DTC Subscription Retention: A Playbook by Subscription Model

Most DTC retention guides give you seven tactics in a list. The list is the problem.

A discount works for a curation box because the cancellation reason is product fatigue. The same discount fails for a replenishment subscriber whose cancellation reason is "too much stockpile." A retention strategy that ignores the subscription model is wrong about half the time.

The three subscription models — replenishment, curation, and access — were defined by McKinsey's 2018 consumer survey of more than 5,000 US shoppers, and the framework still organizes how operators think about the space today. What's changed since 2018 is the data layer underneath it: SUBTA now sizes the broader subscription, membership, and loyalty economy at $3 trillion (with a 10% year-over-year decline in active subscriptions in 2024 underscoring that the market has matured into one where retention matters more than acquisition), Chewy reports 83% of its $12B+ in net sales flowing through Autoship, and platforms like Recharge and Recurly are publishing cohort behavior at a granularity that didn't exist a decade ago. The framework holds. The execution playbooks have sharpened.

This guide is for operators who already know what their churn rate is and want to know what to do about it.

For benchmarks across verticals, see our churn rate benchmarks. For general retention strategies that apply across subscription types, see how to reduce subscriber churn. This guide fills the gap with the model-specific patterns that determine whether a DTC subscription business compounds or leaks.

Sources: McKinsey 2018 subscription survey for the three primary models; Bernstein and Second Measure analyses of HelloFresh cohorts for the meal-kit row; Costco FY2025 10-K for the best-in-class access reference (89.8% worldwide renewal). Detailed source tables appear below.

McKinsey's 2018 subscription survey, which defined the three-model framework still used today, found that curation was the largest category by subscriber share (55%), followed by replenishment (32%) and access (13%). The mix has continued to evolve — Amazon Subscribe & Save, Chewy Autoship, and pet-supply replenishment have grown significantly — but the framework's analytic value is in the why people cancel, not the what percentage subscribe. Cancellation causes split cleanly along model lines, and so do the retention plays.

Replenishment: the convenience model

Replenishment subscriptions automate the purchase of products the customer is already buying — pet food, vitamins, razors, household consumables. The value proposition is convenience and, often, a modest savings versus one-time purchase.

McKinsey found that 45% of replenishment subscribers had been subscribed for at least one year — roughly ten percentage points higher than curation or access. The convenience model retains the longest because the product fulfills a genuine recurring need. When the product aligns with actual consumption, there's no reason to cancel.

The cancellation reasons that do drive replenishment churn split into two patterns by cohort stage.

The cohort-1 killer: cadence mismatch. The subscriber needs the product every six or seven weeks. The default cadence at signup was every four. The product piles up. By month two or three, the customer cancels — not because they don't want the product, but because the inventory has outrun their use of it. McKinsey's data is specific here: subscribers say they cancel when "products pile up or they can't customize order volumes to match their actual requirements." This is a configuration problem, not a value problem.

The long-cohort killer: life-stage shift. The kid outgrows diapers. The pet dies. The household composition changes. These cancellations are not preventable — but they are predictable, and they are the right cancellations to lose gracefully. A win-back-ready churn is worth more than a forced retention.

The retention levers that work for replenishment are operational: easy cadence change, quantity adjustment per shipment, one-click skip, and proactive prompts that flag a likely cadence mismatch before the customer notices it. Discount-heavy save offers are largely wasted on replenishment subscribers — the customer isn't price-sensitive, and "20% off your next order" doesn't solve a stockpile problem.

The replenishment benchmark — Chewy Autoship. In its FY2025 annual report, Chewy disclosed that Autoship customer sales reached $10.5 billion, or 83.3% of net sales, up from 79.2% the prior year. Active customers grew to 21.3 million. Net sales per active customer rose to $591. Chewy explicitly identifies Autoship as "a key driver of recurring net sales and customer retention" in its SEC filings. There is no marketing-magic explanation for these numbers. Chewy made cadence flexibility, account-updater services, and operational reliability the entire product — and the result is what a replenishment business looks like when it's run well.

