Amazon Prime: The Exit Named After a War

What happened

Amazon built a Prime cancellation process it named internally after the Iliad, a poem about a ten-year war. The FTC sued in June 2023, and the charge was not that Prime had been oversold but that leaving it was deliberately hard. An amended complaint that September put a number on the asymmetry: cancelling took four pages, six clicks and fifteen options, against the clicks it took to join. The court refused to dismiss the case in May 2024, which meant the internal documents were going to be read aloud, and in June 2025 it granted sanctions against Amazon over the discovery fight.

The case, start to finish

The team's own name for the cancellation flow was Iliad, after the poem about a ten-year war.

A membership worth far more than its price

Prime is not a product people consume, it is a membership that changes how they shop. The member pays a flat fee, stops comparing shipping costs, and buys more often, so the subscription is worth considerably more to the seller than the fee itself. Amazon reports subscription services net sales separately: $40,209M in 2023, $44,374M in 2024, $49,619M in 2025.

That is precisely what makes cancellation friction look profitable on a dashboard. A saved cancellation is not one saved fee. It is a year of changed purchasing behavior, which means the measured value of keeping somebody in is very large and shows up quickly. The cost of being a company people cannot quit is diffuse, slow, and eventually arrives as a legal event rather than a marketing one.

What the FTC alleged, and what an allegation is

In June 2023 the Federal Trade Commission sued Amazon, later adding two executives as defendants, charging violations of the FTC Act and the Restore Online Shoppers' Confidence Act. Allegations in a complaint are allegations, and it is worth holding that distinction while reading them.

The core allegation was not that Prime was oversold. It was that the interface made one direction easy and the other hard. According to the amended complaint, a consumer could enroll in one or two clicks, while cancelling ran through a four-page, six-click, fifteen-option process the company called Iliad internally. Reaching that process at all meant finding the eleventh option in the third column of a dropdown menu. The complaint says the flow launched in 2016 and was not substantially changed in the United States until about April 2023, and that even the revamped version still took five clicks on desktop and six on mobile.

The decision nobody quite made

What makes this instructive is that there is no meeting in the story where somebody decides to trap customers. The FTC's allegation is that Amazon declined, repeatedly and in the ordinary way, to change something that was performing, because the changes that would have made cancelling easier hurt the numbers.

That mechanism deserves to be taken seriously, because it needs no villain. Cancellation-save rate is measurable within a week and shows up in a review. Resentment is not measurable at all, and the customer who finally escapes does not fill in a survey explaining why they will never come back. A retention metric that rewards friction simply never charges for what friction produces, so the flow keeps winning its own review, for seven years.

The bill, and the sentence it had to ban

A stipulated order in September 2025 settled it: a $1 billion civil penalty, the largest ever in a case involving violation of an FTC rule, plus $1.5 billion in consumer redress covering an estimated 35 million people. That is roughly $43 each, about three months of a membership at the $14.99 monthly price cited in the complaint. Amazon also agreed to a redesign monitored by a paid independent third party, with clear disclosure of price and renewal terms, a clear button to decline, and cancellation by the same method used to sign up.

The order had to name a specific sentence that decline button may no longer say: "No, I don't want Free Shipping." The option to decline had existed all along. It had simply been written so that choosing it felt like choosing against yourself.

Set $2.5 billion against $49,619M of 2025 subscription revenue and it is about eighteen days' worth, which is the honest way to read the penalty. This did not threaten the business. It put a price on a design decision that had been free for seven years, and it took that decision away from the company.

The test that works without a lawyer in the room

The useful rule is symmetry, and applying it requires no legal training. However many clicks it took to join, cancelling should take no more, through the same channel. Somebody who signed up in a browser at midnight should not have to reach a human in business hours, which is the exact asymmetry ROSCA is written about. A pause option offered alongside it, however genuinely useful, does not undo what moving the exit does.

You will lose some measured saves doing this. Size that loss honestly rather than pretending it is free. What you buy is a churn number that means what it says, a save rate you could explain in a deposition, and the ability to win somebody back later, which is close to impossible with a person who remembers escaping you.

