A Dark Pattern That Won Short-Term and Cost Trust Long-Term

The cancellation flow redesign was pitched as 'reducing friction around the decision' and measured purely on one number: reduction in completed cancellations. It worked, by that number. I think it was still the wrong call, and the longer-term data eventually agreed with me.

What the redesign actually did

The new flow buried the cancel button behind a 'pause your subscription' offer presented as the default path, required confirming through three additional screens each offering a discount or alternative, and made the actual cancel action a small text link at the bottom rather than a button. It's a well-documented pattern, and it did reduce completed cancellations by about 15% in the first month.

A dark pattern doesn't make people want to stay. It makes leaving harder, and those are not the same outcome, no matter what the retention dashboard says.

What showed up later

Support ticket sentiment around cancellation attempts turned sharply negative. Trustpilot and app store reviews started mentioning the cancellation flow specifically, by name, as a reason for a low rating — unprompted, which is a strong signal that a pattern has become reputationally costly rather than just annoying. And a chunk of the users who did eventually cancel did so through a chargeback or bank dispute instead of the in-app flow, which is worse for the business than a clean cancellation in almost every way, including the processor fees and dispute-rate risk that come with it.

• We measured completed cancellation rate and declared success without measuring cancellation method, which hid the chargeback shift entirely for two months.
• We eventually replaced the flow with a single clear cancel button plus one genuine retention offer, no forced multi-screen gauntlet.
• The simplified flow's cancellation rate landed close to where the dark-pattern flow had been, within a few points, because most of the 'reduction' had just been diverted into a worse channel, not actually prevented.
• Reputational damage from dark patterns shows up in review text and support sentiment well before it shows up in any retention metric, which is exactly why it's easy to miss if you're only watching the metric the pattern was designed to move.

If a retention tactic's defense is 'the number went up,' ask which number didn't get measured. In this case it was trust, and trust turned out to have a price the dashboard just wasn't tracking.