Two Clicks to Subscribe and Twenty Minutes to Leave

The Asymmetry

Signing up for a subscription takes seconds. One form one card number one confirmation click

Canceling has often meant something completely different: a phone number you have to call during business hours a trained retention agent to keep you online or a hidden setting with four menus inside an account page you forgot existed

That interval between two clicks and twenty minutes is not an accident of bad design. It's a number that someone in finance modeled approved and renewed every quarter to pay off

Why It Works

The behavioral mechanisms behind this are well studied and each of them is used purposefully

Friction. A task with more steps is completed less often even by people who have every intention of doing it. Add a phone call and a thirty-second task becomes one that requires an hour of free time during a work day something many people never encounter

Bias present. Canceling will save you money later but will cost you effort now. That's exactly the type of decision that most people put off

Status quo bias. Staying subscribed requires nothing. Defaults are powerful precisely because they require nothing of you

Retention offers. A discount offered at the time you try to leave converts a fraction of those who would. That's the only reason the call center exists

The Price of a Retained Month

This is the part that's left out in most articles on this topic and it's the part I really find interesting: Friction isn't so much a moral failing as it is a line item. A subscriber who wanted to leave in January but got caught in a retention flow and is still paying in March generated two additional months of revenue at near-zero marginal cost because the content servers and support staff were already paid for. From a spreadsheet standpoint that subscriber is close to receiving free money

The honest way to evaluate practice then is not "is this annoying?"but "Does the retained revenue exceed what friction costs the company once everything that friction produces is counted?" That includes the support calls it generates the chargebacks it pushes customers into and the expected cost of getting caught weighted by the probability of that happening and the amount the resulting fine tends to be. Most of the companies that generated these flows did the first half of that calculation. Fewer did it in the second half and I think that's exactly where the problem lies.error

A Worked Example: Pricing the Friction

Let me build this with round illustrative numbers not actual figures from a single real company

Let's say a subscription service has 2,000,000 subscribers who pay $15 a month and its gross margin on that revenue is 70 percent so each subscriber who stays one more month generates $10.50 of contribution margin. Let's say that in a given month 100,000 subscribers 5 percent of the base really want to cancel. Call this true churn intention. It doesn't change based on how difficult it is.make it cancel. What changes is how many of those 100,000 actually succeed

In an easy-churn world all 100,000 leave so the measured monthly churn rate is 5 percent. Now add friction: a retention flow that only lets 80,000 of them through and silently keeps 20,000 paying another month. The measured churn drops to 4 percent which looks like a retention gain on any dashboard

How much is that worth? Twenty thousand additional subscribers retained in a given month each worth $10.50 of contribution margin is 20,000 times 10.50 or $210,000 of additional margin that month. In a year that's 210,000 times 12 or $2,520,000. That's the beneficial part of the ledger and it's real money

Now the cost side using the same illustrative basis. Let's say half of the 20,000 people trapped each month 10,000 of them make an extra support call because the flow frustrates them at $12 per call to staff. That's 10,000 times 12 or $120,000 a month $1,440,000 a year. Let's say 5 percentOf the 20,000 people stuck or 1,000 per month they forgo the flow entirely and simply charge back through their card issuer instead of finalizing the cancellation each chargeback costs the merchant a $25 fee plus the $15 in revenue they recover $40 total so 1,000 times 40 is $40,000 a month $480,000per year. Then add the expected cost of enforcement: let's call it a 5 percent chance in a given year that a flow like this will trigger regulatory action leading to a $50 million settlement. Five percent of $50,000,000 equals $2,500,000 a year in expected value although most years nothing happens at all

Add up the three costs: 1,440,000 plus 480,000 plus 2,500,000 is $4,420,000 a year. Compare that to the $2,520,000 profit. The friction in these illustrative numbers isn't even worth it. It costs $4.42 million a year to generate $2.52 million of retained margin a loss of about$1.9 million a year and that's not counting a single dollar of reputational damage that falls completely outside of this arithmetic

I want to be honest about what this example shows and doesn't show. It doesn't prove that all companies that add friction are losing money. Change the churn intention to 10 percent instead of 5 or reduce the expected regulatory cost to almost zero because enforcement in a given market is rare and friction can look profitable again quite easily. What the exercise shows is that the answer depends entirely on assumptions that are really hard to pin down and that companies that only used the profit side of thismath the retained income line without ever pricing the other side they were doing something closer to hope than modeling

The Legal Position

The application of the law in this area was developed on a simple theory: that making cancellation much more difficult than registration is itself an unfair or deceptive practice regardless of anything else the company has done

Legislation governing online subscriptions generally requires clear disclosure of terms before the customer is charged informed consent to that charge and a simple mechanism for canceling the charge

Several states passed specific rules that a subscription accepted online must also be able to be canceled online directly targeting the phone call requirement

Federal regulation approached the same practices from a different angle: Cancellation must be at least as simple as enrollment consent for recurring charges must be obtained separately from consent for other terms and material terms must be clearly disclosed before the customer pays anything

Enforcement actions have produced large settlements against major subscription companies generally focused on the difficulty of canceling rather than the underlying subscription itself

The Free Trial Interaction

These practices are combined with what regulators call a negative option A free trial that automatically converts to a paid subscription unless the customer actively cancels first

That structure is legal on its own. It requires clear disclosure of the conversion the price after conversion and the cancellation deadline before the customer signs up

The recognizable error pattern is the fine print in the conversion terms no reminder before the charge arrives and a cancellation process that takes longer to complete than the remaining time of the trial

Several jurisdictions now require a reminder before it becomes a trial. I think it is the most effective remedy of anything discussed here because it fixes forgetfulness which is the actual failure mode rather than fixing friction which is a separate overlapping problem

