Why Is It So Hard to Cancel a Subscription?

Signing up for a subscription can make you feel unusually capable.

The button is enormous. The form remembers what a form is supposed to do. Your credit card slides into the system with the confidence of a diplomat arriving under armed escort. If there is a free trial, the company is prepared to begin your new life before you have finished deciding whether you wanted one.

For a brief moment, you and the system want exactly the same thing.

You want to join. It wants you to join. The path is obvious because nobody in this relationship has a reason to discuss the path.

Then one day you want to leave.

Now the relationship gets interesting.

The service may still let you cancel. But first it may want to know whether you would rather pause. Or downgrade. Or accept a discount. Or remember the features you are about to lose. Or explain why you are leaving. Or confirm that you understood the warning. Or confirm that the confirmation was not an emotional decision made during a difficult afternoon.

A system that accepted your credit card after three clicks has suddenly become your most concerned relative.

The interesting question is not whether every company is secretly trying to trap everyone forever. The evidence does not support that. The better question is what happens to a system when the user's direction changes.

Because the entrance tells you how a system behaves when everybody agrees.

The exit shows you what happens when they might not.

Friction rarely introduces itself as friction

Most difficult cancellation processes do not begin with a locked steel door marked CUSTOMER RETENTION.

They begin with one more reasonable thing.

Choose a reason. Review what you will lose. Consider a cheaper plan. Confirm. Wait—would you rather pause? Confirm again. Check your email. Follow the link. Return to the account page. Locate the setting whose name sounds almost, but not quite, like the thing you are trying to stop.

Individually, each step can look defensible. Together, they can turn a small decision into administrative work.

This matters because people often describe cancellation as simply “hard,” which is emotionally satisfying but not very useful. Hard in what way? More steps? More time? More searching? More choices? A different channel? A human retention conversation? A screen that keeps offering exits from the exit?

Once the problem is broken into those parts, frustration becomes something researchers can actually examine.

You can measure when the exit gets longer

Julian Runge and colleagues examined account creation and account deletion across 40 online platforms. They measured required and actual clicks, elapsed time and features of the process.

In that bounded sample, creating an account averaged 18 actual clicks and 1.7 minutes. Deleting one averaged 27 actual clicks and 4 minutes.

Those numbers are useful, but only if we keep them attached to what the researchers actually studied. They are not universal averages for canceling paid subscriptions. Account deletion and subscription cancellation are different endpoints, and the study examined a particular set of platforms at a particular time.

What the findings give us is more durable than a universal number: proof that entry and exit can be compared instead of merely complained about.

The researchers also documented features such as repeated confirmations, suggestions to deactivate instead of delete, grace periods and information intended to discourage quitting.

Suddenly the familiar experience has components.

The system can require more actions. It can require more time. It can introduce more decisions. It can make the destination less obvious.

Cancellation has gone from “why is this annoying?” to “where, exactly, did the extra work appear?”

That is a much better question.

It is also where click counting stops being enough.

Sometimes the distance is made of decisions

Imagine two exit processes.

One takes eight obvious clicks. The other takes six, except one is hidden under Manage Membership, another opens a retention offer, another asks why you are leaving, and the last one presents a button that looks suspiciously like it might undo everything you just did.

The six-click process can feel longer.

Dominique Kelly and Victoria Rubin studied account-disabling interfaces on 25 social-networking sites. They identified five broad types of dark patterns and 13 subtypes, documenting 59 coded instances across the sample. Emotional pressure or distraction appeared in 13 of the 25 sites, account reactivation in 10, and last-minute solutions in nine.

Again, that is account disabling on a bounded sample of social-networking sites, not every paid subscription on the internet.

But it reveals another form of friction: the exit can become harder without becoming physically much farther away.

A system can add uncertainty. It can add persuasion. It can add alternatives that require the user to keep deciding whether the decision they already made is still the decision they would like to make.

At this point, canceling an $8.99 subscription can begin to resemble a parole hearing conducted by a dropdown menu.

And this is where the obvious conclusion becomes tempting.

If leaving is harder than joining, the company must be doing it on purpose.

The evidence tells us to slow down.

A longer exit is evidence. It is not mind reading.

Some extra exit work can be legitimate.

Permanent account deletion may require identity verification. A service may need to explain what happens to stored data. Billing rules can make the endpoint more complicated than the original signup. A user who says “cancel” might mean stop renewal, end access immediately, delete the account, deactivate it temporarily or change how personal data is used.

Those are not interchangeable requests.

This is why the vocabulary becomes so irritatingly important exactly when the user is least interested in vocabulary.

