THESTANDINGORDER

A note before you start

A subscription is a promise to stop asking. The customer stops having to remember, and you stop having to win the next order. When it works, it’s the best business in ecommerce. When it’s built on the customer forgetting to leave, it’s a slow leak with a legal problem attached.

Those two numbers frame the guide. The first says the fight is early. Most subscribers who leave on their own do it in the first few boxes, before any loyalty program, surprise gift or points balance has a chance to matter. The second says the old shortcut is closed. A cancel path four pages deep is no longer a retention strategy. It’s evidence.

What’s left is the fair version, and it’s more profitable than people expect. A subscription that removes a chore, prices the convenience fairly, bends when life changes, and doesn’t let a declined card end a relationship the customer wanted to keep. This guide is about building that, and knowing from your own numbers whether you have.

Build the subscription people would choose again on the day it renews.

It’s written for consumable DTC brands: coffee, supplements, pet food, skincare, razors, anything used up at a steady rate. Most of it applies to memberships and boxes too. It builds on The Whole Machine, which covers contribution margin, payback and cohorts; you don’t need to have read it, but the numbers here assume you know your contribution per order.

How to read it

Start with The Subscription Audit. Your lowest checks name the chapters to read first. Or follow a path:

• Thinking about launching one: A Standing Order, Not a Trap, Should You Sell One?, What a Subscriber Is Worth, The Rules on One Page.
• Running one that leaks: Churn, Counted Properly, A Declined Card Isn’t a Goodbye, The Cancel Flow, The First Two Renewals.
• New to the program this month: A Subscription on One Page, The Subscription Audit, The First Thirty Days.
• Analyst: Churn, Counted Properly, The Scorecard, For Your Analyst.

Five tools run in the page. Nothing you type leaves your browser.

What’s proven and what isn’t

• Filed: numbers from securities filings and company results, here Blue Apron’s, Chewy’s and Peloton’s.
• Published: statutes, court opinions, regulator releases, card-network rules and peer-reviewed research, with the date they were true. One working paper is cited and labeled as such.
• Reported: what a named outlet, research firm or company published. Subscription billing vendors publish most of the benchmarks in this field, and they sell the fixes they measure, so their numbers are marked as vendor data.
• Derived: my arithmetic on filed, published or reported numbers, shown with its inputs.
• From my work: methods I use on client subscription programs. Clients aren’t named and their figures aren’t here.

Examples that open with Say or Picture use made-up round numbers. Every source is listed in Appendix C. The chapter on the rules is an operator’s summary as of September 2026, not legal advice.

Every subscriber passes through the same six moments. Most programs measure the signup and little else.

The moments interact. A deeper first discount lifts moment 1 and quietly worsens 2 and 3. A cancel flow that fights too hard raises the number of people who dispute the charge with their bank instead, which shows up in moment 4 as a problem with your payment processor. Fixing one moment in isolation usually moves the leak rather than stopping it.

Do this

• Write the six numbers in the table for your program, from whatever you have today. Leave blanks blank. Most programs can fill in two.

The audit finds the one or two places your program leaks most, so the next month goes there. It’s also the fastest way to learn whether anyone on the team knows how the program is doing, as opposed to how much revenue it books.

Open your subscription app (Recharge, Skio, Loop, Stay, Ordergroove or whatever runs billing), your email and SMS platform, your payment processor’s dashboard and your store’s checkout. Score each check 0 to 2: 0 if it failed or nobody can answer it, 1 if partly true, 2 if clean. “The app probably does that” is a 0 until someone has looked.

Do two of the checks by hand rather than by report. Sign up for your own subscription on a phone, and then cancel it. Time both. Most teams haven’t done either since launch, and the cancel path is often not what anyone remembers building.

If nobody on the team has canceled your subscription this quarter, nobody knows what your customers go through.

The twelve checks

1. Churn is defined, and split by cause · 5 minLook at: Last month’s churn rate, and how it was calculated.
   Good: Lost subscribers divided by subscribers active on the first of the month, reported as two numbers: canceled by the subscriber, and ended on a failed payment.
   Cost if wrong: You fix the wrong leak. Failed payments and real cancellations need opposite fixes.
   Read next: Churn, Counted Properly
2. A survival table by start month · 4 minLook at: Of subscribers who started in each month, the share still active after renewals one, two, three, six and twelve.
   Good: It exists, it’s updated monthly, and someone reads it.
   Cost if wrong: A big month of new starts hides a falling survival rate for a quarter or more.
   Read next: Churn, Counted Properly
3. You know what a subscriber is worth over a one-time buyer · 4 minLook at: 12-month contribution per subscriber, against one-time buyers of the same first product.
   Good: A dollar difference, after the subscription discount, that someone can produce today.
   Cost if wrong: The discount may cost more than the subscription adds, and nothing on the dashboard would show it.
   Read next: What a Subscriber Is Worth
4. The terms are clear before payment · 3 minLook at: Your subscribe option on a product page and in checkout, on a phone.
   Good: Price per delivery, how often, what happens to the price after the first order, and how to cancel, all visible before the card is entered. The subscription box is never pre-checked.
   Cost if wrong: Surprised customers cancel at the first renewal, dispute the charge, and complain to regulators.
   Read next: The Rules on One Page
5. Canceling online takes a few taps · 5 minLook at: Cancel your own test subscription from the customer account, on a phone. Count the screens.
   Good: Found without searching, finished online in the same place they signed up, with any save offer shown beside a visible way to finish canceling.
   Cost if wrong: Chargebacks, complaints, and in several states, a violation.
   Read next: The Rules on One Page
6. The first-order offer is priced on contribution · 4 minLook at: The subscription offer on your top landing page, and first-renewal rate by offer.
   Good: Someone has checked what each first-order discount does to the first renewal, and there’s no free or near-free first order running without that proof.
   Cost if wrong: You fill the program with people who wanted one cheap box.
   Read next: The Offer That Starts It
7. The first two renewals are managed · 4 minLook at: What a new subscriber receives between the first order and the second charge.
   Good: A reminder before each of the first two renewal charges, with one-tap skip, change date and change frequency. How-to-use content timed to the product.
   Cost if wrong: The product piles up and the second charge becomes the reason to cancel.
   Read next: The First Two Renewals
8. Failed payments are worked, not written off · 5 minLook at: The retry schedule in your billing app, whether the card updater is on, and the failed-payment emails.
   Good: Retries spread over one to two weeks, an account updater switched on, a plain notice with a one-tap card update, and no retries after a decline that says never retry.
   Cost if wrong: You lose subscribers who never decided to leave.
   Read next: A Declined Card Isn’t a Goodbye
9. The cancel flow asks why and answers it · 4 minLook at: The reasons your cancel flow offers, and what each one leads to.
   Good: Each common reason gets the option that fixes it: skip, delay, change frequency, swap, pause, a smaller size. A discount is one option among several, not the default.
   Cost if wrong: You pay everyone to stay, including those who would have stayed for a skip.
   Read next: The Cancel Flow
10. Saves are judged at 60 days · 3 minLook at: How the team reports the cancel flow’s results.
    Good: Saved subscribers still active 60 days later, by the option they took. Not “save rate.”
    Cost if wrong: The flow reports a win while most saves leave at the next renewal.
    Read next: The Cancel Flow
11. Winback has a reason and a holdout · 3 minLook at: What former subscribers receive, and when.
    Good: Sent at a moment that gives them a reason (a new product, a season, a fix to what they complained about), with a random group held back to measure it.
    Cost if wrong: A coupon on a timer teaches people to cancel and wait.
    Read next: Coming Back
12. One weekly scorecard · 3 minLook at: Where the team sees the program’s health each week.
    Good: Active subscribers, starts, churn by cause, failed-payment recoveries, saves and their 60-day survival, in one place, every week.
    Cost if wrong: Problems surface in the monthly revenue number, a month late.
    Read next: The Scorecard

Score as you go; your band appears when all twelve are in.

Read your score

One exception to fixing in check order: if check 4 or check 5 scored 0, fix those first, whatever else scored. Everything else in this guide is about earning more from the program. Those two are about whether you’re allowed to run it.

