Mato just raised pre-seed Read more

Subscription Cancel Flows | Pixel Audit

  • Oct 27, 2026
  • 20 min

Show notes

What the episode covers

The subscription cancel button is the screen on trial in this UX Autopsy episode, and our spicy take is that a retention screen is only a negotiation if the customer can walk away in one tap. Maya and Grant disagree on that, and they leave it unresolved. This is the Tuesday episode for week 44 of 2026, a product design and UI design critique built around New York City's click-to-cancel rule, which took effect October 1, 2026.

  • The single "are you sure?" modal: friction or accident guard?
  • The phone-only exit, and why it only passes beside a button
  • Confirmshaming copy with lopsided accept and decline buttons
  • The stacked save flow of offer, survey, second offer and confirm

You will walk away knowing how to audit your own cancel flow with a five-line checklist: same medium, no delay-only steps, a one-tap skippable save offer, non-shaming copy and honest cancellation consequences. You will also see why passing the city test does not guarantee compliance under state auto-renewal laws. Opinions here are design views, not legal advice.

Timeline

In this episode

7 moments worth skipping to. The timecodes match the player above.

  1. 0:12Introduction
  2. 1:55The Rule Is Live, the Federal One Isn't
  3. 5:14The Rule as a Design Spec
  4. 8:23Audit: Modals, Phone-Only Exits, Confirmshaming
  5. 11:50The Turn: When a Retention Offer Becomes an Obstacle
  6. 15:26Red Flag, Green Flag, and the Stealable Audit
  7. 18:37Outro

Quick answers

Straight from the episode

The questions this one settles, without the listen.

Is the NYC subscription cancellation rule in effect, and what are the penalties?
According to the episode, NYC's rule took effect Oct 1, 2026 as the first municipal cancellation rule in the US. Penalties start at $525 per violation, and the city can enforce against any company serving NYC consumers. The federal FTC rule was vacated by the Eighth Circuit in 2025 on procedural grounds, but Section 5 and ROSCA enforcement continue and roughly thirty states have their own auto-renewal laws.
What does the NYC cancellation rule require from a design standpoint?
Cancellation must be as easy as signup and use the same medium, covering every consent medium. Online signup means online cancellation, each channel needs its own cancel path, and in-person enrollment needs an online option. Companies can't refuse to acknowledge, obstruct, or unreasonably delay cancellation, and the cancel must actually complete. The hosts' own heuristics are a tap-and-screen parity test and a check for whether the step can be skipped.
Is a phone-only cancellation flow allowed?
The hosts treated it as a red flag. Grant called phone-only exits red, using a toll-booth phone line example. Maya would accept a phone option only alongside a button. They cited DLA Piper on the rule barring hanging up, false cancellation information, and misrepresenting consequences. The verdicts were framed as opinion.
Are retention or save offers allowed in a cancellation flow?
The episode says the rule permits retention offers only if they aren't another hurdle. A good save screen offers a pause or lower price with an equal-size continue-to-cancel button. A stacked wall of offer, survey, second offer and confirm is the problem. Minnesota and other state laws restrict unsolicited offers, so check sequences against city and state rules. Grant and Maya left the exact line unresolved.
What is confirmshaming, and should it be avoided in cancel flows?
Confirmshaming is lopsided accept/decline copy that shames the user for declining. Both hosts rated it red, and the checklist includes non-shaming copy.
Does passing the NYC rule mean a cancellation flow is compliant everywhere?
No. The episode says the NYC rule tracks state law but omits some state obligations, such as affirmative consent and post-consent notices, so passing the city test isn't clean under state rules. The hosts also stressed this is not legal advice. Their five-line checklist covers same medium, no delay-only steps, a one-tap skippable save offer, non-shaming copy, and honest cancel consequences.

Transcript

The full conversation

Every word of the episode, 3,326 of them, in the order they were said.

Read the transcriptHide the transcript

GrantOrder in the court. The cancel button will now take the stand.

MayaOh, it's gonna plead the fifth. It always hides.

GrantExhibit A: three screens, two modals, and a button that says, "Never mind, keep me." Sworn testimony, please.

MayaObjection, leading the witness.

GrantOverruled. It's been leading me for years.

MayaLet the record show this witness has never once been easy to find.

GrantNoted, and the witness will answer one question only. Can the customer leave or can't they?

