The trial that ends politely Release slices (delivery order, slice-rN labels on each story): - R1 Measure the fury — the surprise join — Surprise Rate and the silent-surprised proxy computable per cohort. Dark - R2 No-surprise surfaces — the day-12 banner and honest receipts. In-app only — no email infra, no vendor risk - R3 The reminder + the experiment — the T-48h email with the holdback experiment attached — the readout that ends the finance argument [BLOCKED BY BLK-01, BLK-05] - R4 One-tap cancel, everywhere — one-tap cancel from all three surfaces; the parity rule asserted in CI per market [BLOCKED BY BLK-02] Every story cites the press-release paragraph or FAQ answer that demands it.
As the analyst sizing this problem, I want charges joined to refunds, tickets and activity, so that Surprise Rate is a number instead of an argument.
Acceptance criteriaGIVEN a cohort with charges, refunds, tickets and activity data WHEN the join runs THEN Surprise Rate and M4 compute with cohort sizes stated
As a trial user who never opens email, I want the app itself to warn me, so that an unread inbox doesn't cost me $12.
Acceptance criteriaGIVEN a trial user with the email unopened WHEN they open the app on day 12+ THEN the banner shows date, price, keep, and one-tap cancel
As a paying customer, I want every receipt to say what happens next, so that no future charge is a surprise.
Acceptance criteriaGIVEN any successful charge WHEN the receipt renders THEN next renewal date, price and a cancel link are present
As a trial user who forgot I signed up, I want to be told before I'm charged, so that the first I hear of it is never my bank statement.
Acceptance criteriaGIVEN a trial reaching day 12 WHEN the reminder sends THEN it precedes the charge by ≥48h (M1 records the ordering) AND a send failure is logged and visible — never silently absorbed (F5)
As the VP in the finance standoff, I want the revenue fear tested by a real experiment, so that neither belief ships unverified.
Acceptance criteriaGIVEN random assignment at trial start WHEN the day-60 readout runs THEN it states effect size, confidence interval and cohort sizes — or declares itself underpowered rather than extending silently
As a user who decided to leave, I want cancelling to be one tap from wherever I am, so that leaving is as easy as joining was.
Acceptance criteriaGIVEN a user in the default flow WHEN they tap cancel from any of the three surfaces THEN cancellation completes on that screen with inline confirmation
As a user in any market, I want cancelling to never take more steps than subscribing did, so that the exit is never a maze.
Acceptance criteriaGIVEN the subscribe flow takes N steps in a market WHEN the cancel flow is measured in that market THEN its step count is ≤ N, asserted in CI per market config
Rendered from the committed artifact — the raw file: jira-import.csv on GitHub. The Surprise Charge is fictional; this is the skill's real output.