Every subscription business earns from a mix of two kinds of customer. The first is glad the order keeps coming. The second has stopped using the product and hasn’t gotten around to canceling. Most programs can’t tell you how much of their revenue comes from each, and the difference decides whether the program is an asset or a liability.

What research says about the second kind

In a study published in 2006, the economists Stefano DellaVigna and Ulrike Malmendier followed 7,752 members of three US health clubs over three years. Members on monthly contracts of over $70 went an average of 4.3 times a month, paying more than $17 per visit when a ten-visit pass would have cost them $10 a visit. When they stopped going, an average of 2.31 full months passed before they canceled, with $187 in payments along the way Published.

Liran Einav, Ben Klopack and Neale Mahoney studied the same thing with card-network data on ten subscription services, from entertainment and home security to newspapers and retail goods, in a paper published in 2025. They used a natural experiment: in the month a subscriber’s card is replaced, any subscription the new card doesn’t reach has to be set up again, which forces a decision. In those months, the drop in retention was four times the normal monthly drop. Their models put total revenue at roughly double what the services would earn if every subscriber paid attention, holding the number who signed up fixed Published.

A card replacement is the moment a subscriber is asked whether they still want you. Plan as if every month were that month.

The most direct test comes from a working paper by Adam Miller, Navdeep Sahni and Avner Strulov-Shlain. A large European newspaper offered 1.4 million readers trial subscriptions, at random either set to renew automatically or set to end unless the reader chose to continue. Auto-renewal produced more paid subscribers right after the trial. Overall it did the opposite: auto-renewal cut the number of readers who took a trial by 35%, and over 20 months it cut total subscribers by 23%. The early advantage faded and reversed after about a year Published (working paper, not yet peer-reviewed). The researchers’ reading is that many readers knew they’d forget to cancel, and so didn’t sign up at all.

Why the second kind is borrowed money

Inertia revenue has three lenders, and all three are calling in the loan.

• The customer. People who discover a charge they forgot about don’t just cancel. They tell friends, write reviews, and call their bank. A disputed charge costs you the sale, a fee, and a mark against your merchant account.
• The card networks. Visa requires express consent, a cancel link and a reminder before any trial or introductory price turns into a full charge Published. And about 30% of the card accounts in Visa’s card-updater service get a new number or expiration date, or close, every year Reported. If inertia is what keeps someone subscribed, the new card is where you lose them, as Einav and his co-authors found.
• The regulators. In September 2025 Amazon agreed to pay $2.5 billion to settle the Federal Trade Commission’s case over how Prime was sold and canceled. The complaint described a “four-page, six-click, fifteen-option” cancellation process Published. States have written cancellation into law. Chapter 7 covers the rules.

None of this means the subscription model is in trouble. It means one way of running it is. The other way, where people stay because the subscription saves them a chore, is getting relatively more valuable as the shortcut closes.

The standing-order test

The phrase comes from banking: a standing order is an instruction to pay the same amount on a schedule until told to stop. Nobody resents their standing orders. They set them up because they didn’t want to remember. That’s the bar. A subscription passes when a customer, asked on the day it renews, would say yes again. Three questions tell you whether yours would:

1. Does it remove a chore?Reordering the heavy bag, remembering the refill, running out on a Sunday. If the subscription removes a task the customer would otherwise do, it has a reason to exist that survives attention.
2. Is the price fair for the commitment?A modest discount for letting you plan inventory is a fair trade. A price that starts low and rises quietly isn’t, and it’s the pattern regulators now look for.
3. Can they bend it without leaving?Skip a delivery, delay it, change how often, swap the flavor, pause for a month. If the only way to adjust is to cancel, customers cancel to adjust.

The rest of this guide is about passing those three, and measuring whether you do.

Do this

• Size your inertia base. Pull subscribers still being charged who haven’t clicked an email, visited the site, logged in or changed an order in 90 days. Opens don’t count, because Apple Mail preloads them. Add up their monthly revenue and divide by the program’s total. That’s the share of the program that depends on nobody asking.

Many brands add a subscription because the app makes it easy and the investor deck likes recurring revenue. The better question comes first: would your customers be better off, and would you make more than you would from the same customers buying when they choose?

Three kinds of subscription

A McKinsey survey of 5,093 US online shoppers in late 2017 sorted ecommerce subscriptions into three kinds Reported:

ReportedMcKinsey & Company, “Thinking inside the subscription box,” February 2018. The survey data is from 2017; use it for the shape, not as a benchmark.

The survey also asked why people cancel. The answers that stand out for replenishment are practical: customers are much more likely to cancel when product piles up, or when they can’t adjust what they get. That’s a cadence problem, and a fixable one. It’s the subject of chapter 9.

Curation is harder to sustain. A surprise box sells novelty, and novelty wears off on a schedule of its own. The box has to get better every month to stay as interesting as it was in month one. Access programs, like paid memberships, live or die on whether members use the perks. This guide is mostly about replenishment, where the math is kindest, with notes where the other two differ.

The fit test

1. Is it used up at a steady rate?A dog eats the same amount every day. A candle burns at whatever pace its owner lights it. The steadier the rate, the easier it is to set a cadence that doesn’t pile up.
2. Does running out hurt?Running out of dog food on a Sunday is a crisis. Running out of hand cream is a shrug. The more it hurts, the more the subscription is worth to the customer.
3. Is the reorder a chore?Heavy, bulky, easy to forget, bought on a schedule nobody enjoys tracking. Subscriptions thrive on chores.
4. Can the margin carry a discount?Most programs discount recurring orders. If a 10% discount leaves a subscriber’s 12-month contribution below what the same customer would earn you without subscribing, the program costs you money. Chapter 6 has the tool.
5. Can the customer steer it?Skip, delay, change frequency and swap, in two taps. If your billing setup can’t do this, fix that before launch.

Yes to three of the other four questions, plus a margin that passes question 4, makes a candidate. Fewer than that and you may be better served by a well-timed replenishment reminder, an email or text sent when a one-time buyer is about to run out, which costs no discount and captures much of the convenience.

The cannibalization question

A subscription doesn’t only create orders. It also discounts orders you’d have gotten anyway. Picture a supplement brand whose one-time buyers already reorder about every 35 days, reliably, at full price. Putting them on a 30-day subscription at 15% off buys about 17% more orders and gives up 15% of the price on every one. At a 45% margin, that’s a third of each order’s contribution. For that brand the subscription might lose money on the customers who were already loyal, and make it back only on those it keeps who would otherwise have drifted.

A subscription pays for itself on the customers it keeps who would have drifted, not on the ones who would have come back anyway.

That’s why the only fair comparison is subscribers against one-time buyers of the same first product, over the same period. Subscribers almost always look better on a dashboard, partly because the most committed customers are the ones who choose to subscribe. Some of the gap is the subscription. Some is just who signed up.

Do this

• For your five best-selling products, pull the share of one-time buyers who reorder within 120 days, and the median days between their first and second orders. Products with high repeat and a tight interval are candidates for subscription. Products with high repeat and a loose interval want a reminder flow. Products with low repeat want a better product or a better first order before they want either.

Blue Apron and Chewy both went public on the promise of customers who come back automatically. Their filings tell two very different stories, and the difference comes down to what the subscription asked of the customer.

Blue Apron: refilling a leaking bucket

Blue Apron sold weekly boxes of ingredients and recipes. It went public in June 2017. The expected price range was cut from $15–$17 a share to $10–$11 Reported, and the offering priced at $10 Filed. The quarter before, it reported numbers that most DTC brands would envy:

FiledBlue Apron Holdings, first quarter 2018 report and results, with 2017 comparisons; IPO prospectus, June 2017.

The number that mattered was the last one. A quarter of revenue went to marketing, much of it to replace customers who didn’t stay. Independent card-panel data from Second Measure in 2016 found that 28% of Blue Apron’s customers were still subscribed six months after their first purchase Reported. The pattern outlasted the IPO. Of Blue Apron’s customers who started in January 2022, 15% were still buying 11 months later, and that was the best of the meal kits Second Measure tracked: Home Chef kept 11%, HelloFresh 9%, Marley Spoon and Sunbasket 5% Reported. Second Measure changed its dataset in 2022, so the two years aren’t directly comparable.