Table A — 1-year retention by subscription model. Source: McKinsey, "Thinking inside the subscription box" (Feb 2018), based on a US consumer survey of 5,093 respondents.

Curation: the discovery model

Curation subscriptions deliver new or surprising selections on a recurring basis — beauty boxes, meal kits, hobby boxes, apparel discovery. The value proposition is novelty: every shipment is supposed to feel like a gift, not a refill.

The structural challenge is that novelty fades. McKinsey's survey found that 28% of curation and access subscribers cited "personalized experience" as the most important reason for continuing to subscribe, and that consumers expected personalization to improve over time as the service learned their preferences. The brands that win are the ones that turn early subscriber feedback into measurably better month-six and month-twelve selections. The brands that don't, churn.

The cohort-1 killer: the second-shipment cliff. The first box was discounted, hyped, and engineered for unboxing. The second one arrives at full price, with no introductory framing, and often less surprise. A measurable share of subscribers cancel right here — before they've even had a chance to form a habit. We treat this dynamic as its own section below because it's so distinctive to ecom.

The long-cohort killer: product fatigue and personalization decay. By month eight, the curation that delighted in month one feels predictable. The subscriber has accumulated products they don't use. They have feedback they've never been asked for, or they gave feedback that nothing seems to have been done with. The cancel button starts to feel like a relief.

The retention levers that work for curation are different from replenishment's. Personalization that visibly compounds: if the subscriber rated three lipsticks as too dark in month two, month three should reflect that. Preview emails 7–10 days before billing: anticipation is a retention lever. Build-a-box and swap mechanics: let subscribers exert control over what they receive without canceling. Pause-not-cancel placement: discussed in detail below.

The lever that does not work for curation is cadence flexibility on its own. The subscriber's problem isn't that the box arrives too often; it's that the box isn't delighting them anymore. Offering "delivery every other month" to a fatigued curation subscriber just delays the cancellation.

Subscription box benchmark numbers from Swell's 2026 analysis: monthly churn rates of 10–15%, with 44% of cancellations happening in the first 90 days. Top performers maintain rates below 3%. The spread between average and best-in-class is wider in curation than in any other subscription model — which means the operational ceiling is high if you build the right loops.

Access: the membership model

Access subscriptions are memberships. Subscribers pay a recurring fee in exchange for benefits — discounts, exclusive content, free shipping, priority access. The product isn't shipped; the standing is.

This model behaves differently from the other two because there's no physical fulfillment cycle to skip or swap. The retention question is whether the member is actually using what they're paying for.

The cohort-1 killer: failure to activate the benefit. A member who joins a discount-club access program and then doesn't shop in the first 30 days has just rented an unused membership. The likelihood they remain through renewal is low. The first 60 days of an access subscription are an onboarding and engagement problem more than a fulfillment problem.

The long-cohort killer: renewal shock plus alternatives. Annual access plans hide their cost — the member doesn't see the charge every month, so the relationship feels free until renewal day. If by renewal the member can't articulate what they got from the membership, they cancel. The same is true if a competing free or cheaper alternative has emerged in the interim.

Access retention is run through usage reminders, exclusive content drops, renewal previews that surface the dollar value the member captured ("you saved $312 this year"), and tiered benefits that reward longer-tenured members. None of this looks like the replenishment or curation playbook. Skip mechanics are irrelevant; there's nothing to skip.

The benchmark example is Costco. Costco's FY2025 10-K reported a worldwide membership renewal rate of 89.8% and a US/Canada renewal rate of 92.3% across 81 million paid memberships. Costco's own framing in the filing is that "membership loyalty and growth are essential to our business." The lesson is not "be Costco" — the financial scale and warehouse-format moat are not generally replicable. The lesson is that access subscriptions retain in proportion to how often the member touches the benefit, and Costco members touch the benefit constantly. Lower-utilization access plans, where the typical member rarely activates the perks, churn at curation-like rates.

Across all three models, some share of cancellations were never decisions. The card expired, the bank declined the charge, the account had insufficient funds at billing time — and the subscription quietly ended without the subscriber ever knowing.

