Subscription doesn't create loyalty. It changes when you find out you didn't have any.
A one-off customer who doesn't come back simply doesn't come back, quietly. A subscriber who cancels tells you the date. That's genuinely more useful information — but it's information, not retention, and the difference is where most subscription programmes go wrong.
Churn isn't a metric in this business model. It's the business model. Get it wrong and you've built infrastructure and paid acquisition costs to earn the same revenue you'd have had anyway.
The three models have completely different economics
They get listed as equivalent options. They aren't remotely.
| Model | What it sells | Why it holds or doesn't |
|---|---|---|
| Replenishment | Removal of a chore | Solves a recurring problem; strongest retention |
| Curation | Novelty and discovery | Novelty decays predictably; weakest retention |
| Access / membership | Ongoing benefit | Holds only if the benefit is used regularly |
Replenishment wins because the customer was going to reorder anyway and you removed a step. Nothing has to stay exciting.
Curation is the model that produces the dramatic churn figures in published research — discovery boxes historically lose a large proportion of subscribers within the first six months, because surprise is not a durable need. That doesn't make curation unviable, but it does mean you're running a business where continuous novelty is the product, and that's a content operation rather than a logistics one.
The question to answer before building anything Is the customer subscribing to avoid a task, or to feel something? The first is durable and boring. The second decays, and you'll be replacing the audience continuously.
Replenishment only works if the cadence is right
The detail that decides whether the model delivers, and most brands guess it.
Set the interval too short and the customer accumulates stock, which produces the single most common cancellation reason. Set it too long and they run out and reorder elsewhere, defeating the purpose entirely.
Neither is a marketing problem — it's a measurement one. The interval should come from your own reorder data: how long, on average, between purchases for non-subscribers of the same product, segmented by pack size. That number exists in your order history and almost nobody looks at it before launching.
Three practical additions:
- Let people change the interval easily, and make it obvious. Someone adjusting from four weeks to six has stayed; someone who couldn't find the setting has cancelled.
- Send a pre-shipment notice with a skip option. It reads as courteous and it prevents the unwanted delivery that triggers cancellation.
- Watch for skip patterns. Two consecutive skips means the interval is wrong, and that's a prompt to ask rather than a problem to ignore.
The delivery experience matters more here than in one-off retail too, because a subscriber experiences it repeatedly — the date-certainty argument in delivery promises compounds across a subscription rather than applying once.
Subscribe-and-save is usually a discount on revenue you already had
The most common way this gets built, and frequently the least examined.
Offering a standing 15% off for subscribing converts a portion of customers who were already reordering into customers who reorder at 15% less. Subscription numbers go up, which reads as growth. Margin goes down, which doesn't get attributed to the programme.
The test is incrementality: are subscribers buying more often or in greater quantity than comparable non-subscribers, or the same amount more cheaply? That's answerable from your own data and it's the number that decides whether the programme is working — the segmentation logic in calculating lifetime value properly applies directly, since a subscriber cohort and a repeat-buyer cohort need comparing separately rather than averaging.
Alternatives worth testing before defaulting to a discount. Non-price benefits — free shipping, early access, a bonus item, flexible sizing — cost less margin and are harder to compare against competitors. Volume rather than rate, where the saving comes from a larger pack. And convenience alone, which works better than most brands expect for genuinely tedious reorders.
If you do discount, structure it as part of the offer rather than as a permanent price cut on the same product — the reasoning in packaging what you sell, where changing composition beats changing the number.
Omnisend
Combines email SMS and push notifications for a complete e-commerce marketing stack
Best for: E-commerce Automation
Involuntary churn is the invisible chunk
Subscribers you lose to failed payments rather than to any decision, and it's frequently a large share of total churn.
Expired cards, insufficient funds, bank declines, address mismatches after a move. These customers never cancelled and never complained. They simply stopped being charged, and your reporting shows them as churn indistinguishable from people who chose to leave.
Two details that matter. Retry timing should avoid the days immediately after a failure and target dates when funds are more likely present. And the failure email should lead with the practical consequence — the delivery that won't arrive — rather than with a billing message, which reads as administrative and gets ignored.