Timeline

• 2016 Amazon launches the Prime cancellation process it calls, internally, "Iliad", after the poem about a ten-year war. According to the FTC, it is not substantially changed in the United States until about April 2023.
• June 21, 2023 The FTC sues, charging violations of the FTC Act and the Restore Online Shoppers' Confidence Act. The core allegation is not that Prime was oversold but that the interface was built to make one direction easy and the other hard. Prime is $14.99 a month at the time.
• September 20, 2023 An amended complaint adds senior executives as defendants and puts a number on the asymmetry: the Iliad Flow required a four-page, six-click, fifteen-option process to cancel, while a consumer could enroll in one or two clicks. Reaching the flow at all meant finding the eleventh option in the third column of a dropdown menu.
• May 28, 2024 The court denies the defendants' motions to dismiss. The case is going to trial, which means the internal documents are going to be read aloud.
• June 23, 2025 The court grants the FTC's motion for sanctions against Amazon. Whatever the merits, the discovery fight has now become part of the story the jury would hear.
• July 8, 2025 In a separate case, the Eighth Circuit vacates the FTC's new Negative Option (“click-to-cancel”) Rule in its entirety on procedural grounds, days before its compliance date. The specific rule disappeared. ROSCA and Section 5, which is what Amazon was actually charged under, did not.
• September 25, 2025 Stipulated order. Amazon pays a $1 billion civil penalty, the largest ever in a case involving violation of an FTC rule, plus $1.5 billion in consumer redress covering an estimated 35 million people. It also agrees to a redesign: a clear button to decline Prime, clear disclosure of price and renewal terms, and cancellation by the same method used to sign up, monitored by a paid independent third party.

You're in the owner's chair

You run a subscription with rising cancellations. Your growth team has measured a four-step save flow, with an offer at each step, that keeps 18% of people who click Cancel. Legal has no objection today. What do you ship?

• Cancel in one click, one offer shown once, and a confirmation email
• Keep the flow but add a “pause my membership” option, and route cancellation through a chat agent who can help
• Ship the four-step save flow — it is measured, it works, and nobody is being lied to

Clicks to get in, clicks to get out (per the FTC amended complaint)

• Enroll in Prime: 2 clicks
• Cancel via the Iliad Flow (2016 – April 2023): 6 clicks
• Cancel after the 2023 revamp — desktop: 5 clicks
• Cancel after the 2023 revamp — mobile: 6 clicks

The complaint states consumers could enroll in “one or two clicks”, charted here at two as the generous reading, against a four-page, six-click, fifteen-option cancellation process. It also alleges the 2023 revamp left five clicks on desktop and six on mobile, and left the flow hard to find. The gap between the first bar and the rest is not friction that accumulated by accident; the FTC's case is that it was defended on purpose.

Business model

Prime is a membership that changes how people shop rather than a product they consume. The member pays a flat fee, stops comparing shipping costs, and buys more often, so the subscription is worth far more than its own price, and every retained month compounds into retail volume. That is precisely what makes cancellation friction look profitable on a dashboard: a saved cancellation is not one saved fee, it is a year of changed purchasing behavior.

Revenue model

Amazon reports subscription services net sales (Prime plus digital content) separately: $40,209M in 2023, $44,374M in 2024, $49,619M in 2025. That is the machine the cancellation flow was protecting. It is also the number that puts the penalty in proportion: the entire $2.5 billion settlement is about eighteen days of Amazon's 2025 subscription-services revenue.

Cost structure

The friction itself was nearly free to build (pages, buttons, an offer or two), which is exactly why this pattern spreads. The costs were all deferred and all landed at once: $2.5 billion in money, a court-supervised redesign of the enrollment and cancellation flows, a sanctions ruling, executives named personally in a federal complaint, and a permanent public association between the brand and a cancellation process its own staff named after a ten-year siege.

Strategic challenge

Every retention metric said the flow worked. Cancellation-save rate is measurable in a week; the cost of being the company people cannot quit is measurable in years, and lands as a legal event rather than a marketing one. The FTC's specific allegation is that Amazon knew: that leadership slowed or rejected changes that would have made cancelling easier because those changes hurt the numbers. That is the whole case in one sentence: the metric was real, and following it was the mistake.

Key decision

The decision was made repeatedly, in the ordinary way, by declining to change something that was performing. The FTC alleges Amazon revamped the flow in about April 2023, seven years after launch and shortly before the complaint was filed, which locates the decision precisely. It was not that nobody could fix it. It was that nothing until an investigation made the fix worth its cost in saved cancellations.

