The Rules Behind the Retention Offer
Earlier this year, PCWorld reviewed the discounts streaming services put in front of subscribers who try to cancel. Some Peacock users reported being offered six months of Premium for $2.99 a month, others were offered Premium for $7.99, and some offers disappeared after a browser refresh.[1] PCWorld did not report why the offers differed; it may have been timing, testing or something else entirely. Readers could reasonably conclude that the cancel button works as a negotiation. Regulators in several consumer markets now want businesses to account for gaps like this one, where two customers on what appeared to be the same plan were quoted different prices.
As Streaming Media's Tyler Nesler wrote, "Preventing subscriber churn is the oldest problem in the streaming subscription business."[2] Cost has since become the main reason people leave. Parks Associates found that 30% of U.S. consumers cite cutting household expenses as their top reason for canceling a streaming service.[3] It also found that lower-priced ad-supported plans are the strongest lever for keeping or winning back subscribers,[4] which fits the Antenna data cited in this publication's State of Streaming Monetization 2026: "Some 71% of Net Additions in the past 9 quarters came from ad plans."[5] Services have answered with discounted downgrades, free months and win-back pricing. Once subscriber data is used to estimate what one person will pay, those offers start to look like the personalized pricing the FTC has begun to examine. That may expose services to scrutiny, although how far the FTC's approach reaches into cancellation flows is unsettled.
What the FTC has proposed
On Aug. 19, the Federal Trade Commission released a proposed enforcement policy statement on personalized pricing. It says that "Where consumers reasonably expect that prices for a product or service will not vary based on their personal data," a business should "clearly and conspicuously disclose not just that the price is personalized" but also the basis for it and the types of data used. The Commission says failing to do so "is likely to constitute an unfair or deceptive act or practice." Its example of an adequate disclosure is a price "based on a consumer's estimated willingness to pay" derived from the consumer's purchases, provided the disclosure is accurate and complete; telling a consumer only that a price is "specially selected" would likely mislead.[6]
The statement is still a proposal, and the comment period closed Sept. 25.[7] It does not mention streaming or cancellation offers. Many subscribers already expect long tenure to earn a better deal when they try to leave, so it is an open question whether they reasonably expect retention offers not to vary. Plan pages are another matter, because most viewers assume the price listed there is the same for everyone. Scrutiny is more likely to land on a service that uses behavioral data to estimate willingness to pay when its subscribers do not know it does.
State and city rules
New York's disclosure law, in force since Nov. 10, 2025, covers less of this than its wording suggests: it requires businesses that set prices algorithmically from a consumer's personal data to display "THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA," but it exempts a price offered to someone with an existing subscription when that price is lower than the price for the same good or service under the subscription.[8] A discount on the plan a subscriber already has may therefore not need the label. Personalized sign-up pricing can still be covered, and an offer that moves a subscriber to a different tier needs closer analysis.
California's amended automatic renewal law, in effect since July 1, 2025, puts more immediate limits on retention offers and addresses them by name.[9] On the phone, a business may make a discounted retention offer only after telling the consumer that cancellation can be completed at any time by saying "cancel." Online, a retention offer may appear only if a prominently placed "click to cancel" link or button stays displayed alongside it. New York City's click-to-cancel rule took effect Oct. 1, 2026. It requires cancellation to be as simple as enrollment and bars businesses from obstructing it, but it does not address retention offers directly.[10] Federally, the Eighth Circuit vacated the FTC's 2024 click-to-cancel amendments on July 8, 2025;[11] in March the agency issued an advance notice of proposed rulemaking that cites "difficult cancellation processes" and "unlawful retention tactics."[12]
The FTC proposal and the cancellation rules ask for different things: the FTC is concerned with disclosing that a price is personalized, its basis and the types of data used, where consumers reasonably expect prices not to vary with their personal data. California and New York City govern how retention offers and cancellations are handled, and neither asks why a subscriber got a particular discount. Being able to explain an offer does not, on its own, satisfy any of them. A written offer rule that gives the same result each time is evidence of how an offer was selected and what data determined it. Whether that evidence is enough under a given law is for counsel to decide.
How a churn model picks the offer
A holiday promotion that gives every new subscriber three months at half price would ordinarily fall outside the personalized-pricing question, because neither eligibility nor price depends on individual data. A churn model or customer data platform works differently. It scores each subscriber's likelihood of leaving and sensitivity to price, then decides which save offer, if any, to show.
