Journal · June 2025
A quieter way to instrument cancel flows
Cancel sheets in subscription apps often try to do three jobs at once: retain the person, collect a reason, and satisfy a lawyer. Instrumentation then inherits the mess. Every extra question becomes an event, every skip becomes a null, and involuntary churn — failed cards, expired methods — gets filed under “too expensive” because that was the only checkbox left.
The quieter method we teach in Cancel & Winback Signals is to instrument states, not sermons. Did they open the leave sheet? Did they confirm leave? Did the partner report a payment failure in the same window? Those three are enough to stop win-back copy from reaching people who never chose to go.
Reasons are optional; states are not
A single optional reason field is fine. A twelve-step interrogation that blocks cancel is not instrumentation; it is friction dressed as research. If you need richer qualitative notes, run a separate interview programme. Do not smuggle it into the leave path and then wonder why your cancel_reason export is 70% empty.
Pause is not cancel
Wellness and education apps in Thailand often offer pause. If pause and cancel share an event name, your funnel will celebrate “saves” that were merely delays. Give pause its own confirmed state. Finance already knows the difference on the invoice; analytics should catch up.
This work is unglamorous. It also prevents the most embarrassing email a subscription team can send: a discount offer to someone whose card simply expired on a Tuesday in Phayathai.