What worked

As a retention mechanism, on its own terms, for seven years. Nothing in the record suggests the Iliad Flow failed to keep subscriptions alive; the FTC's complaint is built on the premise that it worked, and that Amazon's leadership knew it worked and protected it for that reason. Amazon's subscription-services revenue kept compounding throughout, at $40.2B, $44.4B and $49.6B across 2023 to 2025, which is the uncomfortable part. Friction is not a failed tactic. It is a tactic that succeeds until it is priced.

What failed

The accounting. A save that a customer experiences as a trap is a liability that has not been booked yet, and here it was booked all at once: $1B in civil penalty, $1.5B in refunds to an estimated 35 million people, roughly $43 each, or about three months of the membership many of them did not mean to buy. The settlement then took the design decision away entirely. Amazon may no longer offer a decline option worded as a punishment; the order specifically forbids a button reading “No, I don't want Free Shipping.” The FTC also alleged that even the 2023 revamp still required five clicks on desktop and six on mobile, which is a fair description of a company fixing the letter of the complaint.

Risk factors

Retention metrics that reward friction and never charge for resentment; save flows that use a discount to convert an intent to leave into an unbooked liability; cancellation available only through a channel the customer did not sign up through, such as a phone call or a chat agent, which is the specific asymmetry ROSCA is written about; state automatic-renewal laws in California and New York that bind regardless of federal rulemaking; and the discovery risk unique to this pattern, that the internal Slack message describing the tactic honestly becomes the plaintiff's best exhibit.

Lesson summary

Friction that keeps a subscriber does not convert them; it stores their intention to leave and charges you interest on it. Amazon's cancellation flow worked for seven years, and the bill was $2.5 billion, a court-supervised redesign, and a permanent story about a company that made quitting a siege. The honest test is symmetry: if signing up takes one click, cancelling should take one click. Retention you have to trap people into is churn with a delay.

Key data

• $1B — largest ever for an FTC rule violation Civil penalty
• $1.5B Consumer redress
• ~35 million Consumers the FTC estimated were affected
• ~$43 Average redress per affected consumer
• 1–2 vs. 6 Clicks to enroll vs. clicks to cancel
• 2016 to about April 2023 Years the Iliad Flow ran unchanged in the US
• $14.99/month Prime price cited in the complaint
• $49,619M Amazon subscription-services net sales, 2025

Sources & basis

The company here is real and named, and nothing about it was invented to make the story land. The list below is where each fact came from — public filings, court records, published reporting — so you can open a source and check it against the sentence that used it.

1. FTC, “FTC Takes Action Against Amazon for Enrolling Consumers in Amazon Prime Without Consent and Sabotaging Their Attempts to Cancel” (June 21, 2023) — the original complaint, the ROSCA and FTC Act charges, the $14.99/month price, and the “Iliad” naming View source ↗
2. FTC v. Amazon.com, Inc., Amended Complaint (filed September 20, 2023, public redacted version) — paragraph 128 for the four-page, six-click, fifteen-option Iliad Flow against one-or-two-click enrollment; paragraph 130 for the 2016 launch and April 2023 revamp; paragraph 257 for the five-clicks-desktop / six-clicks-mobile revamped flow View source ↗
3. FTC, “FTC Secures Historic $2.5 Billion Settlement Against Amazon” (September 25, 2025) — the $1B civil penalty, $1.5B in redress, the ~35 million affected consumers, the banned “No, I don't want Free Shipping” button, and the same-method cancellation requirement View source ↗
4. FTC case page, Amazon.com, Inc. (ROSCA), FTC v. — full docket including the May 28, 2024 order denying the motions to dismiss and the June 23, 2025 order granting the FTC’s motion for sanctions View source ↗
5. Amazon.com, Inc., Form 10-K for fiscal 2025 (SEC EDGAR) — subscription services net sales of $40,209M (2023), $44,374M (2024) and $49,619M (2025) View source ↗
6. Cooley LLP client alert (July 11, 2025) on Custom Communications, Inc. v. FTC — the Eighth Circuit vacating the FTC’s Negative Option “click-to-cancel” Rule in its entirety on procedural grounds, and the continuing application of ROSCA and FTC Act Section 5 View source ↗