Customer counts tell the rest. Blue Apron reported 786,000 customers in the first quarter of 2018, 336,000 in the fourth quarter of 2021 and 298,000 in the fourth quarter of 2022 Filed, about 71% below the quarter before the IPO Derived. In November 2023 the company was acquired by Wonder for $13 a share, about $103 million Filed. That price came after a reverse stock split Reported, so it can’t be compared directly with the $10 offering price.

Plenty of customers liked the food. The problem was the ask. A meal kit subscription asks the customer to plan meals around a box, find an evening to cook, and decide every week whether to skip. It adds a chore. The novelty that sold the first box wore off, and each week gave the customer another chance to notice.

Chewy: taking a chore away

Chewy sells pet food and supplies, and its subscription, Autoship, is the opposite kind of ask. Here’s what its most recent annual results report Filed:

FiledChewy, Inc., fourth quarter and fiscal 2025 results, March 2026. “Autoship customer sales” includes everything Autoship customers buy, not only their Autoship orders.

Look at the terms, as Chewy’s site states them in September 2026: 35% off the first Autoship order, up to $20, then an extra 5% on select brands. Change, skip or reschedule anytime, and change the date up to 48 hours before shipping Reported. The ongoing discount is small. The control is nearly total. And the product is a 30-pound bag that a dog empties on a schedule, which nobody wants to carry home or remember to reorder.

Blue Apron’s subscription asked customers to keep doing something. Chewy’s asked them to stop.

How to read the pair

It isn’t a clean comparison. Chewy is a retailer with a huge catalog and pharmacy and vet services, and its Autoship figure counts everything Autoship customers buy. Blue Apron had problems beyond the design of its subscription. But the lesson survives the caveats. When a subscription removes a task, customers stay because the subscription saves them work. When it adds one, every delivery is a fresh decision, and the program spends its life paying to replace the people it loses.

Do this

• Write two sentences. The chore your subscription takes away, and the chore it adds: deciding whether to skip, storing extra product, remembering to pause before a trip. If the second sentence is longer, fix that before spending another dollar acquiring subscribers.

Churn is the share of your subscribers you lose in a period. That sounds too simple to get wrong, and yet I rarely audit a program where the number on the dashboard means what the team thinks it means.

The definition

Monthly churn is the number of subscribers you lost during the month, divided by the number who were active on the first day of the month. Count only losses from that starting group. Someone who subscribed on the 10th and canceled on the 25th belongs in the survival table for this month’s starts, not in this month’s churn rate.

Then split it. Subscribers leave in two ways, and the fixes have nothing in common:

• Voluntary: the subscriber chose to cancel.
• Involuntary: a payment failed and every retry after it failed too, so the subscription ended without anyone deciding anything.

Peloton’s filings are a good model of a metric defined with care. It reports average net monthly churn for its connected fitness subscriptions: the quarter’s churn events, minus subscriptions that came back from a pause, minus reactivations, divided by the average number of subscribers at the start of each month, divided by three. A pause counts as churn on the day it starts Filed. Its recent quarters show why a single quarter misleads: 1.9% for the quarter ending December 2025, 1.2% for March 2026 and 2.2% for June 2026, with the summer quarter highest Filed. Whatever definition you choose, write it down, count a pause as a loss until the subscriber comes back, and compare the same month year over year before you panic about a seasonal one.

Three impostors

1. Cancels against starts“We started 400 and lost 250 this month.” That ratio tells you whether the program grew. It says nothing about how fast you lose people, because the two numbers come from different groups. A program can have a healthy ratio and terrible churn, if it’s buying lots of new starts to replace the people leaving.
2. Churn on the wrong baseDividing losses by the end-of-month count, or by a base that includes this month’s new starts, makes churn look lower in a growing program. The faster you acquire, the better it looks, which is backwards.
3. Half the lossesSome reports count only cancellations; others count only subscriptions that stopped billing. Either way you see one leak and miss the other.

From my workThe first of these is the most common and the most expensive. I’ve seen a same-month ratio of cancels to starts presented as a churn rate, and a save campaign sized against it. It was the wrong instrument measuring the wrong thing. Split the losses by cause and by the month each subscriber started, and the right fix is usually obvious.

How much is failed payments?

More than most teams guess. In Recurly’s benchmarks, updated with July 2026 data from its network of subscription businesses, failed payments made up about a third of ecommerce subscription churn: 1.38 points of 4.25 Reported. Recurly doesn’t make clear whether those are monthly or annual rates; the share is what matters. The share was about 30% for subscriptions averaging $10 to $25 per customer, and about 6% for those over $250 Derived. The cheaper the subscription, the bigger the share of its losses that come from declined cards. Stripe has said that 25% of lapsed subscriptions are “purely due to payment failures” Reported. Both are billing vendors reporting on their own networks, and neither publishes its sample, so treat the range as a guide and measure your own.

You’ll also see “up to 48%” and “53%” quoted for involuntary churn. I couldn’t trace the first to any data at all, and the second contradicts the numbers in the same article that states it. Don’t use either.

Churn isn’t constant

The average hides the most important fact about subscription churn: it’s heavily front-loaded. Recharge, which runs subscriptions for more than 20,000 Shopify brands, published its active churn rate by renewal for renewals from July 2025 to March 2026, counting only subscribers who canceled Reported:

ReportedRecharge, “Subscription churn is front-loaded,” 2026. Vendor data; subscriber-initiated cancellations only, so failed payments add to these.

Two consequences. First, an average monthly churn rate is a blend of new subscribers leaving fast and old subscribers leaving slowly, so it moves with your mix. A program that just had a big month of new starts will show rising churn next month even if nothing got worse. From my workI pin a note to the scorecard of any program coming off a big launch month: aggregate churn will spike as that group reaches its first renewals, so judge the save and payment fixes on their own numbers, not on the total. Second, the place to fight is the first two renewals, which is chapter 9.

An average churn rate is a blend of new subscribers leaving fast and old ones leaving slowly. Report it by start month.

The survival table

The fix for the blend is a table with one row for each start month and columns for the share still active after renewals one, two, three, six and twelve. Read across a row to see how one group decays. Read down a column to see whether newer groups survive better or worse than older ones. Leave a cell blank when the group isn’t old enough to fill it. The query is in Appendix A.

Do this

• Recompute the last three months with this definition, split into voluntary and failed-payment losses. Then build the first version of the survival table, even if it’s by hand. Put both next to the churn figure your billing app shows and write down why they differ.

The case for a subscription is usually made with one comparison: subscribers are worth three times what one-time buyers are. Recharge reports something close, with subscribers placing nearly three times as many orders as one-time shoppers across the brands it serves Reported. The comparison is true and it’s the wrong one.

Why the obvious comparison flatters

The customers who choose to subscribe were already your most committed. They’d have bought more than average without the subscription. How a customer arrives also changes how they behave. Studying a digital TV service, Hannes Datta, Bram Foubert and Harald Van Heerde found that customers who came in on a free trial behaved differently enough, even after correcting for who chose the trial, that their lifetime value was 59% lower than that of customers who paid from the start Published. Who a program attracts, and how, shapes what it’s worth as much as what the program does.

So compare like with like: subscribers against one-time buyers of the same first product, from the same months. Better still, randomize something. Show the subscribe option, or two discount levels, to a random split of product-page visitors, and compare 12-month contribution per visitor. Selection can’t bias that comparison.

Three pieces

1. Expected ordersFrom your survival table. A subscriber on a 30-day cadence has 12 scheduled orders in a year. Each counts only if they’re still subscribed when it ships, so expected orders are the sum of the survival rates at each one.
2. Contribution per order, after the discountThe discount comes out of contribution, not revenue. Say an order is $40 and leaves $18 after product, packing, shipping and fees. A 10% subscription discount is $4, which is 22% of the contribution. The discount always looks smaller than it is.
3. The comparisonWhat a similar one-time buyer is worth over the same 12 months, at full price. The subscription’s value is the difference, not the total.

