Engagement metrics | 10% |
- 0–30 — low risk, routine check-in
- 31–60 — medium risk, proactive outreach
- 61–80 — high risk, personal call
- 81–100 — critical, same-day intervention
A score like this is a sorting device, not a prophecy. It is only worth having if you can click into it and see the four facts it was built from.
What to actually say
- Attendance drop
- "Noticed we haven't seen you as much lately — everything okay? Happy to hold your spot in Thursday's class."
- Missed booking
- "Missed you in class yesterday. Want me to book you into the next one?"
- Payment issue
- "Looks like there was a problem with your payment — want me to help you update it so nothing lapses?"
- Contract expiring
- "Your membership is up for renewal soon. Before it is, how has everything been going?"
Spreadsheets versus software
| Approach | Works when | Breaks when |
|---|---|---|
| Manual spreadsheet | Under ~100 members and one person owns it | Anyone takes a holiday |
| Standard software reports | You want to know what already happened | You want to know who is drifting now |
| A scored, daily queue | The signals are computed from real records | The score cannot be traced back to its inputs |
Key takeaways
Members signal before they cancel
Typically two to four weeks of warning.
Attendance velocity matters most
Watch the rate of change, not the raw count.
Intervene at the first sign
Not after weeks of absence.
Match the message to the signal
A payment problem and a motivation problem need different calls.
Rank, then work the list
A single ordered queue beats six dashboards nobody opens.