MayaOkay. I'm Maya, that's Grant, and this is UX Autopsy. Today, we audit subscription cancel flows against New York City's new click-to-cancel rule.

GrantNow, I'll say it up front. I think a retention screen is a negotiation. You show the discount once, the customer says no, done. I've sat across enough tables to respect a counteroffer.

MayaNo, no, no. Come on. Nobody opens the cancel page hoping for a counteroffer. They've already decided.

GrantSome people haven't decided. Some are just annoyed at the price.

MayaThen let them hit the button, Grant. One click. If the offer's good, they'll come back.

GrantUh, you've never run a book of subscribers.

MayaWait, I'm sorry. Is that your defense? Resume?

GrantIt's a defense, a weak one. We won't settle this today, by the way.

MayaNot a chance. Honestly, I like it that way.

GrantFair. But before we grade anything, we need to know what law we're grading against.

MayaRight, because the rule that's live isn't the one everybody expects.

GrantOctober first, that's the date that matters. New York City's click-to-cancel rule took effect that day, and the law firm Davis & Gilbert says it makes the city the first municipality in the country to write its own cancellation requirements.

MayaFirst city, not state. A city.

GrantA city, which is odd because you'd assume Washington handled this already.

MayaWait, didn't it? I remember the FTC had a click-to-cancel rule.

GrantIt did. The Eighth Circuit vacated it in twenty twenty-five. Jones Day's writeup says that was on procedural grounds, so the court never ruled on whether the idea was bad.

MayaProcedural, so the paperwork was wrong, not the principle.

GrantRight, and the FTC launched an ANPRM in March of this year to rebuild it. That's an advance notice of proposed rulemaking, basically step one of starting over.

MayaSo the federal rule is a pile of lumber right now.

GrantLumber with a lot of lawyers standing around it. But the FTC didn't stop enforcing. Jones Day says it keeps going after the same conduct under Section Five and ROSCA.

MayaUm, ROSCA. Remind me.

GrantThe Online Sales Consumer Protection statute. Older law, still on the books, still usable against a sneaky cancel flow.

MayaOkay, so nobody's off the hook, even without the rule.

GrantNot even close. Jones Day also says roughly thirty states have their own auto-renewal laws, and some are stricter than the vacated federal rule.

MayaThirty. Hmm. So a company can't just say, "The Feds lost. We're fine."

GrantThat's the comfortable story and is wrong.

MayaFine, but here's my worry. I'm a designer at some startup in Ohio. Why do I care about a New York City rule?

GrantBecause of where your customers live. Kilpatrick Townsend says any company serving NYC consumers is subject to enforcement by the city's Consumer Protection Department, wherever it's headquartered.

MayaWherever. So Ohio, Texas, a laptop in a basement in Lisbon?

GrantIf you've got a New Yorker paying you monthly, you're in scope. It doesn't matter if they signed up once on a whim and forgot about it.

MayaNo way. And what's the damage? Because a fine is a number, and numbers I can argue with.

GrantNY1 reports violations can bring restitution plus civil penalties starting at five hundred and twenty-five dollars per violation.

MayaPer violation. Per what, though? Per customer?

GrantThe reporting doesn't say how a violation gets counted, so I won't guess. But starting at five twenty-five is a floor, and I spent fifteen years watching what floors turn into when you multiply them.

MayaSpreadsheet Grant is coming out.

GrantHe never left. Picture a cancel flow with ten thousand New Yorkers stuck in it. The floor starts to look pretty tall.

MayaOkay, that's a vivid picture, and I hate it.

GrantHate it productively. It's the reason to test your flow now, not after a complaint lands on someone's desk.

MayaFine. Test first, panic never.

GrantSo the rule's live, the federal one isn't, and the exposure follows your customers, not your office.

MayaThen I want to read this thing like a spec. Forget the legal summaries. What does the rule actually tell a screen to do? Okay, so I've got the rule open in one tab and Exhibit A in the other. First line that jumps out, "Cancellation has to be at least as easy as sign-up."

GrantAnd in the same medium. DLA Piper's summary of the city's final rule says companies have to offer cancellation in every medium where they accept consent.

MayaEvery one? Wait, really?

GrantEvery one. Sign up in an app, cancel in the app. Sign up on the web, cancel on the web.

MayaHuh. So the old call-us-during-business-hours exit just got a lot more awkward.

GrantDepends how you signed up. And here's where people trip. Say you have several sign-up channels. The city's notice of adoption, as amNY reported it, says you need cancellation through each one.