A 10% discount on price is often a 20% discount on what the order leaves you.

Where the value sits

Run the example and look at the last number. Five fewer points lost at the first renewal adds $4.12 of contribution per subscriber, with no extra discount, on every new subscriber. Every later order depends on the subscriber getting past that first renewal, so survival at the start compounds through the whole year. It’s why the most valuable work in a subscription program is usually unglamorous: the right cadence, a reminder before the charge, and an easy skip.

The same math says when a subscription isn’t worth having. If your one-time buyers already reorder almost as often as subscribers would, and your contribution is thin, the discount can cost more than the extra orders earn. The tool will show a negative gap. Believe it, and consider the replenishment reminder from chapter 3 instead.

And if you acquire subscribers with a different first-order offer from other customers, include that in the comparison too, through payback. The Whole Machine has the payback tool; feed it the subscriber’s contribution by month.

Do this

• Run the tool twice: once with your subscribers’ real survival, and once with the survival you’d have if first-renewal losses fell by ten points. The difference, times the number of new subscribers you add in a year, is the budget for the work in part three.

You don’t need to be a lawyer to run a compliant subscription. You need to know the handful of rules that shape the signup and the cancel button, and to have counsel check your version. This is an operator’s summary as of September 2026, not legal advice. The rules differ by state and change often.

Federal: the rule that didn’t happen, and the law that did

The Federal Trade Commission adopted its “click to cancel” rule on October 16, 2024, by a 3–2 vote. It would have required canceling to be as easy as signing up. On July 8, 2025, the US Court of Appeals for the Eighth Circuit vacated the whole rule, not on its content but on procedure: the FTC had skipped a required preliminary analysis of its economic impact. In February 2026 the FTC formally restored its older rule, and in March 2026 it asked for public comment on starting again. As of September 2026, no new rule has been proposed Published.

None of that made hard-to-cancel subscriptions legal. The Restore Online Shoppers’ Confidence Act, passed in 2010, still applies to anything sold online on a recurring basis. It requires three things Published:

1. Clear terms before billing informationThe seller must clearly and conspicuously disclose all material terms before collecting the customer’s billing details.
2. Express, informed consentThe seller must get the customer’s express informed consent before charging.
3. A simple way to stopThe seller must provide simple mechanisms for the customer to stop recurring charges.

The FTC and the Justice Department have kept enforcing it Published:

The states

• California amended its automatic renewal law in 2024; the changes apply to subscriptions entered into, amended or extended on or after July 1, 2025. Customers who signed up online must be able to cancel online, at will, without steps that obstruct or delay them. You may show a save offer online only if a “click to cancel” button, or words to that effect, is displayed continuously next to it. Price changes need notice 7 to 30 days ahead. Keep proof of each customer’s consent for at least three years, or one year after the subscription ends if that’s later Published. The law also requires an annual reminder. The text covers annual agreements and continuous service agreements, and the Attorney General’s summary reads more broadly still; confirm the current wording with counsel. The safe practice is a yearly reminder to every California subscriber.
• Minnesota, from January 2025 Reported: once a customer asks to cancel, you can’t present a save offer unless they first agree to hear one, and you may ask that only once per attempt. You may describe pausing or downgrading. Continuous subscriptions need a written notice at least once a year Published.
• New York amended its law in 2025, effective around November 2025 Reported. Canceling must be as easy as the way the customer agreed and available through the same medium. You may present a discount or retention offer, but not with unreasonable conditions. A price increase needs the customer’s consent, or a cancellation window of at least 14 days after the charge with a prorated refund Published.

Other states have their own versions. Build to the strictest one you sell into, which for most brands means California’s cancel button and Minnesota’s ask-first rule.

The card networks

Visa has required since April 2020 that merchants selling subscriptions with a free trial or introductory price get the cardholder’s express consent at signup, send a confirmation that includes a simple way to cancel, and send a reminder at least seven days before the trial or introductory price ends and the recurring charge begins Published. Mastercard introduced a similar rule for free trials of physical products in 2019 Reported. These rules bind you through your payment processor, whatever any state says.

Design the signup and the cancel path as if a regulator will screen-record them. One might.

The operator’s checklist

1. An unchecked choiceSubscribing is a separate, active choice: a toggle or button the customer selects, never a box that comes pre-checked.
2. Terms beside the buttonPrice per delivery, how often, what the price becomes after any introductory offer, and how to cancel, next to the subscribe button, before the card is entered.
3. A confirmation with the terms and a cancel linkSent straight after signup.
4. A reminder before the price changesAt least seven days before an introductory price ends, and before any price increase, within the notice window your strictest state requires.
5. Cancel online, where they signed upIn the customer account, found without searching, finished in a few taps.
6. Save offers that ask first and show the exitAsk once per attempt whether they’d like to hear options; show a clear cancel button beside any offer; if they decline, let them cancel.
7. A yearly reminderTo every subscriber: what they’re subscribed to, how often and how much they pay, and how to cancel.
8. RecordsKeep proof of consent (what the customer saw and chose, and when) for at least three years, or one year after the subscription ends if that’s later.

Do this

• Screen-record your own signup and cancellation on a phone. Send both recordings to counsel with the checklist above, and ask which states you sell into have rules beyond it.

Every subscription starts with an offer, and the offer is a filter. It lets in the people it appeals to. The deeper the first-order discount, the more of those people wanted the discount rather than the product.

What the evidence says

In a study published in 2006, Michael Lewis used customer records from a newspaper and an online grocer and found that customers acquired with a 35% discount were worth about half as much over time as customers who paid full price Published. That’s the long-run risk.

Recharge’s data on subscriptions in particular suggests the risk is small at modest discounts and steep at the extremes. Across about 29.8 million new subscriptions started from July to December 2025, here’s the share that made it through the first renewal, by the size of the first-order discount Reported:

ReportedRecharge, 2026. Vendor data. Recharge concludes that subscribers acquired at up to about 25% off are worth about as much over 12 months as a store’s typical subscriber. This is a different study from the churn table in chapter 5, with a different definition, so the levels don’t match.

Free is its own category. In Recharge’s data on 132.9 million subscriptions from 2023 and 2024, 25.6% of subscriptions with a $0 first order reached a second order, against 63.6% of those with a first order priced at 99 cents. By the sixth order it was 8.7% against 32.8% Reported.

A price of 99 cents makes someone decide. Free lets them decide later, at the first real charge.

The ongoing discount

The first-order discount gets the attention; the ongoing discount costs more, because it applies to every order. Amazon’s Subscribe & Save is a useful benchmark. Sellers choose to fund 0%, 5% or 10% off, and Amazon adds 5% more when a customer receives five or more subscription items at one address in a month, so the usual ceiling is 15% Reported. Chewy’s ongoing Autoship discount is 5% on select brands. If Amazon and Chewy keep the recurring discount that modest, a DTC brand with a thinner margin should need a reason to go deeper. Run the numbers in chapter 6 before you do.

Where you can, make the ongoing benefit something that isn’t a price cut. Free shipping on subscription orders, first access to new flavors, a member-only product. They often cost less than they’re worth to the customer, and they don’t train anyone to think your real price is 15% lower.

Samples and trials

A free sample or trial that turns into a subscription is the offer most likely to fill a program with people who didn’t mean to join, and the one the card networks and regulators watch most closely. If you run one, the Visa rules in chapter 7 apply: express consent, a confirmation with a cancel link, and a reminder at least seven days before the first full charge.

From my workWhen I audit a sample-into-subscription offer, the first number I ask for is the share of sample takers who reach a paid second order, split by traffic source. Sample volume is a vanity metric. Some sources send people who love free things, and the only place that shows up is the second order.

Choosing the cadence at signup

The first-order offer also sets the cadence, and a wrong cadence is the most common reason product piles up. Recommend one based on how people actually use the product (“most people finish a bottle in about five weeks”) and let the customer change it on the spot. A cadence the customer picked is one they own.

Do this

• Pull first-renewal rate by first-order offer for the last six months, next to the no-discount baseline. Any offer whose subscribers renew more than ten points below the baseline has to justify itself on contribution, with the tool in chapter 6, or go.

