Subscription Retention Benchmarks 2026: Where Your Numbers Should Be
Retention benchmarks get quoted the way pub statistics do: confidently, without a source, and stripped of every condition that made them true. A founder reads somewhere that "good" twelve-month retention is 40%, checks their own dashboard, sees 28%, and panics. Whether that panic is justified depends entirely on what the 40% was measuring, and almost nobody asks.
We spend a lot of time inside subscription analytics for DTC brands on Shopify, and the same mismatch repeats in nearly every audit. The benchmark was cohort-based; theirs is blended. The benchmark excluded trial subscribers; theirs includes them. The benchmark came from coffee; they sell dresses. Same metric name, three different measurements.
So here are the 2026 numbers, with the context that makes them usable.
twelve-month retention composite for consumables subscriptions
the single most dangerous month in a subscriber's first year
cancellations that are typically involuntary, not a decision at all
Three questions to ask before trusting any benchmark
First: blended or cohort? A blended retention rate divides active subscribers by some rolling window of everyone who ever signed up. While you are growing fast, new sign-ups flood the denominator and flatter the number; when growth slows, the same metric collapses even though customer behaviour has not changed. A cohort curve takes one month's sign-ups and tracks what fraction remain active at month one, three, six and twelve. It is the only version of retention worth benchmarking, and it is the version every figure in this post assumes.
Second: trial-inclusive or paid-only? A heavily discounted first box attracts people who never intended to stay. Include them in your cohort and your early retention looks catastrophic next to a benchmark measured on full-price subscribers. Neither population is wrong to measure; comparing one against the other is.
Third: which category? Replenishment products like coffee, supplements and pet food get reordered because they run out. Curated boxes survive on novelty and discovery. Fashion competes with the entire high street every single month. A "good" number in one category is a crisis in another, and category-blind averages tell you almost nothing.
The benchmarking pre-flight checklist
- Cohort-based, not blended. Track one acquisition month over time. If your tooling cannot produce this view, fix that before anything else.
- Paid subscribers only. Measure trial and intro-offer cohorts separately, against their own history.
- Voluntary and involuntary churn separated. More on this below; it changes the diagnosis completely.
- Category-matched comparison. Consumables against consumables, boxes against boxes. Never against an "industry average".
- Consistent month boundaries. Decide whether "month three" means 90 days or three billing cycles, and apply it everywhere.
Five conditions your data needs to meet before any comparison against the table below means anything.
The 2026 benchmark table
The figures below are industry composites: a blend of published industry data and the cohort curves we see across the audits we run. They assume cohort-based measurement on paid subscriptions, with all churn included. Treat the band edges as zones, not pass-fail lines.
Subscribers still active, by category and cohort age
Consumables
Coffee, supplements, pet food
Curated boxes
Discovery and lifestyle boxes
Fashion & apparel
Clothing, accessories, styling
Composite retention bands for 2026. Accent cells are the commonly cited anchors; the remaining cells are our interpolations and will vary more by brand.
The shapes matter more than the cells. Curated boxes start strongest because the unboxing novelty does real work in the first quarter, then decay steeper than anything else once the surprise wears thin. Consumables hold the flattest curve because the product runs out and reordering becomes habit rather than choice. Fashion sits lowest throughout, and that is fine: a fashion brand at 28% twelve-month retention is performing well, while a supplements brand at the same number has a fixable problem.
Ignore any benchmark quoted as a single monthly churn percentage with no category and no timeframe. A flat "5% monthly churn" figure compounds into wildly different annual outcomes depending on where in the curve that 5% sits, and it usually describes a blended rate anyway.
The month-one to month-three cliff
Cancellation does not spread itself evenly across the year. Across the consumables cohorts we audit, well over half of first-year cancellations typically land inside the first ninety days, and the distribution within those ninety days is lopsided in a way most brands never look at.
Month two is the danger zone.
Share of first-year cancellations by cohort month (composite pattern)
Mismatch and trial behaviour: the product was not what they expected, or they never planned to stay.
The danger zone: the intro discount has lapsed, the first delivery is not used up, and the habit has not formed.
Overstock spillover: subscribers still working through earlier deliveries finally pull the trigger.
Lifestyle drift and rigidity: needs change, and inflexible plans cannot bend with them.
The long tail, where involuntary churn becomes the dominant share of losses.
A composite distribution from the consumables cohorts we audit. Your exact percentages will differ; the lopsided shape rarely does.