MayaWhat about the gym? You sign the clipboard at the front desk, then what?

GrantThen you still owe them an online option, a website or an email. In-person sign-up doesn't get to mean in-person-only exit.

MayaHuh. So no more come-back Tuesday and ask for Dale.

GrantDale's days are numbered.

MayaOkay, but I keep hearing, "Just add a cancel button." Is that the whole thing?

GrantNo. Corporate Compliance Insights makes this point directly. The rule goes beyond a button. You can't refuse to acknowledge a cancellation, you can't obstruct it, and you can't unreasonably delay it.

MayaDelay, like your request will be processed in five to seven business days?

GrantUnreasonable delay is the phrase. Where the line sits, I wouldn't bet my spreadsheet on. But a button that does nothing is not a defense.

MayaHmm. So a pretty button can still lose.

GrantEasily. Now, I'm the attacker this round, so let me turn the legal sentence into something a designer can run on Monday.

MayaPlease. Lawyers say, "As easy as." I need a number.

GrantCount taps from your home screen to a confirmed sign-up. Then count taps from home screen to a confirmed cancel. If cancel is bigger, you fail my test.

MayaJust taps? What about reading?

GrantFair. Count screens too. Every modal, every are you sure, every survey is a step. And I'd count the hunt. If sign up is one big button on the home screen and cancel is buried under three menus, that's a gap before you've even started.

MayaWhat about a flow that's two taps, but one of them is a survey you can't skip? Short on paper, miserable in practice.

GrantHmm. Then it fails on the spirit, not the count. So I'd add a third check. Can you skip every step that isn't the cancel itself? A step you can't skip counts double in my book.

MayaOkay, I like that. But hold on, this is your heuristic, right? Not the Citys.

GrantTotally mine. The rule doesn't say, "Count taps," it says, "As easy." I'm just making it testable.

MayaGood, because someone's going to quote you in a legal memo, and then I'm the one answering the email.

GrantPrint it on a mug. Tap parity, screen parity, no scavenger hunt.

MayaAdd a fourth line to the mug. Whatever you built has to actually work on a real phone on a bad connection.

GrantRight. Back to the button that does nothing. Run the whole path to the confirmation, not just to the button, because a cancel that never lands is the same as no cancel.

MayaOkay. Specs on the table, time to walk the exits: the modals, the phone lines and the guilt trip copy.

GrantModals first. I brought a stopwatch. Okay, I'm defending today. Modals, phone lines, guilt copy, bring it.

MayaOkay, phone first, because that's where the rule has teeth. DLA Piper's summary of the city's final rule says it bars hanging up on callers.

GrantWait, that's in there? Like, by name?

MayaBy name. It also bars giving false cancellation information and misrepresenting what happens when you cancel.

GrantOof. So the classic, "You'll lose all your photos forever," line.

MayaIf it's false, yes. That script is a liability now.

GrantFine, but a phone line itself isn't evil. Some people want a human. Red flag or green flag?

MayaPhone-only exit, red. Flat red.

GrantYou didn't even let me finish the question. I've sat on hold for forty minutes to end a membership. I wrote the time down. Of course I did.

MayaLet me defend the human for a second. Some cancellations are messy. A shared plan, a billing mix-up. A person can sort that out in one conversation, and a form can't.

GrantA person can, sure. Offer them. But walk the flow with me. You sign up on your phone in about a minute. You want out, so you open settings, and there's no cancel anywhere, just a number and a line about hours. Now you're calling on your lunch break, listening to a hold song, waiting for a script. That's not service, that's a toll booth. And the fact that hanging up on callers is barred tells me the scene is familiar enough to regulate.

MayaOkay, but my verdict as a fan of options, Green, if it's an option. Phone beside a button is service.

GrantBeside, sure. The moment it's the only door, it's a trap with a hold song.

MayaRight. That one I'll give you. Next, Confirmshaming, the button that says, "No thanks, I hate saving money."

GrantRed. Bright red. You're making the customer insult themselves to leave.

MayaHmm. I'll defend it a little. Brands want a personality.

GrantYour brand's personality is passive-aggressive?

MayaOkay. Okay, red. I can't defend it. Nobody's ever been charmed by that.

GrantAnd look at the pairing. Big, friendly accept button on top, and the decline is a confession in small type.