Recurly's 2026 State of Subscriptions, based on 2,200+ merchants and 76 million subscribers, places involuntary churn at roughly 20–40% of total cancellations across DTC. Swell's analysis of subscription box operators puts it at up to 68% — meaning more than two-thirds of subscription box cancellations were not voluntary at all. These are subscribers who wanted to stay and lost their subscription anyway.

The involuntary-churn tax varies by model in instructive ways.

Access subscriptions show the lowest involuntary rates. Members tend to keep their primary credit card information current because they're paying for ongoing standing they value. Recovery is mostly mechanical.

Replenishment subscriptions sit in the middle. Mature operators like Chewy invest heavily in account updater services, smart retry windows timed to payroll dates, and pre-dunning emails that prompt the customer to update payment before a charge is attempted. The infrastructure is unglamorous and the recovery numbers are strong.

Curation and subscription boxes carry the highest involuntary share, especially at lower price points. Low-AOV boxes lose meaningful percentages of revenue to failed payments because the subscriber's commitment to this particular recurring charge is weaker, and an expired card often goes unupdated. Pre-dunning notifications a week before billing, account updater services, and smart retry windows (rather than the same-time-of-day retry) materially change the recovery rate.

The diagnostic move every operator should make: compute the voluntary/involuntary split of cancellations for the last 90 days. If the involuntary share is above 30% and you don't have smart retry logic, account updater coverage, and pre-dunning notifications running, that's the cheapest churn reduction available to you — and it's purely an infrastructure decision, not a marketing or product one.

Of all the cohort dynamics in DTC subscriptions, the one that matters most is what happens between order 1 and order 2.

McKinsey's survey data established the headline number: more than one-third of subscription e-commerce customers cancel within three months of signup, and more than half cancel within six. The meal-kit category, McKinsey noted specifically, sees cancellation rates of "60 to 70 percent and higher" within the first six months.

The meal-kit data became more precise in subsequent years. Bernstein Research's 2022 deep-dive on HelloFresh ("Paying People to Eat") documented a business built on average discounts above 20%, with 90% of customers not purchasing in Q4 — a pattern Bernstein characterized as cohorts being trained to churn and then reactivated only when new discounts appeared. Independent analysis by Second Measure of HelloFresh cohorts found roughly 50% of customers cancelling after the first month, 85% by month six, and 90% by month 12.

Table B — Meal-kit cohort decay (HelloFresh). Sources: Bernstein Research, "HelloFresh: Paying People to Eat" (Aug 2022); Second Measure cohort analysis of HelloFresh subscribers.

The meal-kit numbers are an extreme version of a pattern that shows up across curation broadly. The cause is structural: when acquisition is discount-heavy, a meaningful share of the cohort isn't a customer at all — they're a discount-taker. The first shipment was the deal. The second shipment, at full price, is a different value proposition than the one they signed up for. The cancellation between order 1 and order 2 was decided at acquisition.

This has implications most DTC operators don't fully internalize.

Discount-heavy acquisition is renting, not buying. If your first-box discount is 50% off and your second-shipment retention is 40%, you didn't acquire that 60% — you paid for 60% of them to take a discounted product and leave. The CAC math has to be done against the cohort that survives the second shipment, not the cohort that signed up.

The second shipment is the most important shipment. It is where the subscriber decides whether the relationship is a thing or a fling. The unboxing of shipment two should be deliberately better than shipment one — not a duller version of the trial. Best-in-class curation operators write content, ship personal notes, and engineer the contrast on shipment two to feel like the brand is investing in the customer, not the other way around.

The third shipment is the inflection point. Recharge's State of Subscription Commerce reports surface a consistent pattern: subscribers who reach shipment three retain at meaningfully better rates afterward. The cohort curve flattens. Whatever the right phrase is — habit formation, identity adoption, value confirmation — it locks in around shipment three for most curation categories.

The 90-day rule. Subscribers who survive past day 90 retain at meaningfully better rates than the broader population. Swell's subscription box data attributes 44% of all box cancellations to the first 90 days. Inverted, that's the number to manage against: if you can flatten the early curve, the rest of the cohort behaves like a different business.