The first two renewals decide the program. Everything later is a smaller number, as chapter 5 showed. Here’s what those rates do to a group of 100 subscribers who start in the same month.

DerivedFrom Recharge’s reported cancellation rates at each renewal (24.1%, 27.1%, 16.6%, 12.4%, 10.1%), chained together. The rates come from renewals across many cohorts, so the curve is an approximation. Failed payments are not included, so real survival is lower.

Almost half the group is gone by the second renewal, before the program has had a chance to become a habit. The good news is that the reasons people leave this early are mostly practical.

Why they leave early

• Too much product. The second box arrives while the first is half full. McKinsey’s survey found people are much more likely to cancel when products pile up Reported, and Recharge says one of the most common reasons subscribers give is that they “have too much” Reported. Nobody publishes a reliable share for it; you’ll see “40%” quoted, but in its source it’s a hypothetical.
• A charge they didn’t expect. They forgot they’d subscribed, or missed that the introductory price ends.
• They haven’t used it yet, or used it wrong and didn’t see a result.
• The price went up after the first order, as designed, and the full price feels different when it’s a charge rather than a choice.

Four fixes, in order

1. A reminder before each of the first two renewal chargesThree to five days ahead, and at least seven days ahead when an introductory price is ending. Say what’s coming, when, and for how much, with one-tap buttons to skip, delay two weeks, change how often or swap, and a plain link to cancel. Written as a service, not a warning. It turns the surprise charge into a decision the customer makes on your page instead of their bank’s.
2. Easy adjustments, everywhereThe same four actions in the account, in every shipping email, and by text if they’ve opted in. In Recharge’s 2023 report, covering 2022, 35% of subscribers adjusted an order (skipping, changing frequency or swapping products), and 39% of those skipped Reported. A skip is a subscriber who stayed.
3. Use before the second boxTime the how-to content so the customer has used the product properly before the second charge: how much, how often, what to expect in week one and week three. Ask how it’s going, and read the replies.
4. A longer first cycle, where it fitsOffer a two- or three-month supply or a quarterly plan. In Recharge’s data on 65.1 million subscriptions, cumulative churn was 84.2% for monthly plans, 67.7% for quarterly and 30.1% for annual Reported. Recharge itself cautions that committed customers choose longer plans, so much of that gap is selection. Many annual subscribers in that window hadn’t yet faced a renewal, which widens it further, and these figures include failed payments. But a longer first cycle also means fewer early chances to leave, and fewer boxes piling up.

A skip is a subscriber who stayed. Make it easier than canceling.

Two more things belong in the first sixty days. Keep new subscribers out of campaign discounts deeper than the one they subscribed at; seeing a better price a week after signing up is a reason to cancel and buy again. And make sure the card updater from chapter 10 is on before the first renewal, not after.

Do this

• Build the pre-renewal reminder for the first two renewals this week, with skip, delay, frequency, swap and cancel buttons. Hold back a random 10% of new subscribers from the extra reminder, but still send them every notice the card networks or state law require, such as the seven-day notice before an introductory price ends. Compare churn at the first two renewals once both groups have reached the second renewal.

When a renewal charge fails and every retry fails after it, the subscription ends. The customer didn’t cancel. Often they don’t know it happened until the product stops arriving. It’s the only churn you can fix without changing anyone’s mind, which makes it the cheapest churn there is.

Why cards fail

Most renewal declines are soft: the bank said no for now, not no forever. Recurly, a billing vendor, reports that 72% of renewal card declines are generic declines, insufficient funds or temporary holds Reported, all of which can succeed on a later try. Some of the rest are hard declines: the card was reported lost or stolen, or the account is closed, and trying again won’t help.

Cards also change constantly. Visa says about 30% of the card accounts in its card-updater service get a new number or expiration date, or close, every year Reported. And the month a card is replaced is dangerous for another reason: Einav, Klopack and Mahoney found retention drops four times as much as usual in those months, because a customer whose new card didn’t carry over has to decide whether to set the subscription up again Published.

Four fixes

1. Switch on the card updaterVisa Account Updater and Mastercard’s Automatic Billing Updater send card-on-file merchants new card numbers and expiration dates when banks reissue cards Reported. Most processors and billing apps offer an updater, but check that it’s on. Network tokens, which replace the raw card number with a token the network keeps current, go further. Visa reports they lifted approval rates on online payments by 3% to 4.6% and cut fraud by 28% to 30%, compared with raw card numbers Reported. Those are Visa’s own figures; the processor Adyen reports a similar gain of about 3%.
2. Retry within the rulesVisa’s rules say never to retry after a decline that means the issuer will never approve, such as a lost, stolen or closed card. For other declines, fees start after about 15 reattempts on the same transaction within 30 days Published; some acquirers set the limit at 20 Reported. Mastercard sends merchant advice codes with its declines: code 03 means don’t try again, 21 means the recurring plan was canceled, and codes 24 to 30 say how long to wait after insufficient funds Reported. Stripe’s recommended default is eight tries within two weeks Reported. Check that your billing app follows the codes, spreads retries over one to two weeks, and stops when told.
3. Say what happened, plainlySend a notice with the facts: the payment didn’t go through, when you’ll try again, and a one-tap link to update the card. The subscription is still there. No fake deadlines, no alarm. The last notice states the real date the subscription will end.
4. Offer a second way to payLet customers add a backup card or a digital wallet, and ask for it at the moment it’s useful, on the card-update page.

The card bounced, not the customer. Write every failed-payment message as if you believe that.

The message

From my workA failed-payment email is a billing notice first and a marketing message a distant second. The version I use opens with one line of fact (“Your card didn’t go through, so your next delivery is on hold. Your subscription is still here.”), gives the real date of the next attempt, and puts the update-card button above anything else. Every word has to be true. A fake “last chance tonight” when retries run for another week is the kind of pressure regulators describe in negative-option cases, and a spike in disputed charges can threaten your merchant account, which is a bigger problem than the subscription.

Three more rules from the same programs. Send the plain notice to every affected subscriber, and anything more promotional only to those who opted in to marketing. Keep subscribers with a failing payment out of promotional sends until it’s fixed. And build the failed-payment flow on your billing app’s own events, then test it by forcing a real failed charge on a test subscription. When a brand moves between email or billing platforms, the content moves and the triggers often don’t; a failed-payment flow that stopped firing looks exactly like one that’s working, until you check how many times it fired this week.

What recovery looks like

Billing vendors report recovering roughly half of failed renewal payments through retries and messages. Stripe says its users recovered 55% of failed payments on average in 2025, and that a recovered monthly subscription typically lasts another seven months Reported. Every vendor counts “recovered” differently and none publishes an audited sample, so use these as a direction and measure your own rate before and after.

Do this

• Pull the last 90 days of failed renewals by decline reason. Check three settings in your billing app: the retry schedule, whether retries stop on hard declines, and whether the card updater is on. Then force a failed charge on a test subscription and watch every message arrive.

A customer who clicks cancel is telling you something. Most cancel flows respond in one of two ways: with a maze, which state laws and federal enforcement now treat as unlawful, or with a discount for everyone, which pays people to stay who would have stayed for a skip. The better flow listens to the reason and answers it.

The shape

1. A visible cancel buttonIn the account, where they’d look. The rules in chapter 7 apply from the first screen.
2. One question: why?A short list of reasons plus “something else.” Answering is optional; Minnesota explicitly bars making it a condition of canceling.
3. The options that fit the reasonOffered with a plain “No thanks, cancel my subscription” beside them. Ask first whether they’d like to hear an option; Minnesota requires it, and building to it everywhere is simplest.
4. A clean confirmationIf they cancel, confirm it at once, in writing, with the date of the last charge. No survey wall, no second offer.

Reasons and answers

A discount belongs at most in one row, “too expensive,” and after the size and interval options. A discount offered to every canceller is a price for canceling. Customers learn it, and some come back to click cancel whenever they want the lower price.

Never reward a behavior the customer controls and you don’t want. Reward the behavior you do want, after it happens.