Why month two? Because the second charge is the first one that feels like a decision. The first order was an experiment, often discounted, with the excitement of something new attached. The second billing arrives at full price, frequently before the first delivery is finished, and asks the customer to confirm a habit they have not actually built yet. If the timing is wrong or the value is unproven, this is where they leave.
"The second charge is the first one that feels like a decision. Design your entire early experience around that moment."
The practical consequence: if your cohort curve drops hardest between billing one and billing two, no amount of month-six win-back budget will fix it. The intervention has to happen in the first thirty days, before the decision point arrives.
Separate involuntary churn before you compare anything
In the accounts we audit, a third or more of churn is typically involuntary: failed payments, expired cards, processor declines, exhausted retry schedules. The subscriber never decided to leave. They simply stopped being charged, and most of them did not notice for weeks.
This matters enormously for benchmarking. Suppose your twelve-month retention sits at 33% against the consumables band of 35-45%, and a third of your losses were payment failures. Your voluntary retention may already be comfortably inside the band. The customers who chose to stay are staying; your payment recovery is what is leaking. Chasing that gap with offers, onboarding redesigns and survey-driven product changes would be solving a problem you do not have.
That is not a retention problem; it is a payments problem.
So split the two before you put your curve next to the table. Most subscription platforms tag cancellation source if you ask the data nicely, and we have written up how to claw back involuntary churn in detail. If your reporting cannot make the distinction at all, start with our guide to building a subscription analytics dashboard; cohort curves and churn-source splits are the first two views it should produce.
What moves each part of the curve
Once you know which segment of your curve is underperforming, the lever almost picks itself. The expensive mistake is applying the wrong lever to the wrong segment: discounting at month six to fix a month-two cliff, or rebuilding onboarding when the leak is actually card declines spread evenly across the year.
Onboarding
Owns the cliff. Usage prompts, delivery-timing checks and a deliberate touchpoint before the second charge are what carry a subscriber past the decision point. This is where early-curve gaps get closed.
Flexibility
Owns the mid-curve. Skip, pause, swap and frequency changes let a subscription bend instead of break when a customer's needs drift. Mid-curve sag is almost always a rigidity problem wearing a price-objection costume.
Dunning
Owns the whole curve. Smart retry schedules, card updaters and failed-payment messaging recover subscribers who never wanted to leave. The cheapest retention you will ever buy, at every cohort age.
Three levers, three territories. Diagnose which segment of your curve is below band before choosing where to spend.
If the early curve is your weak point, we have a full playbook on the first 30 days of subscriber onboarding, because in our experience that window decides more of the twelve-month number than everything that follows it combined.
A worked example: reading your curve against the table
A speciality coffee brand came to us convinced they had an onboarding problem, because that is what the last article they read told them to fix. Their cohort read-out told a different story.
Cohort read-out — coffee brand vs consumables band
Diagnosis: the early curve is healthy, so onboarding is not the problem. The sag begins around months four and five, which points squarely at the mid-curve lever: flexibility.
Reading one cohort against the composite band. Where the curve first leaves the band tells you which lever to reach for.
The cause was mundane. Delivery frequency was locked to monthly, with no skip and no pause. Subscribers drinking less than a bag a month quietly overstocked, and around month four or five they cancelled, mostly citing "too much product". The fix was not a discount or a redesigned welcome flow: it was adding six-week and eight-week frequencies plus one-tap skip. The following two cohorts tracked roughly eight points higher at month six.
This is the entire value of benchmarking done properly. The table did not tell the brand they were "bad at retention"; it told them where the curve left the band, and the where pointed to the why. Once the curve moves, feed the new numbers into your customer lifetime value model, because an eight-point shift at month six changes what you can afford to pay for acquisition.
Benchmarks are a diagnostic, not a scoreboard
The point of the table is not to feel good or bad about your number. It is to locate which third of your curve is leaking, so the budget goes to the matching lever instead of the loudest one. Most brands we meet are spending on the wrong segment, and the misallocation costs more than the churn itself.
These bands will not sit still, either. Skip, pause and swap are rapidly becoming table stakes rather than differentiators, and as they do, the composite bands will drift upward; a 2027 table will likely judge today's "in band" curve more harshly. The brands measuring cohorts properly will see that drift in their own data months before any published benchmark catches up.
Start with the pre-flight checklist, split voluntary from involuntary, and put one clean cohort next to the table. In our experience, that single exercise tells most subscription operators more about where their next quarter of retention work should go than any amount of dashboard-watching has so far.