The roach motel: easy to sign up, engineered to be hard to leave

What it is

Product teams measure sign-up as a funnel and optimise every step out of it. The roach motel is that same discipline pointed backwards. If a cancellation flow has five screens and each one loses fifteen per cent of the people who reach it, fewer than half the people who set out to cancel will finish. No individual screen has to be dishonest for that to happen. Each just has to be one more thing between a person and the door.

That arithmetic explains why hard-to-cancel flows share a shape rather than a trick. There is a screen that asks whether you are sure. A screen that asks why, with the continue control disabled until you answer. A screen that warns you what you will lose. A screen with a retention offer. And a final screen where the confirm control is smaller, greyer and lower than the button that keeps you subscribed. Sign-up, meanwhile, was a single form.

“Unsubscribe Labyrinth” is a compressed version of the genre. The first screen shows a spinner captioned “Loading unsubscribe page…” — except the spinner is the button, and the page is not loading at all, so the natural response of waiting patiently is the losing move. A support widget covers the continue link on the second screen until you dismiss it. The third screen swaps the positions of “Yes, unsubscribe me” and “No, keep my subscription!”, then tells you the buttons like to move around. The fourth buries the real confirm link inside a paragraph of filler beneath an offer of half a per cent off a ten-thousand-dollar purchase. Five screens, one exit, and a decoy on each.

Harry Brignull originally named this pattern the roach motel, after the line about roaches checking in but not checking out; the taxonomy on deceptive.design now lists it as “hard to cancel”, and files the surrounding techniques as obstruction, nagging and visual interference. The FTC’s 2022 staff report on dark patterns describes the same shape from the enforcement side: cancellation paths that are difficult to find, lengthy, confusing, and lined with promotions that route people back out of the flow.

Why it works on people

The first mechanism is pure attrition, and it does not require anyone to be fooled. Every step is a chance to be interrupted, to decide this is not worth the afternoon, or to tell yourself you will finish it later. Cancelling is rarely urgent. It competes with everything else, and a flow that costs twenty minutes reliably loses to a flow that costs twenty seconds.

The second is that these flows attack the moment of resolve rather than the decision. Someone who has decided to leave arrives with a fixed and fairly small amount of determination. Retention offers, warnings about what will be lost, and copy about how long you have been a member all work on that budget — not by changing the underlying reasons, but by spending the resolve before it reaches the confirm button. A discount that would not have persuaded you yesterday can still persuade you on step four, when the alternative is two more screens.

The third is that the interface stops behaving predictably at exactly the point where behaving predictably matters. A control disguised as a loading spinner, a modal parked over the continue link, buttons that trade places between screens: each one converts a decision you already made into a fresh act of attention. Errors here are asymmetric by construction. Hesitating, mis-clicking or giving up all leave you subscribed; only one specific sequence of actions gets you out.

Fourth, the sunk cost runs the wrong way. By the time you are on the last screen you have invested effort in leaving, which makes finishing feel closer than it is — right up until the last screen turns out to have a sixth. Flows that reset you to the beginning on a wrong click exploit this directly.

And the incentive is unusually clean. Retained subscribers are countable, immediately, in revenue. The cost of a bad exit is a diffuse reputational one that lands later and cannot be attributed to the team that shipped the flow.

Where you meet it

• Subscriptions that can be started on the web in one click but can be ended only by telephone, by post, or during business hours in a single time zone.
• Cancellation buried under a support-article search rather than sitting in account settings, so finding the flow is itself a task.
• Mandatory exit surveys where the continue control stays disabled until a reason is chosen, and each reason is written as a small reproach.
• Retention offers presented as a full-screen interstitial with the cancel path demoted to a text link, repeated once per screen.
• Email footers where “unsubscribe” leads to a preference centre that requires signing in — including for people who never had an account — before anything can be turned off.
• “Pause instead?” controls placed where the cancel control was on the previous screen, so muscle memory selects the retention option.
• Confirmation screens that quietly leave one list active, so the mail continues under a different heading.