From my workThat rule is the one I hold hardest on every program: no bonus, discount or gift on a trigger the customer controls and you don’t want, like abandoning a cart or starting to cancel. Surprises go after behavior you do want, like a completed third order. A reward that reliably appears when someone clicks cancel isn’t retention. It’s a price list, and deal forums often publish it within days.

Pause is the strategic option

A pause turns a cancellation into a subscriber who’s resting. Give it an end date the customer chooses, remind them a few days before it resumes, and let them extend or cancel from that reminder. Count it as a loss: Peloton counts a pause as churn from the day it starts and subtracts it when the subscriber comes back Filed, and that’s the right way for you to count it too. Otherwise a flow that pauses everyone looks like a perfect save rate.

How many pausers come back is a vendor number with a wide range. Recurly reported in 2020 that about a third of paused subscriptions were reactivated, and in 2026 that 75% of customers who pause “eventually return,” without defining either Reported. Measure your own at 30, 60 and 90 days.

Don’t push offers at people who haven’t asked

Save offers belong to people who have asked to cancel. Pushing them at everyone can backfire. In a field experiment at a wireless carrier, Eva Ascarza, Raghuram Iyengar and Martin Schleicher sent customers recommendations for cheaper plans that would have saved them money. Churn in the following three months rose from 6% in the control group to 10% among those who got the recommendation Published. Prompting people to think about their plan prompted some of them to leave it. In a later study Ascarza found that the customers most likely to churn weren’t the ones a retention offer helped most; targeting those who respond to the offer beat targeting those at highest risk Published. This is about unprompted offers, not notices: the yearly reminder from chapter 7 still goes to everyone.

Judge saves at 60 days

Vendors report average save rates from about 10% to about 17%. Recharge reports a 9.6% average among merchants using its pause feature, and Chargebee a 17.4% average across its cancel flows Reported. Claims of 30% and up for tuned flows exist, without published samples. But the save rate is the wrong number to manage. A saved subscriber who leaves at the next renewal was worth one order.

From my workThe governing number on every cancel flow I build is saved subscribers still active 60 days later, by the option they took. Until you’ve measured it, assume half of saves are gone by day 60 and plan the flow’s economics on that. If survival comes in lower, the flow is worth less than you planned, and the report shows which option to cut.

Do this

• Pull the last 90 days of cancellation reasons. For the top three, make sure the flow offers the option in the table, with a visible cancel button beside it. Then set up the 60-day report: saves by option, and how many are still active.

The first two renewals are about removing reasons to leave. After that, the question changes: what does a subscriber in month twelve have that one in month one doesn’t? In most programs, the real answer is nothing but a longer list of past charges.

Status that can’t be bought

Research on loyalty programs has long found that people work harder as they get closer to a goal, and value progress they feel they’ve already made (Kivetz, Urminsky and Zheng, 2006; Nunes and Drèze, 2006) Published. The Whole Machine covers that research in its chapter on loyalty. For subscriptions, it points to one design above others: tenure that earns something.

From my workThe pattern I use is a consecutive-month count that unlocks real perks money can’t buy: first access to new products, a member-only item, a say in the next flavor. It shows on the account page and in the shipping emails, and it resets if the subscriber cancels, not if they pause. Then the cancel flow can say something true without a discount: “You’re at month eight. Canceling resets it.” Status that anyone could buy back the next day holds no one. Status that took months to earn does.

What paid membership does to behavior

A study published in 2022 by Raghuram Iyengar, Young-Hoon Park and Qi Yu followed customers who joined an Asian beauty retailer’s $50-a-year online membership and found their purchases rose substantially and stayed higher. The authors attributed only about a third of the rise to the program’s economic benefits (gift cards, free shipping and member discounts) and about two-thirds to other effects, with evidence of a sunk-cost effect: wanting to get value from the fee already paid Published. Commitment changes behavior. The corollary is that the commitment has to feel worth it, or it becomes the reason to leave.

Five things that build with time

1. Tenure perksAs above: earned, visible, and lost on canceling.
2. Surprises after the behavior you wantA gift in the third box, or the sixth. Unannounced, so it rewards staying without becoming a price for it.
3. An easy add-onLet subscribers add a one-time item to the next delivery in one tap. Chewy’s 83% figure for Autoship customers counts what they buy alongside their Autoship orders Filed, which is the point: a subscriber who can add to the box buys more.
4. A human voiceA short note from the founder or the team at month six. Replies go to a person who answers.
5. Something newA seasonal flavor, a limited scent, first to try a new product. Replenishment doesn’t have to be boring.

Every number you show a subscriber is a promise. Show only the ones that are true today.

A word on showing value back. “You’ve saved $84 this year” is powerful when it’s accurate and damaging when it isn’t: counting a discount on a price the customer never paid, or including orders they skipped. From my workI treat every number shown to a customer as a representation, calculated from the order record, with the date it was true. If the data behind a number can be stale or wrong, the number doesn’t ship.

Do this

• List what a subscriber at month twelve has that one at month one doesn’t. If the list is empty, design one perk that can only be earned by staying, and add it to the account page and the shipping emails.

Recurly, a billing vendor, reported in 2026 that nearly one in four new subscriptions across its network now comes from a former customer Reported. They already know the product, and you know why they left. Most programs waste both advantages with a coupon on a timer.

After the cancel, the odds drop

From my workOn the programs I’ve measured, a winback email sent after cancellation converts at a small fraction of what the cancel flow saves in the moment. That’s why the save belongs inside cancellation. Winback is for the people who left anyway, and it needs something the cancel flow didn’t have: a reason that didn’t exist when they left.

A reason, not a reminder

• You fixed what they complained about. They said it was too much, and now there’s a 60-day option. They said the flavor was bland, and there’s a new one. Use the cancellation reason you collected in chapter 11.
• Something new launched that fits what they bought.
• A season when the product matters more: sunscreen in May, immunity blends in October.
• A pause ended, or a trip they told you about is over.

Send it once per reason, not on a schedule. A standing “we miss you, here’s 30% off” at 30, 60 and 90 days teaches customers that canceling is how to get the lower price.

Winback needs a reason that didn’t exist when they left.

Prove it

Former subscribers resubscribe on their own, some of them in any given month. A winback campaign takes credit for all of them unless you hold a random group back. From my workI hold back a random slice of every winback audience, large enough to read, and compare resubscription over the next 30 days. The difference between the groups is what the campaign did. The rest would have happened anyway. The Honest Test covers how large a group needs to be.

Do this

• Group your former subscribers by cancellation reason. For the two largest groups, write down what has changed since they left. If nothing has, fix that first. When something has, send one message to each group, with a random 10% to 20% held back.

A subscription program usually shows up in the weekly meeting as one line: subscription revenue. By the time that line moves, the cause is a month old. The scorecard below catches it in the week it happens.

A number without its denominator is a mood, not a measurement.

Two rules for the page

From my workFirst, no subscription revenue figure appears without total store revenue beside it. Subscription revenue can rise because the program is working, or because it’s absorbing orders that one-time buyers used to place at full price. Only the store total tells you which. Second, the flow-trigger count is the most boring line on the page and the one most likely to catch a problem nothing else will. A failed-payment flow that fires zero times in a week isn’t a good week. It’s a broken trigger, and nothing else on the page will tell you.

Reading it

Read churn by cause and by start month, never as one blended number; chapter 5 explains why. Expect the aggregate to rise for a month or two after any big acquisition push, as those subscribers reach their first renewals. When you change something, hold back a random group so the result doesn’t depend on the season or the mix. And re-read the free-text cancellation reasons every week. They’re the only thing here written by customers.

Do this

• Build the page this week with whatever you can pull, even if half the cells are blank. Put the blanks on a list with a name next to each. The page gets complete by being used.

Whether you’ve just launched a subscription or just inherited one, the order of work is the same. Fix what could get you in trouble. Switch on what recovers money without changing anyone’s mind. Measure properly. Then build the flows that change minds.

