Every gym owner I have met can tell you their member count. Far fewer can tell you how many of those members trained last week — and that second number is the one that predicts next quarter's revenue.
Cancellation is almost never a decision. It is the paperwork at the end of a decision that was made weeks earlier, on a Tuesday, when somebody skipped class and nothing happened.
Pull the attendance history of the last ten people who cancelled. You will nearly always see the same shape: a steady run of two or three sessions a week, then a gap of a fortnight, then one session, then nothing. The cancellation email arrives four to eight weeks after the pattern broke.
By the time somebody asks to cancel, they have already stopped being a member in every way that matters. They are asking you to stop charging them for a thing they quietly left. Winning that conversation back is hard, and the discount you offer to save it usually just delays the same email by two months.
The useful window is the fortnight nobody notices.
"At risk" needs a definition, or it stays a feeling. The definition depends on your programs:
Whatever numbers you pick, they should be written somewhere your front desk can see them. A threshold that lives only in the owner's head is not a system.
Here is the whole trick: do not sell anything.
The instinct is to send an offer — a free week, a discount, a "we miss you" campaign. It reads as marketing, because it is, and it tells the member that their absence registered as a billing problem rather than as a person missing.
What works is shorter and more boring:
Hey Sam — noticed you have not been in for a couple of weeks. Everything alright? No pressure either way, just checking. Tuesday's fundamentals class has been a good one to come back to.
Three things are doing the work there. It is from a human, it names a specific class to return to, and it gives them an easy out. Half the replies you get will be "work has been mad, back next week" — and a good number of those people actually come back, because somebody noticed.
The other half will tell you something useful: the 6am slot stopped working, they picked up an injury, the drive got longer after they moved. That is retention research you cannot buy.
The part to automate is the flag, not the conversation. Software is good at watching four hundred attendance records for a gap. It is bad at sounding like you.
A workable setup looks like this: the system flags anyone past your threshold, the list lands with one named person each Monday, and that person sends actual messages. Ten minutes, once a week. If nobody owns the list it will not happen, so give it to a specific human — usually the owner at first, later the front desk.
Some gyms do want the first touch automated. If you go that way, keep it to one message, send it from a real person's name, and make sure a reply reaches an inbox somebody reads. An automated "we miss you" that bounces back from no-reply@ is worse than silence.
This is not dramatic. You will not save everyone, and you should not try — some people move, some get injured, some discover they do not like getting punched. Those cancellations are healthy.
What changes is the group in the middle: the members who drifted for no strong reason and would have drifted all the way out. Catching a meaningful share of those, every month, compounds quietly. It also costs nothing, which is more than can be said for the ads you would run to replace them.
Start this week. Pull a list of everyone who has not checked in for fourteen days, and send ten messages by hand. You will learn more about why people leave your gym in that hour than in any report.