Desk note · 12 Mar 2026
Event taxonomy that survives growth
Most paywall experiments fail before the copy changes. They fail because the taxonomy starts at a single “wall viewed” event, and that name is easy to inflate once you add recap screens, settings gears, and failed restores. Conversion then looks like a product win when you merely trained the logger. Growth multiplies surfaces; a sloppy taxonomy multiplies lies.
In the Paywall Timing Lab we ask studios to freeze three events before they move anything: qualified exposure, attempt, and first paid receipt. Qualified exposure means the user could have acted. A wall that renders behind a loading spinner does not count. An attempt is a store sheet or purchase flow opening, not a tap on “maybe later.” First paid receipt is the only success that may carry money.
Why day-zero walls poison later tests
If the hard wall appears before the user has done the one job the app promised, every later timing test inherits a hostile baseline. Users who would have paid on day two already churned, and the remaining population is either highly intent or highly confused. You cannot A/B your way out of that selection effect by adding a “softer” wall on day one for half the traffic.
Utility apps in Thailand often ship with a day-zero wall because the template did. We do not moralise about it. We ask them to log a task-complete event — first document scanned, first workout finished — and to delay the hard wall until that event or a 24–36 hour clock, whichever comes first. The clock is written down. It is not “when it feels right.”
Metrics we ignore on purpose
Impression rate, tap-through rate on the wall itself, and “paywall engagement time.” They are useful debugging signals for design. They are not revenue analytics. If your weekly review leads with them, the conversation will drift toward visual tweaks while the clock stays wrong.
Net new receipts per qualified exposure, by cohort clock, is slower to compute. It also survives a finance question. That is the trade we teach.
A mild warning
Delaying a wall can drop short-term conversion while you wait for the new clock to fill. Teams that need a chart to rise this Friday should not start a timing test. They should finish restore logging first; otherwise the test will credit purchases that already existed on another device.
If you want the exercises, the Paywall Timing Lab is two weeks. The flagship covers the same clock inside a wider event contract.