Subscription Retention Benchmarks: Data From 8 DTC Programs

Subscription retention is the share of a signup cohort still subscribed after a given number of months. Across four of the DTC subscription programs DTC Pages measures, covering 438,961 subscribers, 51% were still active at month three. A fifth program in the same set was at 14%.

One benchmark curve hides a 3.6x spread#

Subscription benchmark pages publish the same thing: one retention curve, one number per month, presented as what a DTC subscription looks like.

We have dashboard access to eight subscription programs across four platforms, and five of them had cohort curves clean enough to compare. Four of those five stayed within 2x of each other at every month, tightest at signup and widest by month seven.

The fifth sits 3.6x below the average of those four by month three. It has more than 30,000 subscribers and its curve declines smoothly from one month to the next, so it is neither a small sample nor a reporting glitch. It also sells a consumable, and a single benchmark curve would average it in with the other four.

Key numbers at a glance#

Every figure below comes from a cohort curve or a twelve month window we pulled from the platform ourselves. None of it is survey data.

8 DTC subscription programs measured by DTC Pages, 2026. Every average here is across programs, never across subscribers.

The two month-three retention rows are the main finding: the same metric at the same month, 37 points apart.

How we collected this#

This comes from eight subscription programs DTC Pages has dashboard access to, across four platforms: Recharge, Skio, Stay AI and Loop.

Every figure uses the same window, 23 September 2025 to 22 September 2026, with one exception flagged where it appears. We set the window by hand on every platform, because these dashboards open on a thirty day view by default and thirty days answers a different question than twelve months.

Cohort curves cover 471,175 subscribers across those five programs. Month 0 is the month a subscriber signs up. The platforms label it differently: Recharge calls it Month 0 and Skio calls the same month 1st Month.

Programs are grouped and reported by category. No store is named, and no figure is published at a level that identifies one.

Every figure is an average across programs, never an average across subscribers. One program in this set holds 67% of the subscribers, so an average across subscribers would mostly describe that one program.

Inside the band of four, the average and the median never differ by more than one and a half points, and the only month where they round more than a point apart is month zero, at 91% against 93%. We report the average because it is what most dashboards show.

Four programs share a band. The fifth does not.#

Four programs, from three categories on three platforms, stay within 2x of each other at every month. The fifth drops out of that band at the first renewal and never comes back.

Base: the four programs in the band, 438,961 subscribers. The right hand column is a single program, more than 30,000 subscribers of its own. Measured by DTC Pages, 2026.

The two best curves both come from supplements programs, run by separate companies on separate platforms. That is the closest thing to a replication in this data.

The fifth also sells a consumable, so whether the product runs out does not separate it from the rest. And since every category here except supplements has a single program, this data cannot tell you whether category predicts retention in general.

Program size does not separate them either. Two programs of similar size, both between 30,000 and 50,000 subscribers, sit 41 points apart at month three.

The one cliff we could trace#

The one sharp drop we could trace to a cause is in the pet program, inside the band. It comes from how much product one shipment carries against how often the subscription bills.

Its starter shipment carries about two months of product, and a monthly billing interval is among the options on offer. A subscriber on that interval accumulates a month of unused product every cycle. By the second renewal they have paid twice, received four months of product and used two, so a third charge lands while a full shipment sits unopened. That program's curve loses 7 points at the first renewal and then 27 at the second, and the step appears in every cohort we have.

We have not checked the same arithmetic in the other four programs, including the outlier, so we do not know how much of the spread it explains. It is still the first thing we would check in any program with a sharp drop at one specific renewal.

Half life: one number per program#

Half life is the month at which half of a signup cohort is gone. You find the two months that straddle 50% and interpolate between them. One of the supplements programs is at 52% in month four and 48% in month five, so its half life is 4.5 months. The outlier is at 87% in month zero and 47% in month one, so its half life is 0.9.

Band average 3.3 months, range 2.4 to 4.5. Half lives calculated by DTC Pages from cohort curves covering 471,175 subscribers.

You can calculate yours from any cohort report in a few minutes, and it gives you one number instead of eight. That helps because single months rank programs inconsistently: the pet program is the best of the four in the band at month one and the worst of them by month four, and the two supplements programs trade places between month two and month three.

Unlike a churn rate, half life does not change meaning when your growth rate changes. It follows a closed group, so it does not matter how many people signed up afterwards.

One caution. The program with the best half life also bills least often: its billing interval grows with bundle size, up to about three months on the largest plan. A subscriber who is not charged for three months has fewer moments to reconsider, so part of that half life comes from how the plan is built. Compare half lives only between programs that bill on similar intervals.

Every store loses people at a different month#

The month-over-month drops show when each program loses people, which is more useful for deciding what to fix than the curve itself.

Month over month drops across 5 programs measured by DTC Pages, 2026.

The outlier loses 40 points between month 0 and month 1, so almost half of its signups never take a second delivery. Most of its gap with the band opens at that first renewal. We have not traced the cause.

Pet barely loses anyone at that first step, 7 points, then drops 27 at the next one. It shows up in every cohort we have for that program, so it is not one odd month. This is the cliff explained by the shipment arithmetic above.

