Your retention save flow is manufacturing churn that would not exist otherwise. Every subscription team tracks "save rate." That one number hides four completely different users — and only one of them should ever see a save offer. Persuadables will cancel unless you intervene. The save flow earns its keep here. Sure Things were never leaving; a 20% discount just burns margin on a user who clicked cancel by accident. Lost Causes already decided; a 4-page flow just makes them angry — and looks exactly like the Amazon "Iliad Flow" the FTC fined $2.5B in September 2025. Sleeping Dogs are quietly renewing, until your "we'd hate to see you go" email reminds them they're paying $49/month for something they haven't opened in three months. Run the math on a 200K-subscriber SaaS at 3% monthly voluntary churn: roughly 6,000 cancel-intent users a month. Industry research puts 10–20% of them in the Sleeping Dog segment. Contact all 6,000 — which is what every off-the-shelf cancel-flow tool does — and you push 600–1,200 people toward a cancellation they weren't going to make. At $50 ARPU that's $360K–$720K a year your own retention system destroyed. Telenor, the Norwegian telecom, learned this the hard way: their retention campaigns drove higher churn in the treatment group. They only caught it because they ran a proper holdout test. Most subscription businesses never do. Standard churn models ask "who will leave?" Uplift models ask "who will leave because of our intervention?" Different question, different answer — and the only one that can tell a Persuadable from a Sleeping Dog. Here's the part most teams miss: ROSCA doesn't require proving a specific dark pattern. The FTC only has to show cancellation wasn't "simple." The plain test — if your cancel flow has more steps than your signup flow, you already have exposure. Save this before you design your next one. #SubscriptionRetention #SaaSChurn #UpliftModeling #FTCCompliance #CausalAI