Inside the "Iliad" Flow: What Amazon's Cancellation Design Reveals About Retention Metrics vs. Exit Design
Meta description: A four-page, six-click, fifteen-option cancellation sequence didn't happen by accident — it happened because a retention metric and a simplicity proposal kept losing to each other in the same review process, for years. A design-and-governance breakdown for subscription businesses.
The FTC's complaint against Amazon, settled in 2025 for $2.5 billion — the largest civil penalty the agency has ever obtained for a rule violation — centers on a single flow: the sequence a Prime subscriber had to complete to cancel. Internally, according to the complaint, that flow was named "Iliad," after Homer's epic about the decade-long Trojan War. The complaint describes it as four pages, six clicks, and fifteen options, with discount offers and benefit reminders interspersed along the way. (Amazon settled without admitting or denying the FTC's allegations, which is standard practice in FTC consent orders; what follows is drawn from the complaint's allegations and the public settlement record, not a court's adjudicated findings on contested facts.)
What makes this case worth a full breakdown, rather than a line in a table, is that a subscription cancellation flow of this complexity is never the output of one decision. It's the accumulated output of a business model, a metric, and a review process — each individually defensible — compounding over time. That's the more useful lesson for anyone building or scaling a subscription product, because the business model, the metric, and the review process are the same three ingredients in nearly every subscription company, regulated or not.
The environment that produces a flow like this
Prime's core economics make retention unusually central to the product's justification. Free or discounted shipping is a loss-leader; the subscription only pays for itself across a large base of members who stay subscribed long enough, and often enough, to make the logistics investment worthwhile. In a business built this way, "subscribers who cancel in month one" isn't just a churn number on a dashboard — it's a number that speaks directly to whether the underlying economics work at all. That's the environment: a retention rate that isn't one KPI among many, but close to the KPI the entire loss-leader logic depends on.
Put a metric in that position, inside any large organization, and proposals that would move it in the wrong direction face a structural headwind by default — not because anyone reviewing them is acting in bad faith, but because the proposal is asking the organization to accept a worse number on the metric closest to the business model's core assumption. The FTC's complaint alleges that proposals to simplify the Prime cancellation flow were repeatedly brought forward internally and not approved, and cites a draft memo stating that clarifying the enrollment process was not the "right approach" because it would cause a "shock" to business performance. Read charitably, that's not a story about someone deciding to trap users. It's a story about a retention number that had become load-bearing enough that a proposal to reduce friction had to clear a bar that "this will make the product harder to sell to leadership" made progressively harder to clear, review after review.
What the design actually did
Three elements of the Iliad flow, as described in the complaint, are worth separating out, because each maps to a distinct design lever rather than a single trick.
Sequential friction. Four pages and six clicks is a length decision, not a single dark pattern — it's the accumulation of many individually-plausible "let's confirm the user really wants this" steps. Each step is defensible in isolation (confirming intent, surfacing an alternative, checking for account issues); the effect only becomes clearly disproportionate in aggregate, which is exactly why length-based friction is harder to catch in a single design review than an obvious deceptive claim would be. Nobody signs off on "make cancellation absurdly long." Reviewers sign off on one more step, one more offer, one more confirmation — each of which passed review on its own merits.
Interstitial offers as a persuasion lever, not just a delay. The discount offers and benefit reminders inserted into the flow are a specific behavioral mechanism: reintroducing loss aversion at the exact moment a user is about to give something up. This is a standard, effective retention tactic used constructively across subscription products all the time — the same mechanic that makes a win-back offer work. The design question isn't whether persuasive retention offers are legitimate (they are); it's whether they're placed inside a mandatory sequence a user cannot bypass, versus offered as a genuine, skippable choice. Placement inside a forced sequence is what converts a retention offer from persuasion into an obstacle.
Channel routing. The complaint alleges that Amazon had a policy directing customer service agents to route callers who wanted to cancel toward the online Iliad flow, despite agents having the technical ability to process the cancellation directly during the call. This is a process decision, not a UI decision, and it's the detail that generalizes furthest beyond this specific case: a company can build one clean self-service cancellation path and still produce the same outcome as a manipulative flow, if the operational policy around that path routes users away from the easiest version of it.
The behavioral principles doing the work
None of these levers require inventing anything new. They're standard behavioral-design tools, applied here to an exit rather than an entry:
- Loss aversion, deployed via the interstitial offers — the same mechanic that makes "you'll lose your progress" warnings effective in onboarding, redirected toward retention.
- Status quo bias and effort heuristics, deployed via sequence length — the well-documented finding that added steps reduce completion regardless of each step's individual reasonableness, because users weigh the cumulative effort of a task before starting it, not just each step in isolation.
- Choice overload, deployed via the fifteen options — presenting many branches at a decision point is a documented way to slow and confuse a user even when every individual branch is a legitimate choice, because evaluating branches itself becomes the cost.
Each of these is a tool used constructively somewhere else in the same product — loss aversion sells upgrades, thoughtful step sequencing improves onboarding completion, and structured choice helps users self-select the right plan. The Iliad case isn't a story about newly invented manipulation tactics. It's a story about the same toolkit, pointed at an exit instead of an entry, without a design principle in place to flag that the direction had changed.
What this means for subscription product and growth teams
The generalizable lesson isn't "make cancellation a single click regardless of context" — a company can legitimately want to understand why a user is leaving, offer a real alternative, or confirm intent once. The generalizable lesson is that a subscription business needs an explicit design principle for exit flows the same way it has one for onboarding flows, because without one, the same incremental-approval process that improves an onboarding flow over years will, by default, also lengthen a cancellation flow over years — the review process doesn't distinguish between the two unless a principle tells it to.
Two things a product, growth, or legal team can build into that review process directly:
- An exit-symmetry budget. Set the maximum acceptable step/click count for cancellation as a fixed ratio of the sign-up flow's step count — say, no more than 1.5x — and treat any change that would push cancellation past that ratio as requiring the same sign-off tier as a pricing change, not a routine UX tweak.
- A single-offer rule for retention interstitials. Retention offers inside a cancellation flow are legitimate; an unlimited or branching sequence of them is what turns a fair chance to reconsider into an obstacle course. Capping it at one clearly-skippable offer preserves the legitimate retention use case while removing the design lever that scales into something like Iliad over successive "just one more offer" approvals.
Both are cheap to implement and don't require predicting every future compliance question — they require the organization to decide, once, that the review bar for adding friction to an exit is not the same as the review bar for adding friction to any other flow, and then holding that line by policy rather than by relying on each individual reviewer to catch it.
This is part of a series on the design and decision-making behind marketing-compliance enforcement actions. See the full series or get in touch if you're building or auditing a subscription retention program and want an outside read on where exit design might be quietly drifting.