Live Prompt: The Redesign Shipped and Abandonment Got Worse

New live prompt: "You shipped the payment-step redesign we discussed — earlier cost disclosure, equal-weight guest checkout, trust signals, mobile touch-target fixes, clearer decline errors. Payment- step abandonment went from 68% to 74%. What do you do in the next 20 minutes?"

1. What's the first thing you'd do, before proposing any new fix — and why does jumping straight to "revert everything" or "add another fix" both skip a necessary step?
2. Name three distinct ways a redesign grounded in reasonable hypotheses could still make the target metric worse, each implying a different diagnosis and a different next action.
3. How would you decide, live, whether to roll back, iterate, or hold and gather more data?