MayaRight. The accept gets, "Yes, save me," and the decline makes you announce you're cheap. That's not two choices. That's one choice and a dare.

GrantShame has a conversion rate. It also has a refund rate.

MayaOoh, fair. Now the real fight, the single are you sure modal.

GrantRed.

MayaGreen. One tap and it stops the accidental cancellations. Thumbs are clumsy, Grant.

GrantAccidental cancellations are rare. Accidental subscriptions are the whole industry.

MayaBut a cancel is destructive. Banks do it for deleting an account. You confirm, you move on.

GrantA bank confirmation says, "Delete account, confirm." It doesn't say, "Wait, think about the memories." The question is whether the modal is a seatbelt or a speed bump.

MayaAnd if it's just one screen with a clear yes cancel button?

GrantEh, I'd squint at it, but I'd still call it friction. You added a step that sign-up never had.

MayaSign-up has a confirm step too. Everybody's checkout has a review step.

GrantHmm. That's a fair hit. Though the review step serves the company, so I don't love using it as cover.

MayaSo we're stuck.

GrantStuck. And honestly, I'm fine with that. Reasonable people can land on either side of one modal.

MayaThese are just our verdicts, by the way, not legal advice. Nobody go quote us to a judge.

GrantPlease don't, because modals are the small stuff. The save offer is where it gets hairy. When does a discount stop being a feature and turn into a wall?

MayaOoh, yes, that's the one I wanna fight about.

GrantOkay, so the rule doesn't ban the save screen. I went in expecting it to.

MayaWait, really? None of it?

GrantDLA Piper's summary says retention offers are allowed, just not in a way that obstructs cancellation. That's the whole carve out.

MayaObstructs. That's a squishy word.

GrantIt is. The Corporate Compliance Insights piece gives a sharper test. An offer is fine only if it isn't another hurdle between the customer and cancellation.

MayaAnother hurdle. So one screen with a discount is a hurdle?

GrantDepends if it blocks the exit. Offer plus a visible cancel button, probably fine. Offer where the only button says, "Claim it," no.

MayaAh, the button that's a different shade of gray than the other button.

GrantYou've seen it.

MayaEverybody has.

GrantLet me draw both sides. Good side, you tap Cancel. A screen says, "Pause for a month or stay at a lower price," and right under it is a plain Continue to cancel button, same size. One tap and you're out.

MayaAnd the wall side?

GrantYou tap Cancel, get an offer, decline it, then hit a survey you have to finish, then a second offer, then a confirm. Each screen is its own hurdle. Stack three, and that's the kind of sequence I'd want a lawyer looking at.

MayaSo it's the stacking, not the single offer.

GrantNow, New York isn't the only rulebook. A Kilpatrick Townsend alert notes that several state laws restrict unsolicited retention offers in cancel flows.

MayaWhich states?

GrantIt names Minnesota specifically, and it advises anyone with an aggressive save offer sequence to check it against both the city and the state.

MayaHold on. Unsolicited. So an offer the user asked for is different.

GrantThat's how I read the word, yes. I wouldn't go further than that. I'm not a lawyer. I just read spec sheets for a living now.

MayaSpec sheets and car spreadsheets.

GrantSame muscle.

MayaSo a company ships one save flow nationwide, and it might be fine in New York and a problem in Minnesota.

GrantWhich is why the law question stops being the interesting one. The rule doesn't kill the save screen, so now it's a design argument.

MayaAnd I'm going to lose it because I think the screen is a tax.

GrantGo.

MayaSomeone decided to leave. They did the thinking already. They tapped Cancel. Any screen after that says, "We don't believe you."

GrantSure, but some people tap Cancel because it's the only door they can find. They'd stay for the cheaper plan.

MayaThen put the cheaper plan on the pricing page before they're annoyed.

GrantPricing pages don't reach someone who's already at the exit. My test is two things: skippable and one tap and real value on the screen. Pause the plan, drop the price, something that isn't a guilt trip.

MayaReal value according to who? The company writing the screen?

GrantFair. A discount that expires in ten minutes isn't value.

MayaOkay, that one I'd call manipulation.

GrantAgreed. And if the skip is a tiny gray link under a giant green button, it's not skippable in one tap. It's skippable in theory.

MayaSee, that's my point. Every save screen I've met drifts that way. Somebody gets a retention target, and the skip link shrinks by a pixel a quarter.

GrantI have met a few good ones.

MayaName one that didn't make you feel handled.

