The visibility problem

You don't notice the client is gone until they're already gone. Three skipped check-ins, four unanswered messages, two sessions logged out of a planned four, and then the cancellation email lands on a Tuesday. The retrospective is always the same: I should have caught that two weeks ago.

The reason you didn't is that every signal in isolation looks like a normal bad week. Anyone can skip a check-in. Anyone can be slow to reply. Anyone can miss a session. The leverage is in noticing when two or three of these stack inside the same four-week window.

The three signals

Pick three measurables that you can pull from your platform without manual work. Mine, in order of predictive power:

  1. Check-in skips. Did they reply to the Sunday check-in within 72 hours? Binary, every week.
  2. Message latency. When you send a check-in reply, how long until they respond? Median over the last four weeks.
  3. Training compliance. Sessions logged divided by sessions planned. Rolling four-week average.

The four-week sliding window

Once a week, scan every active client across the three signals. The rule is simple: any two of three below threshold for two consecutive weeks, you intervene. Not three. Not one. Two of three is the sweet spot, it filters the false positives without missing the real ones.

My thresholds: check-in skips at >0 in the window, message latency at >48 hours, compliance at <70%. Yours might differ. The thresholds matter less than the consistency.

The intervention message

When the rule fires, the message is not "how can I help?" That puts the work on the client and they will not reply. It's a specific observation:

“"Noticed you missed last Sunday's check-in and Tuesday's session. Want to swap to two sessions this week and skip the check-in this Sunday? Take the pressure off, we can pick up next week."”A real message that has resaved at least 9 clients

The numbers

Across 11 months of running this on my own 30-client roster: the rule fired on 22 clients. Of those, 17 stayed past the next quarter. The other 5 left anyway. Without the rule, my historical attrition on that cohort would have predicted 13–14 leaves. The intervention saved 8–9 clients I would have lost.

Two of those eight are still on the roster 18 months later. The math on lifetime value writes itself.