1. Week one: the rules and the settingsScreen-record signup and cancellation on a phone and send them to counsel with the checklist from chapter 7. Fix anything pre-checked, hidden or missing. Switch on the card updater, check the retry schedule and hard-decline handling, and force a failed charge on a test subscription (chapter 10).
2. Week two: the numbersRecompute three months of churn, split by cause (chapter 5). Build the first survival table. Run the worth tool with your real survival (chapter 6). Export 90 days of cancellation reasons and failed renewals by decline reason.
3. Week three: the first two renewals and the cancel flowLaunch the pre-renewal reminder for renewals one and two, with skip, delay, frequency, swap and cancel, and a random 10% held back from the extra reminder, never from required notices (chapter 9). Map the top three cancellation reasons to the options in chapter 11, with the cancel button visible beside each.
4. Week four: the scorecardPut the page from chapter 14 in front of the team. Start the 60-day save report. Group former subscribers by cancellation reason and write down what’s changed for each (chapter 13). Decide the ongoing discount from the worth tool, not from habit.

At day thirty you won’t have results yet; the saves need sixty days and the survival table needs a few more start months. What you’ll have is a program that’s legal, instrumented and fixing its cheapest leaks.

Settings before copy. Measurement before opinions.

Do this

• Put the four weeks on the calendar today, with an owner for each line. If the same person owns billing settings, the cancel flow and the scorecard, give them this guide.

Whoever owns the program, a new hire, an agency, or you on the Monday you decide to take it seriously, needs six things on day one. Without them, the first month goes on hunting for things they should have been handed.

1. Admin access to billingThe subscription app, the payment processor’s dashboard, and the settings for retries, the card updater and the cancel flow.
2. A subscription exportEvery subscription with start date, first offer, cadence, status, end date, how it ended, pause and resume dates, and the cancellation reason if one was given.
3. A charge exportEvery renewal attempt with date, amount, result and decline code.
4. Contribution per orderFrom finance, for subscription and one-time orders, so the discount can be priced.
5. The historyEvery change to the offer, discount, price, cadence options and cancel flow since launch, with dates. Gaps are findings.
6. A name at counselSomeone who will look at the signup and the cancel path within a week.

Do this

• Ask for all six in one message today. Whatever comes back slowest is where the program knows least about itself.

• The Customer-Base AuditPeter Fader, Bruce Hardie and Michael Ross, 2022How to read your own customer data: cohorts, survival and the reports that lie. The best companion to part two.
• SubscribedTien Tzuo with Gabe Weisert, 2018The case for the subscription model from the founder of a billing company. Read it for the ambition, and read parts three and four of this guide for the fine print.
• The Automatic CustomerJohn Warrillow, 2015A practical survey of subscription models beyond boxes, useful for deciding which kind you’re building.
• The Membership EconomyRobbie Kellman Baxter, 2015On memberships as relationships rather than billing arrangements, and what members expect in return.
• Trustworthy Online Controlled ExperimentsRon Kohavi, Diane Tang and Ya Xu, 2020How to run the holdouts this guide keeps asking for, and how to believe their results.

And the research: DellaVigna and Malmendier (2006) on paying not to go to the gym; Einav, Klopack and Mahoney (2025) on inertia and card replacement; Miller, Sahni and Strulov-Shlain (working paper, 2026) on auto-renewal defaults; Lewis (2006) on acquisition discounts; Datta, Foubert and Van Heerde (2015) on free-trial customers; Kivetz, Urminsky and Zheng (2006) and Nunes and Drèze (2006) on progress toward a goal; Ascarza, Iyengar and Schleicher (2016) and Ascarza (2018) on retention offers; Iyengar, Park and Yu (2022) on paid membership. Full references are in Appendix C.

Andrew Lauchner runs Growth Legend, embedding inside consumer brands to own lifecycle, email and SMS, and revenue operations. He wrote The Second Order, on turning first-time buyers into second-time buyers; Close the Loop, on getting customers to bring the next customer; The First Offer, on the offer that wins the first order; The Whole Machine, on the fundamentals of DTC growth; and The Honest Test, on conversion work and testing.

As Senior Director of Growth and Retention Marketing at Gallery Furniture, he rebuilt the customer journey and the sales playbooks together. He has worked on growth and retention at Binance and 3Commas, and has been Head of Growth and Retention at Greatness Wins and at Nexus Agriscience.

The methods marked “from my work” come from auditing and running client subscription and membership programs in 2025 and 2026. Clients aren’t named and their numbers aren’t here.

Andrew answers every note from people running subscription programs, including those looking for someone to own one. Write to andrew@growthlegend.com or message him on LinkedIn.

Everything below runs on two tables: one row per subscription, and one row per charge attempt. Most subscription apps can export both. The queries are written for Postgres; other warehouses need small changes to the date functions.

The data contract

Treat a pause as an end on the day it starts. Record a resume as a new subscription row with reactivation = true and the original ID in previous_subscription_id, and leave reactivation rows out of the survival table and the first-renewal-by-offer query. That’s close to how Peloton counts pauses, and it stops a pause-heavy cancel flow from hiding churn.

Monthly churn, split by cause

-- losses among subscriptions active on the 1st, by how they ended
-- months: a calendar table with one row per month start (m)
WITH base AS (
  SELECT k.m, s.subscription_id, s.ended_at, s.end_type
  FROM months k
  JOIN subscriptions s
    ON s.started_at < k.m
   AND (s.ended_at IS NULL OR s.ended_at >= k.m)
)
SELECT m,
       COUNT(*) AS active_on_1st,
       COUNT(*) FILTER (WHERE ended_at < m + INTERVAL '1 month'
                          AND end_type = 'voluntary')      AS voluntary,
       COUNT(*) FILTER (WHERE ended_at < m + INTERVAL '1 month'
                          AND end_type = 'failed_payment') AS failed_payment,
       COUNT(*) FILTER (WHERE ended_at < m + INTERVAL '1 month')::numeric
         / COUNT(*)                                        AS churn
FROM base
GROUP BY m
ORDER BY m;

Pauses fall into churn but not into either named cause; report them on their own line.

The survival table

-- share of each start month still active when renewal r was due
SELECT DATE_TRUNC('month', started_at) AS start_month,
       k.r AS renewal,
       COUNT(*) AS subscriptions,
       AVG(CASE WHEN ended_at IS NULL
                  OR ended_at > started_at + k.r * cadence_days * INTERVAL '1 day'
                THEN 1.0 ELSE 0 END) AS still_active
FROM subscriptions
CROSS JOIN (VALUES (1), (2), (3), (6), (12)) AS k(r)
WHERE started_at + k.r * cadence_days * INTERVAL '1 day' <= CURRENT_DATE
  AND NOT reactivation
GROUP BY 1, 2
ORDER BY 1, 2;

The WHERE line keeps each subscription out of a column until that renewal was due, so young groups show blanks, not zeros. Measuring “still active” rather than “paid the renewal” counts a skipped delivery as survived, which is what it is. If many subscribers change cadence, rebuild the due dates from the cadence history table.

First renewal by offer

SELECT first_offer,
       COUNT(*) AS subscriptions,
       AVG(CASE WHEN ended_at IS NULL
                  OR ended_at > started_at + cadence_days * INTERVAL '1 day'
                THEN 1.0 ELSE 0 END) AS reached_first_renewal
FROM subscriptions
WHERE started_at + cadence_days * INTERVAL '1 day' <= CURRENT_DATE
  AND started_at >= CURRENT_DATE - INTERVAL '6 months'
  AND NOT reactivation
GROUP BY first_offer
ORDER BY subscriptions DESC;

Saves still active at 60 days

-- cancel_attempts: one row per attempt, with outcome and the option taken
SELECT c.save_option,
       COUNT(*) AS saves,
       AVG(CASE WHEN s.ended_at IS NULL
                  OR s.ended_at > c.attempted_at + INTERVAL '60 days'
                THEN 1.0 ELSE 0 END) AS active_at_60_days
FROM cancel_attempts c
JOIN subscriptions s USING (subscription_id)
WHERE c.outcome = 'saved'
  AND c.attempted_at <= CURRENT_DATE - INTERVAL '60 days'
GROUP BY c.save_option
ORDER BY saves DESC;

Put the same query next to one for subscribers who never tried to cancel, from the same months. The gap between saved subscribers and untouched ones is the true measure of what a save is worth.

