Fair-treatment traps in pricing, cancellation and free-trial promotions

Amazon's internal codename for the Prime cancellation flow was "Iliad", after the long journey. The FTC sued over it in June 2023; in September 2025 Amazon settled for $2.5 billion, split as a $1 billion civil penalty and $1.5 billion in refunds to affected subscribers. Look at the split before you look at the total. The refund half is bigger than the fine, and it was sized by years of accumulated enrolments, not by whatever the sign-up page happened to look like on the day the complaint landed.

This lesson stays on the commercial terms themselves: trial-to-paid rollovers, intro pricing that steps up, renewal disclosure and cancellation friction. Which body can act, with what powers, and how a complaint becomes a ruling is the ground the watchdogs lesson covers.

Why media subscriptions sit in the blast radius

SVOD (subscription video-on-demand), music streaming, news paywalls and games storefronts run the same four mechanics, and each carries its own failure mode:

• Free trial: charging before the trial ends, or leaving the trial-to-paid moment unclear.
• Discounted intro pricing: the step-up date and the full price sit somewhere other than where the offer price sits.
• Auto-renewal: continuous billing without express consent and an easy exit.
• Cancellation flow: more steps, more menus and more persuasion to leave than to join.

Dark pattern: a design choice built to push users toward a decision they would not otherwise make. Confusing button hierarchy, pre-ticked boxes, buried opt-outs, forced continuity. The FTC's 2023 action against Epic Games put $245 million in refunds against exactly this in Fortnite: a purchase interface where a single unintended press could charge a saved card, and account locks for players who disputed charges with their bank. Games are media, and that purchase logic is the same logic sitting in your upgrade and add-on paths.

The rules that bite on commercial terms

In the US, negative-option billing runs through ROSCA (Restore Online Shoppers' Confidence Act, 2010): clear disclosure of terms before payment, express informed consent, simple cancellation. The FTC's 2024 click-to-cancel rule would have codified "as easy to cancel as to join" across sectors; the Eighth Circuit vacated it in July 2025 on procedural grounds, weeks before it bit. ROSCA itself is untouched, and California and New York run their own auto-renewal statutes with comparable requirements. Planning on the assumption that the vacated rule removed the obligation is a mistake teams have already made.

In the EU, the Consumer Rights Directive and the Unfair Commercial Practices Directive set the baseline on pre-contract information and misleading omissions, and the Commission's digital fairness work has dark patterns and subscription traps squarely in scope. In the UK, the DMCCA 2024 lets the CMA fine up to 10% of global turnover for consumer law breaches without going to court first.

Adobe is the case to study if you sell annual commitments billed monthly, a structure now common in media bundles and games passes. In June 2024 the Department of Justice, on FTC referral, sued Adobe over its "annual, paid monthly" plan: an early termination fee worth a large share of the remaining year, disclosed behind a small link rather than on the plan selector, and a cancellation path that dropped users between steps. Two executives were named personally in the complaint. That is the detail worth carrying into your own risk conversation.

The pre-launch audit

Run this before any pricing page, trial offer or cancellation flow ships. It belongs inside the gate the sign-off lesson maps out, owned by marketing rather than handed to legal at the end.

1. Price transparency

• Is the full post-promo price visible before checkout, at the same visual weight as the offer price?
• Is the step-up date stated at sign-up, not only in a linked T&Cs page?
• If the price rises mid-contract, does the notice give real time to leave before the new rate applies?

2. Consent and billing

• Does paid billing start on an affirmative action, never a pre-ticked box?
• Does a reminder go out before the first charge, with enough lead time to cancel and act on it?

3. Cancellation parity

• Count the steps to subscribe. Count the steps to cancel. A self-serve online sign-up paired with a phone-only cancellation is the pattern regulators reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → for first.
• Can a retention offer delay completion, or is there a "cancel anyway" route at every step? Retention offers are not banned. Blocking the exit with them is.
• Edge case: if the subscription was sold through Apple or Google in-app billing, cancellation lives in the store, not your app. You still wrote the offer copy, and you still receive the complaint.

4. Visual and language design

• Is "Keep my plan" a bright button while "Cancel" is a grey text link? That asymmetry is the flag.
• Countdown timers, "3 people just subscribed" prompts and limited-time pricing must be true and evidenced, or they are deceptive claims.

A simple internal scoring tool:

Audit item                          Pass/Fail   Evidence link
--------------------------------------------------------------
Full price shown pre-checkout          [ ]        screenshot
Renewal date/price disclosed           [ ]        screenshot
Cancel steps <= Signup steps           [ ]        step count log
No pre-ticked consent boxes            [ ]        screenshot
Retention offer has "cancel anyway"    [ ]        screenshot
Urgency/scarcity claims verifiable     [ ]        data source

Run it per region: US, UK and EU thresholds differ, and a flow cleared in one market can fail in another on notice periods alone.

For a practical external reference, the CMA's guidance on subscription contracts is a genuinely useful, free primer: CMA guidance for businesses on subscription contracts

What a clean audit still misses

A page that passes today does not clear what you shipped in 2021. Three second-order problems:

Cohort exposure. Redress is calculated across everyone ever enrolled under the defective flow, so the bill scales with your growth history. Amazon's $1.5 billion refund pot dwarfed its $1 billion penalty for that reason. A mid-size streamer that ran an ambiguous trial rollover through two years of aggressive acquisition carries a liability that has nothing to do with its current UX.

Evidence you cannot produce. If you cannot show the exact screen a 2022 cohort saw, including the mobile web variant and the A/B arm that lost, you cannot defend the enrolment. Keep timestamped screenshots of every checkout and cancellation variant, tied to the date range each was live, in a store that outlives the growth team that built it. Most companies discover this gap only when asked.

The churn bill for fixing it. Removing friction from cancellation loses subscribers, and the loss lands in the quarter you fix it while the avoided penalty never shows up in any dashboard. Forecast that hit before you ask for the change, or the change will be deferred every planning cycle until a regulator makes the decision for you.

The counter-case is worth stating plainly: parity does not mean surrendering retention. You can still offer a pause, a downgrade or a discount inside the cancellation flow. You cannot make those offers the only route through it, and you cannot hide the exit behind them.

🎬 [VIDEO: "The FTC's Click-to-Cancel Rule Explained" - youtube.com - search for FTC or consumer-law explainer channels covering the click-to-cancel rule and ROSCA basics for a plain-language regulatory walkthrough]

Key Takeaways

• The Amazon Prime settlement was $2.5 billion, of which $1.5 billion was consumer refunds. Redress scales with historic enrolments, so old flows cost more than current ones.
• The test is parity and clarity: cancelling should be no harder than joining, and renewal terms should be as visible as the promotional price.
• ROSCA obligations survived the Eighth Circuit vacating the click-to-cancel rule in July 2025, and state auto-renewal statutes apply regardless.
• Annual-commit pricing billed monthly is the sharpest trap in media bundles. The Adobe complaint targeted a hidden early termination fee and named executives personally.
• Keep timestamped screenshots of every pricing, trial and cancellation variant, tied to its live date range. Without them you cannot defend an old cohort.
• Model the churn cost of fixing a flow before you propose the fix, or it gets deferred until enforcement decides for you.