DTC Subscription Retention (2026)

If you run an annual retention program on a consumer subscription, you're running it after the fact. More than half of e-commerce subscription-box buyers have canceled by month six, and more than a third by month three (McKinsey Subscription Box Research, 2018). Specialty streaming keeps 72% of subscribers at month one and 54% by month three (Antenna Specialty SVOD Landscape, 2026). Meal-kit cohorts that were 57% to 69% still buying in the month after their first order are down to between 5% and 15% by month eleven (Second Measure Meal Kit Competition, 2023).

Three verticals, three methods, one shape. A survey of consumers, a measured panel of streaming subscriptions and a card-transaction panel of meal-kit buyers all describe loss that is front-loaded and mostly complete before a SaaS-shaped retention program would have run its first review.

That one fact reorders what's worth doing. Win-back campaigns, annual plan pushes and loyalty tiers all operate on subscribers who are, statistically, already gone. The work that changes the outcome happens in the first two or three billing cycles. So measure the first three billing cycles before you fund anything that runs later.

One correction worth flagging up front: the subscription model, replenishment against curation against access, turns out to be the weakest of the variables here, and the source usually cited for its importance says so directly.

The three panels below measure different things and can't sensibly be laid on one axis. McKinsey asked consumers what they had done; Antenna observes subscriptions; Second Measure observes card transactions. What they share is the shape, not the level.

The drop is early and steep

Specialty streaming holds 72% of a cohort at month one and 54% by month three (Antenna Specialty SVOD Landscape, 2026), so more than a quarter of every cohort is gone inside thirty days on our subtraction from those survival figures. Premium streaming, which is the larger and more established end of the same market, still loses 60% of its sign-ups within a year (Antenna State of Subscriptions: Adds and Ads, 2026). In physical goods the pattern is at least as sharp, though the measure underneath it is different: 57% to 69% of the January 2022 meal-kit cohorts bought again in the month after their first order, and 5% to 15% were still buying at month eleven (Second Measure Meal Kit Competition, 2023). That counts purchases, not open subscriptions, so a subscriber who skipped a month reads here as lost.

Why annual figures hide it

An annual churn rate is a single number covering a period in which the curve is at its steepest at the start and nearly flat by the end. Two businesses reporting the same annual figure can have completely different problems: one losing steadily all year, one losing almost everyone in the first quarter and holding the remainder. Mostly it is the second that onboarding work can fix, and the annual number can't tell you which you are. Cohort survival by month can, which is why every source in this section reports one.

You can put your own number on a comparable annual basis with the retention rate calculator, but for a consumer subscription it is worth reading month by month first.

The model is usually treated as the thing that determines a retention playbook, with replenishment, curation and access each getting their own treatment. The source cited for that framing doesn't support it.

McKinsey's own finding is that about 40% of e-commerce subscribers have canceled a subscription, and that these rates are similar across replenishment, curation and access services (McKinsey Subscription Box Research, 2018). The model does not sort subscribers into good and bad retention on the measure the report leads with.

Where the model does show up

It isn't nothing. On a different measure, one-year membership, replenishment holds 45% against roughly ten points lower for curation and access (McKinsey Subscription Box Research, 2018). That's a real gap, and it points the way you'd expect: consumable goods on a predictable cycle hold better than surprise boxes.

But ten points is small next to what sits inside a single model. Meal kits, a category within curation, run cancellation of 60% to 70% and higher within six months on the same report's reading (McKinsey Subscription Box Research, 2018). The variance within a model is larger than the variance between models, so a retention strategy chosen on the model is being chosen on the weaker variable.

One caveat on all of it: this is a single survey fielded in November 2017, self-reported rather than measured, and it has not been repeated. It's used here for the comparison between its own segments, which is the part least likely to be distorted by recall, and not as a level to benchmark against.

Everything above is consumer subscription broadly. If you ship physical goods on Shopify, there is a narrower benchmark, and it is the one to read first.

Average monthly churn across 20,000 brands, by category, for calendar 2025. The spread is tighter than the category labels suggest: Health and Wellness at 8.8%, Beauty and Personal Care at 8.4%, Pets at 8.1% and Food and Beverage at 7.7% sit within 1.1 points of one another (Recharge Subscription Trend Report, 2026). Home is the outlier at 3.9% (Recharge Subscription Trend Report, 2026), and the report is careful not to read that as health: it reports orders per subscriber down 4.97% year on year in the category, calls the stability something that is starting to mask stagnation, and names passive loyalty rather than cancellation as the pressure point (Recharge Subscription Trend Report, 2026).

So the finding is not that your category decides your retention. Four of the five behave alike, and the one that does not is the one whose subscribers order least often.

These are not the same unit as the annual figures

Stripe's industry cut puts merchandise at 39% annual churn, and furniture at 28%, the lowest rate of any industry it reports and also the one with the highest involuntary share at 46% (Stripe Churn Benchmarks, 2025). Those are annual rates on monthly-billed subscriptions only; the figures in the table are monthly rates for the median brand in a category. The two do not sit on one axis. Compounded, the table's range runs from roughly 38% to roughly 67% a year on our arithmetic, which places Stripe's merchandise figure at the bottom of that range rather than in conflict with it.

Two things follow for a merchant. Compare against your category rather than against a consumer-subscription average, and check the period on any figure before you put it beside your own.

If the model is the weak variable, the strong ones are structural: how the subscription is packaged, and how the subscriber was acquired.