Inputs can include viewing history, device, payment history, how often the subscriber has canceled before and when they last watched. Two subscribers on the same plan who cancel on the same day can see very different offers. When a propensity model makes that call, the service can usually retrieve the score behind each offer, but a score is not a reason it can give the subscriber. Post-hoc explanation tools are of limited help here, because different methods can give different accounts of the same output, and none of them proves which logic actually chose the offer.[13]
Writing the offer as a rule
Retention offers matter too much to churn economics for services to drop them, and most services already segment subscribers in some form. Philo's Steven Chang told Streaming Media this summer, "We segment subscribers on two dimensions: what they watch and how they watch."[14] Deterministic segmentation uses segments like these to set the offer itself. Explicit rules assign subscribers to defined segments. Where two rules could apply, a documented priority order decides, so each subscriber lands in one segment, and each segment carries a set offer relative to the published plan price. Given the same facts about a subscriber, the same rule version and the same operating conditions, such as offer budgets and expiry dates, the result is the same. Data and models can still help decide which segments are worth creating, but once a segment is live, its rule sets the offer.
For a retention offer, the record behind one segment might read:
Segment: Long-tenure ad-free subscribers canceling
Rule: Ad-free plan, 12 or more consecutive paid months, cancellation started; takes priority over the general win-back segment
Data: Billing and tenure history (first-party)
Treatment: Same ad-free plan at $5.99 a month for six months, versus the $10.99 published price
Explanation shown: "Loyalty offer for subscribers of 12 months or more"
Version: v3, effective Oct. 1, 2026
Figure 1: Three subscribers cancel the same ad-free plan. Both approaches use the same data and the same offers; under the segment rules, A and B qualify for the same $5.99 offer, and the record shows why.
Anyone who meets that rule gets the same offer, and the message on screen can state the rule. Disclosure wording has its limits, too: in an OECD lab experiment in Ireland and Chile, online disclosures had "only limited effects on consumers' ability to identify and comprehend online personalised pricing."[1] A message like this one explains eligibility, though whether it would count as an adequate disclosure under the FTC's proposal or any state law is a legal question. If a regulator, a plaintiff or the service's own review team later asks why a subscriber was offered $5.99 last spring, the record shows which segment applied, which data was used and which rule version was in force.
Coarser targeting may cost less than services expect, judging by a large field experiment on a food delivery platform's promotions, in which Walter Zhang and Sanjog Misra found that offering only five unique treatments let the firm recoup "over 99.5% of its expected incremental profits" compared with fully individualized targeting.[2] The setting was not streaming, and results will vary. The study measured profit and has nothing to say about explainability, but it suggests that a handful of well-chosen offers can keep most of what individualized targeting earns.
Rule-based offers are not a safe harbor, because a rule that relies on a sensitive inference, or on data collected or used without the disclosures, consent or other legal basis the circumstances require, can still cause legal problems, and counsel should review the criteria. What the approach adds is traceability, because the service can show what each offer was based on.
Where to start
Streaming services have spent heavily on predicting churn, yet many could not say why a given subscriber received a particular incentive. That gap is a business problem whatever happens to the FTC's proposal. Finance needs to know what each save offer costs, and customer service needs an answer when a subscriber asks why a friend got a better deal. Offer tests depend on the same record, since results mean little without a reliable log of who received what.
A practical check is to pick a subscriber who saw a retention offer six months ago and reconstruct it: the offer shown, the data that selected it and the rule or model version running that day. If that takes several teams and a model notebook, the service knows where its record keeping falls short.
The next step is an inventory of every price and offer a subscriber can see, from plan pages and sign-up promotions to cancellation flows and win-back emails, noting whether a person, a rule or a model chooses each one. Model-driven offers in cancellation flows are the obvious place to begin, since cancellation practices are already a regulatory focus, and written segment rules are one way to make those decisions reproducible.
[Editor's note: This is a contributed article from Klary. Streaming Media accepts vendor bylines based solely on their value to our readers.]
_____________________
[1] OECD, "The Effects of Online Disclosure About Personalised Pricing on Consumers: Results from a Lab Experiment in Ireland and Chile," OECD Digital Economy Papers No. 303, Jan. 18, 2021. https://www.oecd.org/en/publications/the-effects-of-online-disclosure-about-personalised-pricing-on-consumers_1ce1de63-en.html
[2] Walter W. Zhang and Sanjog Misra, "Coarse Personalization," arXiv:2204.05793 (v4, June 30, 2025). The experiment covered more than 1.2 million customers of a food delivery platform. https://arxiv.org/abs/2204.05793
[1] Ben Patterson, "Looking for streaming deals? Try hitting the cancel button," PCWorld, Jan. 28, 2026. https://www.pcworld.com/article/3040095/looking-for-streaming-deals-try-hitting-the-cancel-button.html
[2] Tyler Nesler, "Streaming's Great Rebundling and the Fight Against Churn," Streaming Media, Sept. 27, 2024. https://www.streamingmedia.com/Articles/Editorial/Short-Cuts/Streamings-Great-Rebundling-and-the-Fight-Against-Churn-166089.aspx
[3] Parks Associates, "Streaming Competition and Profitability: Pricing Models & Retention Strategies," as reported in "Ad-supported tiers are key retention tool in maturing streaming market," NewscastStudio, Feb. 12, 2026. https://www.newscaststudio.com/2026/02/12/ad-supported-tiers-are-key-retention-tool-in-maturing-streaming-market
[4] Ibid. NewscastStudio reports that lower-cost plans with advertising "ranked as the top incentive for retaining or winning back subscribers."
