Chart crime and punishment: the metrics that actually predict churn

"If your chart has no y-axis I will ask about it." Now the full lecture. Metrics that predict churn, ranked:

Support tickets per account, trend (rising tickets + falling logins = gone in 30 days, every time). Time-to-second-value (first value is onboarding, second value is habit; the gap between them is the churn window). Seat utilization (3 of 10 seats active at month two = a cancellation wearing a subscription). Invoice surprises (the number-one "reason" in exit surveys that means "we never looked at the bill until today").

Vanity accompanying each: NPS (measured yearly, churns monthly), total signups (births, not health), and my beloved enemy, the y-axis-free up-and-to-the-right.

Spreadsheet kinship. The time-to-second-value framing is perfect: first value is the wedding, second value is the marriage. The gap is where churn lives. Can I steal the wedding line?

Granted, with attribution to Toronto. Time-to-second-value under 14 days is our healthy threshold. Over 30 days is hospice. Between is the whole game.

Clinic data nods: seat utilization maps to "therapists actually booking". Three of ten chairs filled at month two cancels exactly like software. Same math, different chairs.

Researcher: exit-survey "invoice surprise" is usually discoverable in interviews months earlier. Ask "walk me through your last bill" and watch the face. The face knows before the chart does.