Retention is won in the first 90 days of a subscription
Subscribers in their first three months churn at 12% per month; subscribers past the one-year mark churn at 3.2%, per Churnkey's 2025 benchmarks. That is nearly a 4x gap, and it means your retention curve is mostly decided before month four. The levers that move it are unglamorous: the first-delivery experience, expectation setting at checkout, an early check-in, and surviving the first real payment.
Save flows, winback sequences, and offer tuning all matter, and this blog spends most of its time on them. But they operate at the end of the subscriber lifecycle. The cheapest churn to prevent is the churn that never builds momentum, and the data says that churn lives in the first 90 days.
The tenure curve
Churn is not evenly distributed across a subscriber's life. It is front-loaded, heavily, and then decays. Churnkey published the cleanest public version of this curve in its November 2025 voluntary churn benchmarks, combining Stripe's 2024 dataset (200 million subscriptions) with Churnkey's own base of 25 million subscriptions:
| Subscriber tenure | Monthly churn rate |
|---|---|
| Under 3 months | 12.0% |
| 3-6 months | 7.4% |
| 9-12 months | 4.9% |
| 12+ months | 3.2% |
Run the compounding: at 12% monthly, a cohort loses roughly a third of itself in the first quarter. The same cohort at 12+ month rates would lose under a tenth. Churnkey's own framing of the action item is blunt: delaying your churn strategy "by as little as 3 months means you lose the opportunity to retain 5% of the customers."
Usual caveat, stated plainly: this data leans SaaS and Stripe. WooCommerce-specific tenure curves are not published anywhere we know of, and the early ChurnStop install cohort is still too small to add a number responsibly. But the shape - steep early, flat late - shows up in every dataset that reports by tenure, and nothing about a WooCommerce store exempts it. For where your overall rates should sit, see the benchmarks post; this post is only about the shape.
Why early churn concentrates
New subscribers churn more because nothing is holding them yet. Three specific mechanisms, each with its own fix later in this post.
The expectation gap. A new subscriber holds a mental picture formed by your marketing. The first delivery or first billing event tests it. Any mismatch - a smaller box than the photos implied, a charge that arrived sooner than expected, a "monthly" subscription that turned out to be synchronized to the 1st - resolves in month one, and it usually resolves as a cancellation.
No habit, no switching cost. A subscriber of two years has routines built around you. A subscriber of two weeks has a line item on a card statement. Usage habits are the retention asset, and they take weeks to form. The cancel-reason data agrees on what kills new subscribers: infrequent usage was 30.6% of cancellations in Churnkey's survey data (2 million responses), and Recurly's 2026 State of Subscriptions found 52% of consumers had canceled at least one subscription in the past year due to lack of use.
The first renewal is the first real payment test. The signup payment happens with the customer present, watching. The first renewal happens alone, weeks later, against a card balance you cannot see. Insufficient funds accounted for 40.56% of the 5.4 million payment failures in Churnkey's involuntary-churn dataset - a failure mode that has nothing to do with wanting the product.
Lever 1: the first-delivery experience
For physical subscriptions, the window between checkout and the first delivery is the highest-anxiety stretch of the entire relationship. The subscriber has paid, received nothing, and is deciding whether this was a good idea.
Three practical moves. Ship the first order fast, even if that means the first box is not on your normal fulfillment cadence - the renewal cycle can synchronize later. Send tracking proactively; silence between charge and doorstep is where remorse grows. And make the first unboxing explain the subscription: when the next box comes, what will be in it, how to skip or pause. A first delivery that answers "what did I sign up for" removes the expectation gap before it becomes a cancel reason.
For digital products the equivalent is time-to-first-value: the gap between paying and the first moment the thing visibly works. Whatever your product's version of that moment is, instrument it, because the subscriber who never reaches it is the "infrequent usage" cancellation of week six.
Lever 2: expectation setting at checkout
The cheapest retention work happens before the subscription exists: say exactly what will be charged, when, and how to leave. Every surprise you prevent at checkout is a first-cycle cancellation you never see.
Concretely, on a WooCommerce store: show the renewal date and amount on the product page and at checkout, not just in the terms. If you synchronize renewals, make sure a signup three days before the aligned date does not pay twice in a week - that is a settings problem with a settings fix, covered in our Subscriptions settings audit. And make cancellation visibly easy. Counterintuitive but consistent: subscribers who know they can leave in one click extend more trust in month one, and trust in month one is what the tenure curve rewards. The compliance version of that argument lives at click-to-cancel.