Pause beats cancel, if it's honest
The highest-return retention feature available, with a significant caveat.
A large share of cancellations aren't dissatisfaction. They're stock. Someone with three unopened boxes wants to stop receiving deliveries, not to stop being a customer. Offering pause — with a clear end date and an easy resume — converts a meaningful proportion of those into a gap rather than a loss.
The caveat is that pause has to be a genuine alternative, not an obstruction. Regulators have specifically criticised retention flows where a default pause setting, extra screens, mandatory surveys or a prominent "keep subscription" button make cancelling harder than signing up was. Offering pause is good practice; using it as friction is now the thing enforcement actions describe.
The honest version: present pause and cancel side by side, with equal prominence, and let the customer choose. It still works, and it doesn't create exposure.
What's actually true about cancellation rules in 2026
Worth being precise here, because a lot of current guidance states the opposite of the legal position.
The FTC finalised amendments to its Negative Option Rule — the "click-to-cancel" rule — in October 2024. In July 2025 the Eighth Circuit vacated it on procedural grounds. The rule is not in effect. Several widely-shared guides published in 2026 describe it as fully enforced, which is incorrect.
What is true:
- The FTC continues to bring subscription cases under Section 5 of the FTC Act and under ROSCA, and has done so actively — recent actions have described cancellation flows requiring many screens and many separate actions.
- In March 2026 the FTC launched an advance notice of proposed rulemaking to revive the rule, alongside public statements committing to the area.
- Around 30 US states have their own automatic renewal laws, several stricter than the vacated federal rule — California, for instance, requires a pre-conversion reminder before a free trial converts and annual reminders for auto-renewing subscriptions.
The practical conclusion is unchanged, which is why the correction matters more for accuracy than for behaviour: make cancellation as easy as signing up, because state law, general consumer protection powers and an incoming federal rule all point the same way. What changes is that you can't rely on "the federal rule was struck down" as a position — the exposure moved rather than disappeared, which is the same pattern as in tightening ad disclosure rules.
Practitioner's summary rather than legal advice; this is US-specific, positions vary by state and jurisdiction, and it's a moving area — check primary sources and take counsel.
Which products actually suit this
Because the model is being recommended far more broadly than it fits.
Good fit: consumable, predictable consumption, low reorder consideration, and a genuine chore in reordering. Coffee, supplements, pet food, razor blades, contact lenses, cleaning products, nappies.
Poor fit: anything where choice is part of the pleasure, anything with unpredictable consumption, anything where the customer wants to see it before committing, and anything with a high return rate — since a subscription that generates returns is compounding the cost problem described in how retailers are rewriting return policy.
Worth testing rather than assuming: categories where the product is consumable but the choice matters, such as skincare or speciality food. These frequently work with a curated-replenishment hybrid — the customer picks, and the schedule is automatic.
The honest position is that subscription suits a minority of catalogues, and a brand pushing it across everything is applying a mechanic rather than solving a problem. Where it doesn't fit, ordinary repeat-purchase work does more — the flows in post-purchase sequences and the mechanics in increasing average order value.
Where subscribers are actually lost
Three moments carry most of the churn, and knowing which is yours changes what you fix.
The first renewal. The largest single cliff in almost every programme, and it's where the sign-up promise meets reality. If your acquisition offer was a deep first-box discount, this is where the real price arrives and a large share leave — which is why heavily discounted acquisition into subscription frequently produces worse economics than no acquisition at all.
The second or third delivery. Where a cadence mismatch surfaces as accumulated stock.
The payment event. Cards expire on a schedule you can predict, which makes this the most preventable of the three.
Report cohort retention by joining month rather than an overall churn rate, because the shape tells you which cliff you have. A single blended churn figure describes none of them, for the same reason a blended lifetime value describes nobody.
The sign-up flow deserves the same scrutiny as checkout
Brief, but it's where a lot of potential subscribers are lost.