GrantGive me a minute.

MayaTake two.

GrantI'm not conceding. A good one exists. Hard to find, that's all.

MayaThen the rule is doing us a favor. Hard to find means most of them are walls.

GrantMaybe. We can leave that one open. So what do you actually do with this? Time for the checklist?

MayaFinally. I've got a pen.

GrantOne caveat before anyone prints a certificate: passing the New York City test doesn't make a flow clean everywhere else.

MayaWait, really? I thought the city rule was the strict one.

GrantIt tracks state law, but it leaves out some state obligations. Affirmative consent, for one, and notices you have to send after someone consents.

MayaSo a flow can sail through the cancel side and still be sloppy at signup.

GrantRight. Cancel screens get audited. The consent screen three weeks earlier is where somebody forgets a notice.

MayaUgh, so the audit has a blind spot.

GrantOnly if you stop at the city line. Check your state too.

MayaOkay. Checklist time, and I want it short enough to survive a sprint planning meeting.

GrantFive lines. One, cancel in the same medium you signed up in. Signed up in the app, cancel in the app.

MayaTwo, no step that exists only to delay. If you can't say what a screen does for the user, it's there for you.

GrantThree, the save offer is skippable in one tap. One. Not a tiny gray link, not a second confirm.

MayaFour, copy that doesn't shame. No guilt trips, no sad cartoon mascot.

GrantThe sad mascot has cost somebody a lot of legal fees, I'd guess.

MayaProbably. Five, say honestly what canceling does. Does it access end today or at the end of the billing period? Do they lose their data?

GrantRight, and that one's easy to get wrong by accident. A vague line sounds harmless to the person writing it.

MayaTotally. Whoever writes the copy knows what they meant. The customer is guessing.

GrantSo pick the sentence that matches what your system really does and say it plainly. If access runs to the end of the period, write that. If the data's gone, write that too.

MayaAnd test the copy on someone outside the team. Hand them the screen and ask them what happens next. If they can't tell you, rewrite it. Ten seconds of that beats a week of guessing.

GrantLet's run the whole list on a made-up flow just to see it work. Say you sign up in the app with one tap and a confirm. Now you hunt for cancel.

MayaLine one fails already if it's buried under three menus or if the only exit is an email address and a hope.

GrantSay you find it. You get an offer, a survey, and a second offer. Line two and line three both fail. Each of those is a step the signup never had.

MayaAnd if the decline button says, "No thanks, I like overpaying," that's line four. Red flag, no appeal.

GrantThen the final screen says your account will be closed and doesn't tell you whether you keep access until the period ends. Line five. Zero for five.

MayaOkay, that's a painful audit. But honestly, it takes about five minutes with a phone and a test account.

GrantWhich is the point. You don't need a committee. You need one person willing to be annoyed on purpose.

MayaVolunteer yourself first. You'll be a lot nicer to your customers afterwards.

GrantAnd I'd still keep a save offer in mind. One with real value.

MayaAnd I'd still delete it. We're not settling that today.

GrantNot today. And to be clear, this is design guidance from two people with opinions. It's not legal advice. Your lawyer reads the actual rule.

MayaMm. Show them the checklist first, though. It'll make that conversation shorter.

GrantFair. Five lines, a stopwatch optional.

MayaOkay, I think that's the audit. Shall we wrap it up?

GrantYeah. Let's close the file. So the one thing to steal, run your own cancel flow. Sign up, then cancel in the same medium and see if anything stands between you and the exit.

MayaDo it on your phone too, with your thumb mid-commute.

GrantEspecially then. I'll admit, I still want my one honest save offer.

MayaAnd I still want my one-click exit. We're not settling that today.

GrantNot a chance.

MayaOkay, your homework. Screenshot a screen that bugs you, a sneaky modal, a guilt trip button, whatever, and tag the show. We'll put it on the stand.

GrantSubscribe so you catch next week's audit. And if this changed how you look at your own designs, leave a review. It helps more than you think.

MayaGrant, go cancel your gym membership before you lose your nerve.

GrantUgh. Forty minutes on hold, Maya. I'm still recovering.

MayaThanks for listening, everybody. It was a fun one. I'm Maya. See you next week.

More episodes

Keep listening

Other episodes of UX Autopsy, newest first.

All episodes of UX Autopsy

Sources

Where this came from

7 reports behind the episode. Every one of them opens where it was published.