Roach Motel

Making a subscription hard to cancel went unnamed as a problem for a long time. Newspapers could only be cancelled by calling a number that was open during office hours, and gyms required showing up in person to sign a form. When products moved online, that structure moved with them: signup got rebuilt as a 30-second web form while the exit stayed where it was.

What kept the structure alive is that it doesn't look like a problem from the inside, because every step has a reason attached. Identity has to be confirmed by phone. Someone who might stay deserves to hear the offer. A cancellation reason helps the team fix the product. Taken one sentence at a time, each of these is hard to argue with, and the metrics cooperate too — fewer completed cancellations means a visibly lower churn rate that month.

The name comes from the insect trap that's easy to walk into and impossible to walk out of. Having a name is what made it possible to look at the whole flow as a single object instead of six defensible screens. Design that deliberately pushes people toward choices against their own interest is called a dark pattern, and within that group this type belongs to the branch that narrows the exit rather than the branch that distorts the moment of choosing.

The entire mechanism is an asymmetry in what the person pays on the way in versus the way out. One line of evidence settles it: one minute to sign up, eleven minutes to cancel. The typical flow stacks four moves.

1. Change the channel. Signup works any hour from a browser while cancelling requires a phone call on a weekday afternoon. The moment a person has to schedule around you, some of them push it to next week.
2. Add steps. Put the reason selection, the satisfaction survey and the final confirmation on separate screens. Each screen sheds a few people, and across four screens those few add up to something close to half of everyone who started.
3. Insert retention offers. Two months at half price, then a free month when that's declined, then an option to pause. This is nagging — re-asking after a refusal — brought inside the cancellation screen.
4. Twist the wording. Label the last button as giving up your benefits. That's confirmshaming, putting an embarrassing phrase on the decline option, layered on top.

The four moves work because people abandon a goal more easily the further they are from it. With six screens between the first click and the actual cancellation, anyone interrupted mid-flow usually doesn't come back to finish.

One number does the judging: of the people who started cancelling, how many finished. A result of 60 percent means four in ten tried to leave and failed, and those four then sit in the same column as the users who stayed because they were happy, which is what makes the retention number unreadable.

Anything that pulls money automatically every month can host this structure. Phone plans, gym memberships and newspaper subscriptions are the old examples; a subscription bought inside an app that can only be ended on a website outside the app is the recent one. Account deletion sits in the same spot. Plenty of products still let you sign up with nothing but an email address while deleting the account requires writing to support and waiting for a reply.

Regulation is closing in on it. Several countries are writing into law the principle that you must be able to cancel through the channel you used to subscribe, so a web signup obliges a web cancellation. That makes this structure a legal review item rather than only a reputation question.

Checking for it isn't a documentation exercise. Hand the product to somebody seeing it for the first time, have them create an account and pay for it, then have them delete that account and time every stage. At Pagedeck, Grace orders that check every quarter. Noa ran it herself and brought back a table reading one minute to sign up, eleven to cancel, which is what led to cutting two retention screens out of the cancellation flow. The team that built a flow is always the last to notice what's inside it.

"If signup is on the web and cancelling is phone-only, that's a roach motel. The channels have to match."

Asking for the cost of leaving to be set to match the cost of arriving.

"Churn didn't go down, completion of cancellation went down. The same people still want out."

Pointing out that the metric improved because the exit was obstructed.

"Let's keep one retention offer on this screen and drop the second. Three crosses the line."

Drawing the line on how many times a product may ask someone to stay before the asking becomes the obstacle.

This comes up in Pagedeck's quarterly review. Sofia notes that removing the retention screens costs the quarter some saved accounts, and Grace brings up the accounts saved that way that came back as card disputes the following quarter. They're reading the same table over different time windows.

The decision usually turns on records rather than numbers. Reading three support messages from people who tried to cancel and failed makes the meeting shorter.

The most common mistake is assuming that a visible cancel button means this isn't a roach motel. Legal reviews often check only that the button exists, which cements the idea. But if six screens sit behind the button and three of them are retention offers, the person is paying roughly what a phone call would cost them. The judgment rests on the gap in time and steps between signing up and cancelling, not on whether a button is present. Miss that and teams fix nothing while quietly moving the button somewhere less visible.

The other frequent conclusion is that it works, so it can stay. This month's churn rate really does drop. But a person held in place stops using the product while the charges continue, and that state doesn't last. It comes back as refund requests, card disputes and app store reviews. When dispute rates climb, payment processors raise fees or reopen the contract. The money saved lands in this quarter and the cost lands in the next, which is why a single quarter reads like a win.