The operational implication is that retention infrastructure spent on month-eight subscribers is often misspent. The leverage is at the start of the cohort, not the middle.

A DTC subscription business on Shopify, BigCommerce, or a comparable stack is running through a subscription-app layer — Recharge, Ordergroove, Stay AI, Loop, Skio, Bold, or one of several others — that sits between the storefront and the fulfillment system. The native Shopify Subscriptions API handles billing cycles and basic recurring orders, but the retention features that actually move the needle are added by the app layer on top.

There are four mechanics that meaningfully change retention outcomes, and merchants evaluating subscription apps should evaluate against these four — not against feature counts on a vendor landing page.

Portal-resident skip, swap, and cadence change. Subscribers who can self-serve these actions stay materially longer. Ordergroove's internal platform data shows skip availability correlates with 135% longer subscriber lifetimes; swap availability correlates with 71% longer lifetimes. Recharge's State of Subscription Commerce reports document tens of millions of skip and swap actions across their merchant base annually — these are not edge-case features. The implementation detail that matters: whether these actions are one tap from the portal home, or buried under sub-menus. Friction here directly converts into cancellations.

Pause-before-cancel placement. When a subscriber clicks cancel, what they see next determines whether they actually cancel. Recurly's 2026 data shows that 25% of would-be churners choose pause instead when the pause option is presented in the cancellation flow. Chargebee's 2025 global consumer research found 58% of consumers prefer pause to cancel when given a real choice. The mechanic is simple: subscribers in temporary friction (going on vacation, going through a budget crunch, drowning in stockpile) need an off-ramp that isn't a full cancellation. Subscription apps that hide pause behind the cancel button capture this saved volume; apps that don't, lose it permanently.

Dunning sophistication. The difference between a subscription app that retries failed payments three times at the same time of day and one that uses smart retry windows, pre-failure alerts, account updater services, and SMS-plus-email recovery cadences is the difference between 30% involuntary churn and 15%. The technical bar is straightforward; the operational difference is large.

Conditional cancel-flow logic. A one-size-fits-all cancellation flow with the same generic discount offer treats a month-two subscriber the same as a month-twenty-four subscriber. They are not the same customer. Apps that support conditional logic — different offers based on order count, cancellation reason, LTV tier, or product category — convert save offers at materially higher rates because the offer matches what the subscriber actually needs.

The structural reality of the subscription-app layer is that no single app is best for every brand. Recharge's strength is platform maturity and ecosystem depth. Ordergroove's strength is the relationship-commerce framing and primary research on subscriber behavior. Stay AI, Loop, Skio, and Bold each have positioning that fits certain merchant profiles better than others. The evaluation question for a merchant isn't which app is best, but which app's retention surface area maps most cleanly to my subscriber base's cancellation patterns.

This is also the layer at which retention-specialized products like SubJolt operate. Cancellation flow design, reactivation banners, payment recovery, and cancellation analytics sit on top of whichever subscription app a merchant chose — extending the retention toolkit without replacing the billing infrastructure.

Our benchmarks page has the cross-vertical numbers — monthly churn by category, involuntary share by industry, retention curves where they're published. Use it to locate yourself in the broader landscape.

Two diagnostic moves that matter more than benchmarking:

Compute your own second-shipment retention. What percentage of subscribers who completed shipment one also complete shipment two? If you don't know this number, your retention strategy is operating without its single most important leading indicator.

Split voluntary from involuntary churn. If you can't produce a 90-day report broken into voluntary cancellations and failed-payment churn, your retention investment is probably mis-allocated. The fixes for each are different — and the involuntary fixes are usually cheaper.

The cohort math is also worth doing in your own dashboard, even at small scale: cohort retention at M1, M3, M6, M12 by acquisition channel and by first-discount level will tell you which acquisition spends are generating subscribers and which are renting them.

For the calculations themselves, our churn rate calculator and retention rate calculator will run the numbers; the LTV calculator will translate retention improvements into lifetime value.