The pre-renewal reminder

SUBJECT    Your Next Delivery Ships Thursday
PREVIEW    skip, change the date, or swap in one tap

Your next [product] ships Thursday, [date], for $[amount].

Running low?        Nothing to do. It's on its way.
Still have plenty?  [Skip this one]   [Ship in 2 weeks]
Using it faster
or slower?          [Change how often]
Want something
different?          [Swap]
Done with it?       [Cancel online]

Questions? Reply to this email. A person reads it.

[If an introductory price is ending, send this at least
7 days ahead, say so first, and keep the cancel link:
"Your first-order price was $[x]. From this delivery
it's $[y]."]

The failed-payment notices

1. SAME DAY
SUBJECT    Your Card Didn't Go Through
PREVIEW    your subscription is still here
Your card didn't go through for your [date] delivery,
so it's on hold. Your subscription is still here.
[Update your card]
We'll try again on [date]. If your card is fine,
you don't need to do anything.

2. BEFORE THE NEXT ATTEMPT
Same facts, shorter. The retry date, the button.

3. FINAL
SUBJECT    Your Subscription Ends On [Date]
We've tried your card [n] times. If we can't charge it
by [date], your subscription will end on that date.
[Update your card]   [Cancel instead]

SMS (only to subscribers who opted in to texts)
[Brand]: your card didn't go through for your next
delivery. Update it here: [link]. We'll retry [date].
Reply STOP to opt out.

The cancel flow

SCREEN 1   Cancel your subscription
           What's the main reason? (optional)
           ( ) I have too much
           ( ) It's too expensive
           ( ) I'm not using it
           ( ) I'm traveling or moving
           ( ) I want something different
           ( ) Something else: [       ]
           [Continue]   [Cancel my subscription now]

SCREEN 2   (only if a reason was chosen. Ask first, as Minnesota
           requires: "Would you like to see an option first?"
           If they say no, go straight to screen 3.)
           One option that fits the reason:
           too much     -> [Skip next]  [Every 8 weeks instead]
           expensive    -> [Smaller size] [Every 8 weeks]
           not using    -> [How to get the most from it] [Swap] [Pause]
           traveling    -> [Pause for 1 month] [Pause for 2 months]
           different    -> [Swap]
           Always beside it:  [No thanks, cancel my subscription]

SCREEN 3   Your subscription is canceled.
           Your last charge was [date]. Nothing more will ship.
           (A confirmation email follows, with the same facts.)

The yearly reminder

SUBJECT    Your Subscription, One Year On
PREVIEW    what you get, what you pay, and how to change it

You subscribe to [product], delivered every [n] weeks,
for $[amount] per delivery. In the last 12 months you
received [n] deliveries.

Change how often, skip, swap or pause:  [Manage]
Cancel anytime, online:                 [Cancel]

Subject lines in Title Case and preview text in lowercase are a house style, not a rule; keep whatever your brand uses. What matters is that the facts come first and the buttons do what they say.

Filings and company results (chapters 4, 5, 8, 11, 12)

• Blue Apron Holdings, Inc. Prospectus (Form 424B4), June 28, 2017; Form 10-Q for the first quarter of 2018; first quarter 2018 results; fourth quarter and full year 2022 results; agreement to be acquired by Wonder, September 2023; completion of the tender offer, November 2023. CNNMoney on the range cut, June 28, 2017.
• Chewy, Inc. Fourth quarter and fiscal year 2025 results, March 2026; Form S-1, 2019, for the definition of Autoship customer sales; Autoship terms, read September 2026.
• Peloton Interactive, Inc. Results for the second, third and fourth quarters of fiscal 2026, for the churn definition and values.

Research on subscriptions and retention offers (chapters 2, 3, 6, 8, 9, 11, 12)

• Stefano DellaVigna and Ulrike Malmendier. “Paying Not to Go to the Gym.” American Economic Review 96(3), 2006.
• Liran Einav, Ben Klopack and Neale Mahoney. “Selling Subscriptions.” American Economic Review 115(5), 2025. Author copy.
• Adam Miller, Navdeep Sahni and Avner Strulov-Shlain. “Sophisticated Consumers with Inertia: Long-Term Implications from a Large-Scale Field Experiment.” Working paper, version of January 2026.
• Michael Lewis. “Customer Acquisition Promotions and Customer Asset Value.” Journal of Marketing Research 43(2), 2006.
• Hannes Datta, Bram Foubert and Harald J. Van Heerde. “The Challenge of Retaining Customers Acquired with Free Trials.” Journal of Marketing Research 52(2), 2015.
• Eva Ascarza, Raghuram Iyengar and Martin Schleicher. “The Perils of Proactive Churn Prevention Using Plan Recommendations.” Journal of Marketing Research 53(1), 2016.
• Eva Ascarza. “Retention Futility: Targeting High-Risk Customers Might Be Ineffective.” Journal of Marketing Research 55(1), 2018.
• Raghuram Iyengar, Young-Hoon Park and Qi Yu. “The Impact of Subscription Programs on Customer Purchases.” Journal of Marketing Research, 2022.
• Ran Kivetz, Oleg Urminsky and Yuhuang Zheng. “The Goal-Gradient Hypothesis Resurrected.” Journal of Marketing Research 43(1), 2006. Joseph C. Nunes and Xavier Drèze, “The Endowed Progress Effect.” Journal of Consumer Research 32(4), 2006.
• Tony Chen, Ken Fenyo, Sylvia Yang and Jessica Zhang. “Thinking inside the subscription box: New research on e-commerce consumers.” McKinsey & Company, February 2018.

Law and card-network rules (chapters 2, 7, 10)

• Restore Online Shoppers’ Confidence Act, 15 U.S.C. § 8403.
• Federal Trade Commission. Final click-to-cancel rule, October 16, 2024; Negative Option Rule docket; conforming rule, February 12, 2026; advance notice of proposed rulemaking, March 13, 2026.
• Custom Communications, Inc. v. FTC, No. 24-3137 (8th Cir. July 8, 2025).
• Federal Trade Commission. Settlement with Amazon, September 25, 2025; amended complaint, September 2023. Chegg, September 2025. Shutterstock, May 2026. Amended complaint against Uber, December 2025.
• US Department of Justice. Adobe settlement, March 13, 2026.
• California Business and Professions Code § 17602, as amended by AB 2863 (2024); California Attorney General consumer alert, 2025.
• Minnesota Statutes § 325G.57, § 325G.58 and § 325G.60; effective date per Kelley Drye’s 2025 round-up.
• New York General Business Law § 527-a, as amended in 2025; Kelley Drye on the amendment.
• Visa. New rules for subscription merchants, 2020; rules for declined transaction resubmission, effective April 2021; Visa Account Updater FAQ; on network tokens; Visa tokens surpass physical cards, August 2022.
• Mastercard. Automatic Billing Updater. Consumer Reports on the 2019 free-trial rule. Merchant advice codes and retry fees as documented by processors and acquirers: Braintree, Yuno, Adyen on scheme fees, Adyen on Visa integrity fees, TabaPay and TD Bank, 2026. Adyen, on network tokens, 2023.

Vendor data (chapters 5, 6, 8, 9, 10, 11, 13)

• Recharge. “Subscription churn is front-loaded”, 2026; on first-order discounts, July 2026; on $0.99 first orders, August 2026; “The quarterly upgrade”, August 2026; Subscription Trend Report 2026; 2023 State of Subscription Commerce; on pause, 2023.
• Recurly. Churn rate benchmarks, July 2026 data; guide to decline reasons; on pausing, 2020; 2026 State of Subscriptions, January 2026.
• Stripe. “How we built it: Smart Retries”, January 2024; Stripe Billing; Smart Retries documentation.
• Chargebee. Subscription retention insights, updated 2024.
• Amazon. Subscribe & Save for sellers, read September 2026.
• Bloomberg Second Measure. Blue Apron customer retention, June 2016; meal kit customer retention, January 2023.