[5] Nadine Krefetz, "The State of Streaming Monetization 2026," Streaming Media, March 26, 2026, citing Antenna. https://www.streamingmedia.com/Articles/Editorial/Featured-Articles/The-State-of-Streaming-Monetization-2026-173815.aspx
[6] Federal Trade Commission, Proposed Enforcement Policy Statement Regarding Personalized Pricing, Aug. 19, 2026. https://www.ftc.gov/system/files/ftc_gov/pdf/p034101-ftc-enforcement-policy-statement-re-personalized-pricing-proposed-for-public-comment.pdf
[7] Federal Trade Commission, "FTC Extends Public Comment on Proposed Policy Statement Regarding Personalized Pricing," press release, Sept. 3, 2026 (extending the deadline from Sept. 18 to Sept. 25; docket FTC-2026-1057). https://www.ftc.gov/news-events/news/press-releases/2026/09/ftc-extends-public-comment-proposed-policy-statement-regarding-personalized-pricing
[8] N.Y. Gen. Bus. Law § 349-a(2) (required disclosure) and § 349-a(3)(d) (exemption for a price offered under an existing subscription-based agreement that is less than the price for the same good or service in that agreement). https://www.nysenate.gov/legislation/laws/GBS/349-A
[9] Cal. Bus. & Prof. Code § 17602, as amended by A.B. 2863 (2024), applying to contracts entered into, amended or extended on or after July 1, 2025. See Kilpatrick, "California's Latest Automatic Renewal Law Amendments Take Effect in July 2025," Oct. 2024. https://ktslaw.com/insights/alert/2024/10/california-latest-automatic-renewal-law-amendments-take-effect-in-july-2025
[10] NYC Department of Consumer and Worker Protection rule adopted July 10, 2026, effective Oct. 1, 2026. See Skadden, "New York City's Click-to-Cancel Rule: What Businesses Need to Know," Sept. 2026. https://www.skadden.com/insights/publications/2026/09/new-york-citys-click-to-cancel-rule
[11] Custom Communications, Inc. v. Federal Trade Commission, No. 24-3137 (8th Cir. July 8, 2025). See Latham & Watkins, "Eighth Circuit Vacates FTC Click-to-Cancel Rule." https://www.lw.com/en/insights/eighth-circuit-vacates-ftc-click-to-cancel-rule
[12] Federal Trade Commission, "FTC Seeks Public Comment in Response to Advance Notice of Proposed Rulemaking Regarding Negative Option," press release, March 11, 2026 (statement of Christopher Mufarrige). https://www.ftc.gov/news-events/news/press-releases/2026/03/ftc-seeks-public-comment-response-advance-notice-proposed-rulemaking-regarding-negative-option
[13] Satyapriya Krishna et al., "The Disagreement Problem in Explainable Machine Learning: A Practitioner's Perspective," Transactions on Machine Learning Research (2024), arXiv:2202.01602. https://arxiv.org/abs/2202.01602
[14] Brandi Scardilli, "Sneak Preview: The Retention Game: Managing Subscriber Acquisition and Churn," Streaming Media, July 14, 2026. https://www.streamingmedia.com/Articles/News/Online-Video-News/Sneak-Preview-The-Retention-Game-Managing-Subscriber-Acquisition-and-Churn-175720.aspx
Related Articles
The streaming market has entered a new phase that comes with new challenges. The average US household now holds 3.8 subscriptions and is actively looking to simplify. Platforms that will win the next decade are not those with the largest catalogs, but those that focus on subscribers to make their platform feel indispensable.
03 Apr 2026
Faced with the ongoing challenge of reducing churn and increasing retention, video platforms have realized that customer acquisition alone isn't enough and nor is aggressive pricing. What's really needed is a better understanding of how users engage with content and what kind of experience they're having.
29 May 2025