Cancel Plan. Turn Off Renewal. Deactivate. Close Account. Delete Account.

Five phrases enter. One recurring charge leaves. Possibly.

The research has the same problem. Runge and colleagues studied account creation versus deletion. Kelly and Rubin studied account disabling. Federal Trade Commission cases include subscription and recurring-payment cancellation. Those sources point toward a broader phenomenon of exit friction, but they do not measure one identical action.

So asymmetry tells us something happened to the process. It does not automatically tell us why.

That limitation is not a disappointment. It gives us a fairer test.

If we cannot read the designer's mind, we can still inspect the hallway.

The useful question is whether the extra work belongs there

A good exit does not have to be identical to the entrance.

Sometimes it should be longer. Deleting years of stored information probably deserves more warning than creating an empty account. Ending a complicated service may require information that beginning it did not.

The better question is proportionality.

First define the endpoint. What is the user actually trying to stop or remove?

Then look at the work: meaningful actions, elapsed time, searching, channel changes, confirmations, retention offers, waiting periods and moments where the user cannot tell which choice leads to the intended result.

Then ask what those steps are doing there.

Some will protect the user. Some will explain consequences. Some may satisfy technical or billing requirements. Some may offer a useful alternative.

And some may be surprisingly difficult to defend once somebody writes them all down in order.

Stuart Mills and colleagues have proposed combining dark-pattern analysis with “sludge audits,” while also warning against overly simple measures such as assuming equal click counts necessarily mean equal experiences.

That is the important shift. We do not need a universal rule saying every cancellation must take exactly as many clicks as signup.

We need to know where the additional burden appears and whether it makes sense for the task.

Dunce Dynamics' proposed technical standard remains regrettably hostile to dramatic accusations:

Measure first. Become suspicious with supporting documentation later.

Sometimes the hallway gets bad enough that regulators notice

The Federal Trade Commission's 2022 staff report Bringing Dark Patterns to Light discussed designs that can obscure or impair consumer choice, including practices that make subscriptions or recurring charges difficult to cancel.

The FTC's case against ABCmouse alleged a difficult-to-find, lengthy and confusing cancellation path despite an “Easy Cancellation” promise.

In another 2022 case, the FTC alleged that Vonage let customers enroll through multiple channels but required cancellation through a live retention agent, with customers encountering problems locating the cancellation number, getting transferred, receiving callbacks and completing the process. The settlement required a simpler cancellation process and $100 million for refunds; in 2023 the FTC reported sending more than $99 million to hundreds of thousands of eligible consumers.

These are specific enforcement examples, not evidence that every annoying subscription screen violates the law.

What they demonstrate is that exit friction can move beyond subjective annoyance into conduct concrete enough to document, challenge and remedy.

Which brings us back to the strange change in personality between signup and cancellation.

During signup, every design choice seems to ask: How can we help this person finish?

A good cancellation system should still be able to answer that question.

Even when “finish” now means leave.

The reverse-direction test

Subscriptions are only one version of a larger systems problem.

Consider the moments when people reverse a direction an organization normally prefers: returning something instead of buying it, closing an account instead of opening one, unsubscribing instead of subscribing, changing a privacy setting away from the default.

Those situations are not factually identical, and this Article is not claiming research about all of them from subscription evidence.

But they suggest a useful way to look at systems.

Watch what happens when the user's arrow turns around.

When everyone is moving in the preferred direction, good design can be almost invisible. The buttons are obvious. The language is simple. The next step is exactly where a reasonable person expects it.

Reversal exposes different parts of the machinery: warnings, dependencies, incentives, verification, persuasion, ambiguity and all the things that were unnecessary when everybody wanted the same outcome.

Some of that machinery belongs there.

Some of it may not.

The point is that you cannot learn much about an exit by studying the front door.

What this seems to tell us

The easiest conclusion is that companies make cancellation difficult because they want your money.

Sometimes retention incentives may matter. But the evidence does not give us permission to use that explanation as a universal shortcut, and doing so actually makes the subject less interesting.

The better observation is that cancellation changes the relationship between a user and a system.

At signup, success is easy to define because both sides are moving toward the same immediate event: start the service.

At exit, the system may need to handle several things at once—verify, warn, preserve, offer, document, retain, terminate—and the user has to discover which of those things are necessary to accomplish one simple intention: I am done.

That is why the exit deserves to be measured separately.

The entrance shows how efficiently a system handles yes.

The exit shows how competently and proportionately it handles no.

A mature service should probably be judged by both.

Because if a company can remember your credit card from 2019, it can probably remember where it put the Cancel button.