One supplements program spreads its losses evenly, 14 then 12 then 11 points, with no month that stands out. Without a moment to target, it is the hardest of these patterns to act on.

The other supplements program ties superfoods for the largest first step drop inside the band, 17 points, and still finishes with the best curve in the set, so a large first-month loss did not decide where that cohort ended up.

Black Friday cohorts do not churn faster#

The usual assumption is that subscribers acquired during Black Friday signed up for the discount and leave once it runs out. We compared the November and December 2025 cohorts, the promotional window rather than the single weekend, against every other cohort in the same window, across the five programs with monthly cohort data.

One of the five has a Black Friday cohort of 317 subscribers, below the 500 floor we apply everywhere else in this post, so it sits outside the aggregate. Across the four that clear it, three retained their Black Friday cohort better than the rest of the year at month three and one slightly worse. The average difference is +1.8 points and the median is +1.7. That is a small edge rather than a penalty, and not a gap you would plan around in either direction.

The excluded cohort is the one clearly negative result in the set, 5.6 points worse at month three, and at that size a few dozen people move it several points. One program let us look further back. We pulled the same November and December window for 2023 and 2024, the only figures in this post taken from outside the common twelve month window. Of those three seasons, two clear the same floor, and in both of them the gap at month three stayed within a point.

For three of the four this is a single season, so it does not prove that Black Friday cohorts retain well. It is enough to say they do not reliably retain worse.

Churn is almost never a payment problem#

Dunning, card updaters and payment recovery get most of the attention in subscription advice. They are worth setting up, but in our data they address a small share of churn.

12 months of churn events from 4 DTC subscription programs measured by DTC Pages, 2026.

Voluntary means the subscriber went into their account and cancelled. Passive means nobody cancelled: the card expired or the bank declined the charge, and after several retries the platform ended the subscription on its own.

Across four programs where the platform separates the two, voluntary cancellations are 84% to 94% of all churn, average 89%. Failed payments are everything else. Nine in ten of the people you lose chose to go.

We use twelve months of churn events for this split rather than a thirty day window, which gives counts an order of magnitude larger. Across four very different programs, the voluntary share stays between 84% and 94%.

Recharge publishes vertical medians for Health and Wellness that point the same way: 7.3% active monthly churn against 0.5% passive.

This split covers four programs, not eight, because two of the platforms in our set do not separate voluntary from passive churn.

Subscription AOV is lower, until subscription becomes the default#

In four of the five programs where we have both numbers, subscription orders are worth less than one time orders. Across those four the gap runs from 6% to 24.5%, average 15%.

Subscription against one time AOV in 5 DTC subscription programs measured by DTC Pages, 2026.

In those four, a forecast that moves customers to subscription and holds average order value constant will overstate revenue per order.

The fifth runs the other way: its subscription orders are worth 48.5% more than its one time orders. Around 98% of its orders are subscriptions, so the few one time orders left are mostly small add-ons rather than another way to buy the main product. The comparison stops meaning much there, which is why the 15% average covers four programs and not five.

Subscription's share of revenue also says little about whether the program is working. In our set, the vape program sits at 72% of revenue and pet at 71%, and they retain 14% and 47% at month three.

Two metrics that look comparable and are not#

Two numbers in a subscription dashboard look comparable across stores and are not.

Growing fast inflates your churn rate. One program in our set reads 32.1% monthly churn on a thirty day window, and 13.6% when the rate is derived from twelve months of churn events.

The gap is composition. New subscribers churn faster than old ones, which you can see in any cohort curve. That program added close to five times its active base in new subscribers over the year, so at any given moment almost everyone in the base is new. A thirty day churn rate measures how fast you are growing as much as it measures your subscribers. In a program with an older base, the same two methods read 10.8% and 7.6%.

The twelve month churn rate in your dashboard is not a churn rate. Recharge reports churned subscribers over active base. The numerator counts the whole year and the denominator is today's snapshot, so once your base turns over more than once a year the figure passes 100% by construction. We measured 148.5%, 165.4% and 480.8% across three programs. At 480% it is a turnover ratio, not a rate, even though the dashboard shows it in the headline position labelled as churn.

How to read your own curve#

Calculate your half life, not your churn rate. Pull any cohort report and find the month where the cohort drops below 50%. One number, no arguments about denominators, and it does not move when your growth rate moves.

Find your worst month and attack that month. Churn is not one problem. A store losing people at the first renewal and a store losing them at the fourth are dealing with different problems, so the month tells you where to look.

Divide the days of product in your first shipment by your billing interval. If the answer is greater than one, subscribers receive product faster than they use it, and the ratio tells you which renewal to watch.

Check your voluntary split before you invest in dunning. If 89% of your churn is people choosing to leave, a better card updater is working on about one churn in ten.

Do not compare your thirty day churn to anyone growing at a different rate. Compare half lives or cohort curves instead, between programs that bill on similar intervals.

The limits of this data#

One program holds 67% of the subscribers in the curves, which is why every figure here is an average across programs.

Four programs is still a thin band, and we publish the full range for that reason. Add the fifth program and the average at month three drops from 51% to 43%, which is why this post never gives one figure across all five.

