Three minutes before one of our subscriptions finally died, it got 40 percent cheaper. Then 60. Then free for two months, then “paused” — a status the company’s website had never mentioned before. None of these prices existed anywhere a paying customer might actually want them. They lived only on the way out, dispensed one at a time by a chat agent whose job, he explained, was to “understand the reason for your cancellation today.” The honest answer — that it had taken four buried menus and twenty minutes to reach him — did not speed things up. Nothing speeds things up. Speed is not his department.
Over four weeks we tried to cancel fifty online subscriptions, and we kept a log. Eleven required a phone call to a human being, several of whom appeared to work in the same enormous room. Total time spent quitting: fourteen hours and thirty-two minutes. Four we never managed to cancel at all — those four are still billing us, monthly, and we have begun to think of them less as services than as dependents.
But the numbers aren’t the story. The shape is. Every one of those fifty services had let us in without ceremony — two taps, a stored card, a button the size of a barn door. The same companies that engineered joining to feel like falling off a log had engineered leaving to feel like chewing through the log, and the gap is not understaffing, legacy software or a support team having a bad quarter. It is the product. Signing up belongs to a growth team with a conversion target; cancelling belongs to a retention team with a save-rate target. Both teams are excellent at their jobs. Both are pointed at you.
The playbook
Run an experiment like ours and a playbook emerges, too consistent across too many companies to be anything but taught, tested and shared. There is the offer you never asked for — the price that materialises only at the exit. There is the menu four levels down: Settings, Account, Membership, Billing, and then, in a grey slightly lighter than the grey around it, the word “cancel.” There is the chat agent who needs a reason, the exit survey dressed up as concern, the warning that your cancellation “may take up to ten business days to process,” as though unsubscribing were an international wire transfer. And there is the phone line, open nine to five on weekdays — a bold scheduling choice for businesses that take your money at three on a Sunday morning.
Academics have a name for this architecture, and the name is unkind: the roach motel, easy to check into, brutal to leave. In a cross-country study published under that phrase, researchers subscribed and cancelled across 67 news sites in four countries and measured the asymmetry in clicks. Signing up averaged 4.9 clicks; cancelling averaged 6.2. Four sites — all American — required a phone call to quit. Bloomberg’s flow forced leavers to negotiate with a chatbot for ten clicks; others made customers type out a phrase, the digital equivalent of writing lines. In only two jurisdictions, Germany and California, could every site be cancelled online. That last fact is not a coincidence.
A second research team subscribed to 47 Massachusetts newspapers and tried to leave: 21 let them click to cancel, while 26 demanded a phone call, a chat or an email. More telling is California, where the law has forced the exit door into existence. Thirty newspapers there offered click-to-cancel — but 76 percent padded the path anyway, two-thirds of them with a mandatory exit survey, a quarter with last-minute offers. The term of art is “sludge”: friction poured, by the yard, into whichever stretch of the journey a business would rather you never finish. Regulate the button into being and the sludge simply migrates around it.
The rule that died of process
Regulators, to their credit, have stopped filing all this under “customer service” and started treating it as design — which is to say, as intent. Since 2010, the Restore Online Shoppers’ Confidence Act has required a “simple mechanism” for stopping recurring charges, a phrase that has spent fifteen years carrying more weight than it can bear. Germany’s Fair Consumer Contracts Act went further in 2022: anything subscribed to online must be cancellable online. California’s rule is the most pointed of the three — customers who joined online must be able to leave “at will, and without engaging any further steps that obstruct or delay,” language that reads as if the drafter had personally met the four-level menu.
The Federal Trade Commission tried to make symmetry federal law. Its “click-to-cancel” rule, finalised in October 2024, would have required leaving to be as easy as joining, full stop. It never took effect. In July 2025 the US Court of Appeals for the Eighth Circuit vacated the rule — not because the FTC was wrong about the harm, but because the agency had skipped a required procedural analysis. The rule against friction was killed by process. You could not design a more fitting ending, which is presumably why nobody had to.
The clean standard remains symmetry: you should be able to leave a service the way you entered it — same device, same clicks, same hour of the night. Germany and California show the button can be forced into existence, and that the companies produce it the moment they must — which is the proof that the obstacle courses everywhere else are a choice rather than a constraint, even where the sludge simply relocates to the survey in front of it. Until the standard spreads, the cancellation flow stays the most honest screen a company builds: everything before it is designed to please you, and that one is designed to beat you.
So before you subscribe to anything, find the exit first. The businesses that hide it are telling you something the big friendly button never will. Somewhere in the background, our four unkillable subscriptions go on renewing — small monthly monuments to an economy where “forever” is the default setting and “goodbye” is a feature request.