Designing around it

• Match the exit to the entrance. If the service can be joined online, it should be leavable online, in the same session, without a phone call. This is the requirement that has drawn the most legislative attention: California’s automatic renewal rules require a cancellation route in the same medium the consumer used to enrol, and the EU’s Digital Services Act names making termination harder than subscription as a practice the Commission may issue guidance on. In the US the picture is genuinely in flux — see the note on the Negative Option Rule below — but the direction of travel across jurisdictions is consistent.
• Put cancellation where people look for it: in account or subscription settings, at the same level as billing, with a plain label. A cancel route that can only be found through a help-centre search is an obstruction whatever the copy says.
• Budget the flow in screens and defend the number. Confirm once, and make that the whole flow. Every additional screen is a measurable loss of people who intended to finish, and the team should have to justify each one on grounds other than retention.
• Let a retention offer exist, but let it exist honestly. Show at most one, do not make it blocking, and keep the cancel control continuously visible and equally prominent beside it. California’s rules take this literally: a business that presents a discount in the cancellation flow has to simultaneously display a prominently located, continuously and proximately displayed direct link or button labelled “click to cancel”.
• Make surveys optional. Asking why someone is leaving is reasonable research; gating their exit on answering is a toll. Ship the reason field with a skip control that is as easy to press as submit.
• Keep control labels stable and keep the user’s action primary. The button that does what the person came to do should be the one that looks like the primary action, in the same position on every screen of the flow. Reordering choices between screens is called out in California’s privacy regulations as confusing design, and it is a reliable source of mis-clicks anywhere it appears.
• For email specifically, implement one-click unsubscribe properly. Send the List-Unsubscribe and List-Unsubscribe-Post headers described in RFC 8058 so mail clients can offer a native unsubscribe control, and make the link in the message body resolve in a single page with no sign-in. US law is unusually concrete here: the CAN-SPAM Act and its implementing rule bar requiring a recipient to pay a fee, to supply anything beyond their email address and opt-out preferences, or to take any step other than replying to the message or visiting a single web page — and require opt-outs to be honoured within ten business days.
• Confirm the outcome in writing and be specific about scope. “You have been removed from Product Updates. You are still subscribed to Order Notifications and Security Alerts” is an honest confirmation. “Preferences updated” is not.
• Instrument the exit as carefully as the entrance. Log completion rate and step-level drop-off for the cancellation flow, and review them beside sign-up. A step that loses a large share of the people who reach it is either broken or working as an obstacle, and the two are worth telling apart before someone else does.
• An honest version is short enough to describe in a sentence: settings, then cancel, then one confirmation screen that states what ends and when, then an email receipt. It retains fewer people than the labyrinth. It also produces people who might come back, and a flow that survives being screenshotted.

Questions

What is the roach motel dark pattern?

It is a design where joining a service is easy and leaving it is deliberately hard — one-click sign-up against a cancellation flow of several screens, guilt copy, retention offers and hard-to-find confirm controls. Brignull named it after the roach trap; the current taxonomy calls it “hard to cancel”. It works by attrition: each extra screen loses a share of the people who intended to finish.

Is it illegal to make cancellation difficult?

In some places and in some circumstances, yes, but the rules differ and are moving. In the US, online negative-option sales must offer “simple mechanisms” to stop recurring charges under ROSCA, and several states — California among them — require that a subscription started online can be cancelled online; a broader FTC “click to cancel” rule was finalised in 2024 and then vacated on procedural grounds in 2025, with new rulemaking opened in 2026. In the EU, the Digital Services Act names making termination harder than subscription as an interface-design concern, and withdrawing consent must be as easy as giving it. Whether a specific flow breaks a specific law is a question for regulators and courts.

How should an unsubscribe flow be designed?

Same medium as sign-up, reachable from account settings, one confirmation screen, at most one non-blocking retention offer with the cancel control still visible beside it, an optional rather than mandatory reason survey, and a written confirmation naming exactly what has ended. For email, ship RFC 8058 one-click headers and make the in-message link resolve in a single page with no sign-in.

Why do companies keep building hard cancellation flows if they annoy people?

Because the benefit is measurable and immediate and the cost is not. Every retained subscriber shows up in this quarter’s revenue with a clear attribution; the churned goodwill, the support load and the regulatory exposure arrive later and are hard to trace back to the flow that caused them. Measuring step-level drop-off in the cancellation funnel, and reviewing it beside sign-up, is what makes that trade visible to the team making it.

Related patterns

• Confirmshaming and the confirmation gauntlet: when “are you sure?” is the trap
• Interruption cost: the work does not resume where it stopped
• Cookie consent dark patterns: double negatives and asymmetric choice
• Forced consent: unreadable terms and the ritual of agreeing