Case Study: Amazon's Prime Cancellation Flow

The clearest recent example of this exact mechanism is the FTC's case against Amazon over how customers canceled a Prime membership. The complaint filed in 2023 alleged that Amazon's internal name for the cancellation process was "Iliad" a reference the employees themselves allegedly made to an epic poem due to the length of the stream relative to the one or two clicks required to join

The core of the complaint was structural not about a single dark pattern trick. Signing up for Prime required one or two clicks. Cancellation reportedly required browsing several pages each of which presented another offer or another question designed to redirect the customer before they reached an actual cancellation button. That's friction as a multi-step designed funnel which is exactly the mechanism the previous section modeled in the abstract

The case was eventually resolved in a settlement that combined a civil penalty with money returned to affected subscribers. My recollection is that the total was billions of dollars about two and a half billion when you add in the fine and consumer compensation although I would treat that exact figure as something to compare to the FTC's own publication rather than relying on memory

Where This Breaks

I've described friction as a calculation that companies get wrong. The fair reaction is that I assume that law enforcement is safer than it really is and that there is real evidence of that reaction

The Federal Trade Commission finalized a rule in 2024 generally called click-to-cancel that requires ending a negative option subscription to be no more difficult than starting one in most of the market at a time and not on a case-by-case basis. A federal appeals court struck down that rule in 2025 on procedural grounds ruling that the commission had skipped a preliminary analysis step it was required to complete before issuing such a broad rule. That's not a ruling that friction is legal.a decision that this particular attempt to ban it broadly was constructed in the wrong way. But the practical effect is the same: the expected regulatory cost I used in the example above a broad rule that automatically catches all violators became smaller and more complicated returning to a world of case-by-case enforcement that primarily hits the largest and most visible violators

There is a second place where the model breaks down and cuts in the other direction. Some frictions are not misleading at all. A subscription tied to a physical product an insurance policy or a service with real risk of fraud sometimes has a legitimate reason to confirm identity or intent by phone before canceling and treating every telephone requirement as bad faith would be a mistake. The distinction that regulators really make I think correctly is not telephone versus internet. It is about whether the cancellation path is more difficult than the enrollment path for no reason.related to the product itself

The third point this breaks is scale. My practical example assumed a company large enough that regulatory action is even a significant possibility. A small subscription business with ten thousand customers is not on any regulator's radar in the same way so the expected implementation cost in that version of the model hovers around zero and the friction may appear purely profitable with nothing on the other side of the ledger to compare it to. If the model in this article applies primarily to companies large enough to make it worthwhilesue them this is a real limitation not a footnote

What a Consumer Can Actually Do

It is worth knowing the practical remedies in this case regardless of how the regulatory landscape develops

When a subscription was purchased through a mobile app store cancellation is usually done through that platform's account settings rather than directly through the merchant and this route is usually the easiest available

Card issuers can prevent a specific merchant from accepting additional payments and a virtual card number with a spending limit or expiration date structurally prevents recurring charges without requiring merchant cooperation

A fee charged after a cancellation request was requested is debatable and any record of that cancellation request you can present serves as evidence

That last option matters more than it seems. It returns the cost of a difficult cancellation flow to the merchant since chargebacks carry fees and count against the merchant's standing with the card networks which is precisely the cost line built into the example worked above

How I Actually Think About This

My reading having constructed the above arithmetic for myself is that friction is best understood as a gamble with a reward that is easy to see and a cost that is easy to undervalue because the cost shows up late in a different department's budget and often in a year when nothing at all happens

When I look at any subscription business the first question I ask is not whether churn is difficult since almost all companies allow some friction to exist. The question is whether the friction seems designed around the signup flow and whether the company discloses churn in a way that allows an outsider to distinguish between churn driven by product dissatisfaction and churn suppressed by the process. A company that publishes voluntary churn rates separately from involuntary churnlike failed payments it's telling you something honest. A company that only reports a combined retention number is one I read with more skepticism not because the number is made up but because it's exactly the number that a friction strategy is designed to inflate

I admit that the regulatory aspect of this is the part that I find really difficult to model with confidence. Enforcement priorities change with administrations court rulings like the one that overturned click to cancel can reset the baseline overnight and the actual probability of a given company getting caught is not something I have a good way to estimate from outside the company. What I do trust is the direction of the argument even when I can't pinpoint its magnitude: the friction that only survives because it is currently unlikelyGetting caught is a friction that's one compliance cycle away from being costly and a management team that's actually done this calculation rather than just watching the retention number go up is the one I'd bet on in the long run

I also try to apply this to my own subscriptions before writing about others. I caught myself staying subscribed to something I no longer used for months not because I loved the retention offer but because I never opened the tab to cancel it. This is present bias and status quo bias that does exactly what was described in the previous section and knowing the mechanism by name hasn't made me immune to it which is a bit embarrassing to admit in an article about how these systems exploit behavior.human

The Bottom Line

Making expungement more difficult than registration increases retention and it does so because it's a deliberate spreadsheet decision not an accident of product design. The numbers worked out above show why that decision is more fragile than it seems: the benefit is a clear visible number on the retained revenue line while the cost is hidden among support budgets chargeback fees and an expected regulatory cost that depends entirely on assumptions about law enforcement that can change with a single court ruling.Amazon's Prime case shows the mechanism at full scale and the overridden click-to-cancel rule shows that even regulatory support is less certain than a company modeling this in 2024 might have assumed. My own conclusion is that friction is not free it is undervalued and the most exposed companies are the ones that only measured half the calculation that made them look good