If you run trials, expectation setting is most of the game. Recurly's 2024 State of Subscriptions put the median trial-to-paid conversion at 50%: half of trials, at the median, never become revenue. A trial that clearly states when it ends and what happens next converts a decision made calmly; a trial that ends by surprise converts a chargeback.
Lever 3: the early check-in
Somewhere between day 7 and day 21, send one human-sounding message with one question: is this working the way you expected? Not a feature tour, not a review request, not a coupon. One question.
The mechanics matter less than the timing. At day 7-21 the subscriber has enough experience to have an opinion and not enough investment to churn silently rather than answer. Replies split into three useful buckets: confirmation (ignore), fixable problems (fix them - this is the cheapest save you will ever make, weeks before any cancel flow), and expectation mismatches you cannot fix (better to learn now, and occasionally better to refund now, than to take two more payments and a chargeback). The one-question rule applies to this email for the same reason it applies to cancel surveys: every added question costs responses.
The check-in also plants the pause option early. Recurly's 2026 report measured pause usage up 337% year over year and found 3 out of 4 subscribers who pause eventually return. A new subscriber who learns in week two that pausing exists has an alternative to the cancel button in week six.
The involuntary side of the first 90 days
Early churn is not all volitional, and the involuntary slice is the most fixable part. A new subscriber's card has never been charged on a recurring schedule by you before; the first few renewals are where mismatched expiry dates, prepaid cards, and thin balances surface.
Two moves. First, turn on the WooCommerce Subscriptions failed-payment retry system - it ships disabled, and the numbers on recovery are strong enough that we gave it the top slot in the settings audit rather than repeating them here. Second, watch your first-renewal failure rate specifically. If renewal one fails at a much higher rate than renewal five, your problem is not dunning, it is what happens at signup: cards that were never validated for recurring use, trials that collected no payment method at all.
One accounting note. If you run $0 trials with no payment method on file, your first-renewal "failure" rate is really a conversion rate, and it belongs with your funnel numbers, not your dunning numbers. Mixing the two makes both unreadable: the trial cohort will swamp the genuine decline data and convince you that you have a payments crisis when you have a trial-design decision.
What to measure
Track the first 90 days as its own funnel, not as a blended monthly rate. A blended churn number averages your new-cohort problem into your loyal base and hides it. Four metrics, monthly:
- 90-day cohort retention. Of subscribers who started in month X, what share is active at day 90? This is the headline number; everything above exists to move it.
- Second-renewal rate. The share of new subscribers who successfully pay renewal number one. It is the single sharpest early-warning metric because it combines the expectation gap and the payment test in one number.
- Early cancel reasons vs late cancel reasons. Split your reason data at 90 days of tenure. Early cancels skewing "not what I expected" is a checkout problem. Early cancels skewing "not using it" is an onboarding problem. Late cancels skewing "too expensive" is a different post.
- Time to first value. First delivery shipped, first login, first download - whatever fits. Report the median and watch the tail: the subscribers in the slowest quartile are your likeliest month-two cancels.
One reporting habit makes all four numbers sharper: grade them by cohort start month, not calendar month. A March acquisition push read on a calendar basis looks like an April retention problem; read by cohort, it looks like what it is - a batch of buyers with different intent than your usual traffic. Cohort columns cost nothing in a spreadsheet and prevent the most common misread of early-tenure data.
A save flow still belongs on the cancel button from day one - early-tenure cancels enter it too. But if your 90-day cohort retention is weak, the save flow is catching people a better first month would never have sent there.
The 90-day playbook
Condensed to a checklist:
- Before signup: renewal date and amount visible at checkout; cancellation visibly easy; trial end date explicit.
- Day 0: first order ships fast; tracking sent proactively.
- First delivery / first login: explains the cadence, the next charge, and how to skip or pause.
- Day 7-21: one-question check-in; fix what it surfaces.
- Renewal 1: retries on, failure rate watched separately.
- Every month: 90-day cohort retention and second-renewal rate reviewed next to your blended churn rate.
None of this is clever. The tenure curve says it does not need to be: the subscribers you keep past month three mostly stay for years. The work is making month three happen. If you adopt only one line of the playbook, take the second-renewal rate - it is the single number this whole post compresses into.