- Amazon paid $2.5 billion for a cancel button that took 6 clicks to reach.
Uber is now facing 21 state attorneys general over a flow with 23 screens.
Your retention team calls this "optimizing save rate."
Regulators call it an enforcement action. 🧵 - Start with the metric your retention team hides behind: "save rate." The % of users who start cancelling and don't finish.
A 30% save rate sounds like a win.
It's a vanity number that conflates four completely different users into one. - Four people are hiding behind that number:
Persuadables — cancel unless you intervene. The only ones worth saving.
Sure Things — never leaving; your discount is wasted margin.
Lost Causes — gone regardless; a 4-page flow just makes them angry.
Then the dangerous one 👇 - Sleeping Dogs: renewing happily, would keep paying — until your "We'd hate to see you go" email reminds them they pay $49/mo for something they haven't opened in months.
Your retention system just manufactured the exact churn it exists to stop. - The math is brutal. A SaaS company with 200K subs and 3% monthly voluntary churn has ~6,000 cancel-intent users a month.
If 10-20% are Sleeping Dogs and your flow contacts everyone, you push 600-1,200 of them out.
At $50 ARPU: $360K-$720K/yr destroyed. - Not theory. Telenor, the Norwegian telecom, ran retention campaigns that caused 2% HIGHER churn in the treated group.
They only caught it because they ran a real holdout test.
Most subscription companies never do — so they never see the damage. - Off-the-shelf tools can't fix this. ProsperStack, Chargebee Retention — they A/B test offers but can't tell a Persuadable from a Sleeping Dog. And the enterprise option, Pega, runs $500K+ and still won't audit your flow for dark patterns.
- The question every one of them misses isn't "who will churn?" It's "who churns BECAUSE we intervened?"
That's causal uplift modeling — measuring each user's treatment effect — not the predictive churn score every tool already sells you. - And the legal floor keeps rising. ROSCA doesn't require proving a "dark pattern" — only that cancelling wasn't "simple."
If your cancel flow has more steps than your signup, you're exposed.
Jan 2026: the FTC went after JustAnswer's AI chatbot for exactly this. - It compounds across states: California's "One Save" rule caps offers at one per cancellation; NY mandates online-only cancellation. The strictest law is your floor.
Click-to-Cancel was vacated on a technicality, not a green light. Amazon, Chegg, HelloFresh settled after. - Honest question for retention teams: have you ever run a true holdout test on your save flow — measured uplift, not just save rate?
If not, you don't actually know whether it makes you money or quietly burns it. #SaaS #ChurnReduction - We build the causal segmentation + compliant flow design + dark-pattern auditing that no single vendor ships together. How it works: https://veriprajna.com/solutions/ethical-subscription-retention-ai