We dropped the tail of any curve whose cohorts fall under 500 subscribers. Below that a handful of people move the percentage several points.

We dropped two programs entirely for sample size, one under 200 subscribers and one around 1,100.

We dropped one platform from the curve comparison. It forces 100% at month zero and appears to count subscribers with an order that month rather than subscribers still subscribed. That is a different measurement under the same label.

The voluntary against passive split comes from four programs, not eight.

Figures are reported by category rather than by store, and nothing here is published at a level that identifies a client.

The Black Friday comparison rests on a single season for three of the four programs in the aggregate. The three season check on the fourth is the only place in this post where we step outside the common twelve month window, and we say so where it appears.

Frequently Asked Questions#

What is a good subscription retention rate?#

For a consumable DTC product, 45% to 56% of a cohort still active at month three is the range we see. The DTC Pages average across four such programs is 51%, covering 438,961 subscribers. That band is not universal: a fifth program in the same set, which also sells a consumable, sits at 14% at the same month.

What is a good subscription churn rate?#

There is no single figure, because the churn rate your dashboard shows depends on how fast you are growing. A store adding subscribers quickly reads a much higher rate than a store with the same retention and a stable base. Across the eight programs DTC Pages measures, one read 32.1% on a thirty day window and 13.6% derived from twelve months of churn events for the same period. Compare half lives or cohort curves instead.

How do I calculate subscription retention?#

Take everyone who subscribed in the same month and treat them as a closed group. Count how many of that exact group are still subscribed one month later, two months later, and so on, and divide by the size of the group. Do not add new subscribers into the count. That is what makes it a cohort curve rather than a churn rate. It is also how DTC Pages calculated every retention figure in this post, across 471,175 subscribers in the five programs with usable cohort curves.

Why is my subscription churn rate over 100%?#

Because it is not a churn rate. Most dashboards report churned subscribers over active base, so the numerator counts the whole period and the denominator is today's snapshot. Once your base turns over more than once a year the figure passes 100% by construction. DTC Pages measured 148.5%, 165.4% and 480.8% across three of the eight programs in this set. Read it as a turnover ratio.

What percentage of subscription churn is involuntary?#

Between 6% and 16%, average 11%, across the four DTC Pages programs where the platform separates the two. The other 89% is people who went into their account and cancelled, rather than a card that failed. Recharge's published Health and Wellness medians point the same way: 7.3% active monthly churn against 0.5% passive.

Is subscription AOV lower than one time AOV?#

Usually. In four of the five programs where DTC Pages has both numbers, subscription orders are worth 6% to 24.5% less, average 15%. The exception is a program where around 98% of orders are subscriptions, and there subscription AOV is 48.5% higher, because the one time orders left over are add-ons rather than a real alternative way to buy.

What is subscription half life?#

The month at which half of a signup cohort has cancelled. You find the two months that straddle 50% and interpolate between them. Across the five programs DTC Pages measured with full cohort curves, it runs from 2.4 to 4.5 months in the band of four and 0.9 months for the outlier. It is the most comparable single figure we have, because it follows a closed group and does not move when your growth rate moves, as long as the programs bill on similar intervals.

Why does my churn rate go up when I grow?#

Because a thirty day churn rate measures the composition of your base, not the behaviour of your subscribers. New subscribers churn fastest, so a store adding subscribers quickly always has a base made mostly of new subscribers. One of the programs DTC Pages measures reads 32.1% on a thirty day window and 13.6% derived from twelve months, and it added close to five times its active base in new subscribers over the year.

What share of orders should include a subscription?#

Recharge publishes a Health and Wellness vertical median of 24% of checkout orders containing a subscription. Be careful comparing that figure across platforms: some report the share of orders containing a subscription and others report subscription revenue over total revenue, and the first always reads higher.

How many months should a subscription cohort be tracked?#

Long enough to pass your half life, and not so long that the cohort gets too small to read. DTC Pages publishes seven months and cuts the tail of any curve whose cohorts fall under 500 subscribers, because below that a handful of people move the percentage several points.

Why is my subscription retention so far below the benchmark?#

Check first where the loss happens. The outlier in the DTC Pages data sits at 14% at month three against a band average of 51%, and most of that gap opens at the first renewal. A drop concentrated at one renewal points to something specific to fix. In another program, DTC Pages traced a sharp drop at the second renewal to shipments that outlast the billing interval: subscribers pile up product they have not used and cancel before the next charge. Divide the days of product in your first shipment by your billing interval. If the result is above one, check whether your biggest drop lands on the renewal where unused product would start to pile up.

Does Recharge retention compare to Skio retention?#

The curves compare once you line up the definitions, and the labels do not. Recharge calls the signup month Month 0 and Skio calls the same thing 1st Month. Skio also counts subscriptions by default and subscribers behind a toggle, and one person can hold more than one subscription. Check which of the two you are reading before comparing anything. DTC Pages put the five programs with cohort curves on the same convention before comparing them.

How often should I benchmark my subscription program?#

Quarterly is enough for the curve, because a cohort needs months to develop. Check your voluntary against passive split more often than that, since it moves when your payment setup changes.