I Cancelled Nine Subscriptions and Timed Every One

Over three months I cancelled nine subscriptions, and because I had been thinking about this topic I timed each one from the moment I decided to cancel to the moment I had written confirmation. The range was enormous and the pattern in it was not random.

Fastest: eleven seconds. A settings page, a cancel button, a confirmation dialogue, done, email arrives immediately. Three of the nine were in this category and all three were tools I would recommend and two of which I have since resubscribed to, which I mention because it is directly relevant.

Slowest: nine days. This one required a phone call during business hours, and the phone number was not on the website but in the terms; when I called, I was offered a discount, then a pause, then a different plan, and when I declined all three I was told the cancellation would be processed and would take up to seven working days. It was, and it was, and I have not gone back.

The middle band is where most of them sat and where the interesting design lives. The common pattern is a cancel option that exists, is genuinely findable, and is placed behind between three and six screens each of which offers something. Are you sure. Would you like to pause instead. Can you tell us why. Here is fifty percent off. It is not deceptive — nothing is hidden and no claim is false — and it took between four and eleven minutes each time.

What struck me on the fourth or fifth of these is that I was being charged for the cancellation in the only currency that mattered, which was my attention, and that the price was set exactly high enough to deter a certain fraction of people without being high enough to generate complaints. That is a design, and it is calibrated, and someone measured it.

The correlation I noticed, with the caveat that nine is not a sample: the difficulty of cancelling was inversely related to how much I had used the service in the final month. The three easy ones were things I had used regularly and stopped for a specific reason. The hard ones were things I had barely opened. That makes commercial sense — a company whose customers use the product does not need to fight at the exit — and it means the cancel flow tells you something about the business model. A service that has invested heavily in retention friction is telling you what fraction of its revenue comes from people who are not using it.

Two practical things I took away. First, check the cancellation route before subscribing, not after; it takes two minutes, it is in the help documentation, and it is a much better signal of how a company will treat you than the reviews are. Second, if you cancel and are offered a discount, notice what that means: the price you were paying was not the price they were willing to accept. I took two of those discounts and both times regretted it, because the reason I was leaving was that I did not use the thing, and half price for something I do not use is still money for something I do not use.

The one I would flag as genuinely improper, out of the nine, is the one that made the discount offer the final screen before cancellation and put “no thanks” in grey text below a large accept button. Everything else was merely effective.