Bundling moves it about thirty points

Twelve-month survival for the Disney+, Hulu and HBO Max bundle runs 59%, against 31% for Disney+ standalone and 28% for HBO Max standalone (Antenna Premium SVOD Year in Review, 2025). That's the largest single effect here, and it's a packaging decision rather than a retention campaign. Bundled subscriptions have gone from 14% of Premium SVOD in late 2023 to 27% in late 2025 (Antenna Premium SVOD Year in Review, 2025), so the market has noticed.

The ad tier barely moves it, and the direction reverses

It's worth stating carefully, because it's easy to quote in one direction. For Netflix specifically, ad-supported subscribers survive twelve months at 54% against 52% for ad-free, while category-wide the ordering flips to ad-free 39% against ad-supported 37% (Antenna State of Subscriptions: Adds and Ads, 2026). Two points either way is not a retention strategy, and anyone citing one of those pairs without the other is reporting a preference, not a finding.

Promotional acquisition did not produce worse subscribers

The standard advice is that discount-acquired subscribers churn harder. In this panel they did not. Sign-ups from Black Friday 2024 promotions survived nine months at 57%, against 43% for 2024 sign-ups overall (Antenna Premium SVOD Year in Review, 2025).

Read that narrowly. It is one panel, one seasonal window, one vertical, and Black Friday shoppers are a high-intent group who may have been going to subscribe anyway. It is not evidence that discounting improves retention generally. It is evidence that the blanket claim doesn't hold everywhere, and shouldn't be asserted without a number behind it.

A cancellation reads as terminal in most retention reporting. In consumer subscriptions it frequently isn't.

Of Premium SVOD subscribers who canceled in 2024, 23% had returned within three months, 32% within six and 42% within twelve on a weighted average, with Netflix at 50% (Antenna Premium SVOD Year in Review, 2025). Across Recurly's merchant base, 20% of new sign-ups had subscribed before, rising to 27.7% in digital media and entertainment (Recurly State of Subscriptions, 2025).

Two consequences follow. The first is measurement: a business counting gross cancellations without tracking returns is overstating its losses by a margin that, in streaming, approaches half. The second is that the win-back window is short. Most of the return happens in the first three months, so a quarterly re-engagement cycle catches a good deal of it and an annual one catches almost none.

Two mechanisms sit outside the usual retention conversation and account for a meaningful share of what looks like churn.

Pause is chosen, not just tolerated

Where merchants offer it, 25% of subscribers who would otherwise cancel choose to pause (Recurly State of Subscriptions, 2025), and 38% of surveyed consumers say they would prefer pausing to canceling (Recurly State of Subscriptions, 2026). The returns are real, not nominal: in the peak month 61.9% of paused subscribers resumed (Recurly State of Subscriptions, 2026). Pause usage in digital media and entertainment grew 330% in a year (Recurly State of Subscriptions, 2025), which says more about how few merchants offered it than about a change in what consumers wanted. The demand is not in doubt: 58% of consumers say they have paused a subscription instead of canceling it, and 79% say they need an easy pause before they will subscribe at all (Chargebee Global Consumer Insights, 2026). That is a survey of 1,454 US and UK consumers across consumer subscriptions generally, with meal kits inside the frame but no physical-goods cut of its own.

Some of the churn was never a decision

At low order values a large minority of cancellations are payment failures, not choices: 35% of all churn under a $10 order value is involuntary (Stripe Churn Benchmarks, 2025). Consumer subscriptions cluster at exactly that end of the price range, so the involuntary share matters more here than in most categories. The full cut across order values is on the churn rate benchmarks page.

Four things to establish before comparing yourself to anything on this page.

Measure by cohort month, not by period

Take everyone who started in a given month and track what share is still active at month one, three, six and twelve. That's the only shape comparable to the sources here, and the only one that separates an onboarding problem from a steady leak. A period churn rate averages the two together.

Match the vertical before the number

Specialty streaming at 27% twelve-month survival (Antenna Specialty SVOD Landscape, 2026) and Premium streaming, which publishes a 60% first-year cancellation rate rather than a survival figure and so sits near 40% surviving on our inversion of it (Antenna State of Subscriptions: Adds and Ads, 2026), are different businesses, and meal kits are different again. Consumer subscription isn't one benchmark, and the gap between two categories here is larger than the gap between any two subscription models. If you sell physical goods and let subscribers skip a delivery, read the meal-kit row knowing it scores a skipped month as a lost one, so a flexible program benchmarks worse against it than it deserves to.

Net your losses against returns

If a fifth to a half of your cancellations come back, gross cancellation overstates the problem and understates the value of the win-back window. Track returning subscribers separately from new ones. Most billing systems do not do this by default.

Separate what was chosen from what was not

Payment failure and voluntary cancellation have nothing in common except where they land in the report. At consumer price points the involuntary share is large enough to move the headline figure on its own. You can size your own split with the churn rate calculator and, for the revenue-weighted view, the MRR churn calculator.

What good looks like

For a consumer subscription, holding more than half a cohort at month three is a reasonable target and puts you above specialty streaming's 54% (Antenna Specialty SVOD Landscape, 2026). Month-twelve survival above 40% is strong, since Premium streaming as a whole sits near there once its published 60% first-year cancellation rate is inverted (Antenna State of Subscriptions: Adds and Ads, 2026) and specialty sits at 27% (Antenna Specialty SVOD Landscape, 2026). Both matter far less than the direction of your month-three number across successive cohorts, because that's the one you can still change.