Subscription sign-up adds decisions to a purchase — frequency, quantity, start date, commitment length — and every one is a chance to abandon. Defaults matter enormously: a sensible pre-selected interval based on your reorder data converts far better than an empty dropdown.
Two things worth checking. Whether the subscription option is legible at the point of decision rather than a radio button people miss, and whether the total commitment is clear — hidden or ambiguous terms are precisely what the enforcement actions described above are about. The general discipline is the same as in checkout optimisation and form design: fewer decisions, sensible defaults, nothing hidden.
Where the subscription interface itself is the constraint — no self-service interval change, no pause, no easy card update — that's a development job that will return more than any retention campaign, because every one of those gaps converts a manageable situation into a cancellation.
What to measure
- Cohort retention by joining month, not a blended churn rate.
- Voluntary versus involuntary churn, separated.
- Renewal-by-renewal survival — first, second, third — to locate the cliff.
- Incrementality against comparable non-subscribers, not subscriber revenue in isolation.
- Skip and interval-change rates, as the early warning before cancellation.
- Contribution per subscriber after discount, shipping and returns — the only figure that says whether this is worth doing.
Reviews are a useful secondary signal here too, since subscribers write about the experience of the programme rather than the product — the collection mechanics in collecting and using reviews apply, and cadence complaints surface there before they surface as churn. And for anything with a trial or introductory period, the conversion economics in free trials versus demos transfer more directly than most retailers expect.
The short version
Subscription doesn't create loyalty; it tells you the date you lost it. The model choice largely determines churn before any retention work — replenishment removes a chore and holds, curation sells novelty and novelty decays. Get the cadence from your own reorder data rather than guessing, and make interval changes and skips easy, because those are retention features disguised as settings. Test whether subscribe-and-save subscribers are genuinely incremental or just cheaper versions of customers you already had. Separate voluntary from involuntary churn, because failed payments are a large silent share and the cheapest thing you'll ever recover. Offer pause honestly, side by side with cancel, rather than as friction. And be accurate about the rules: the FTC click-to-cancel rule was vacated in July 2025, the FTC is reviving it, around 30 states have their own laws, and the practical answer is unchanged.
The programmes that work are boring ones attached to genuine chores. The ones that struggle are selling a feeling on a monthly schedule — and feelings don't renew.
Subscriber numbers growing and margin quietly shrinking?
We separate the churn, test the incrementality, and fix the cadence before you spend another rupee on acquisition.
Audit my subscription programme →Frequently asked questions
Which subscription model has the best retention?
Replenishment, by a wide margin, because it solves a recurring practical problem rather than selling an experience. A customer subscribing to something they were going to reorder anyway is removing a chore. Curation models sell novelty, and novelty decays predictably — which is why discovery boxes historically show much higher cancellation within the first six months. Access and membership models sit between the two, depending on whether the benefit is used regularly.
Is the FTC click-to-cancel rule currently in force?
No, and a lot of current content states otherwise. The FTC's 2024 amendments to the Negative Option Rule were vacated by the Eighth Circuit in July 2025 on procedural grounds, so the rule itself is not in effect. The FTC continues to bring subscription cases under Section 5 of the FTC Act and ROSCA, launched an advance notice of proposed rulemaking in March 2026 to revive the rule, and around 30 states have their own automatic renewal laws, several stricter than the vacated federal one.
Why is subscribe-and-save often a bad deal for the retailer?
Because a large share of the customers taking it would have reordered anyway. Applying a standing discount converts existing full-price revenue into discounted revenue while the reported subscription numbers grow, which looks like growth and is margin erosion. The test is whether subscribers are genuinely incremental — buying more often or in greater quantity than comparable non-subscribers — rather than simply cheaper versions of customers you already had.
What is involuntary churn and how big is it?
It is subscribers lost to failed payments rather than to any decision — expired cards, insufficient funds, bank declines, address mismatches. It is frequently a substantial share of total churn and it is almost invisible in reporting, because those customers never cancelled and never complained. A retry sequence with sensible timing, card-updater services and a clear pre-expiry prompt recovers a meaningful portion at essentially no acquisition cost.
Should you offer a pause option instead of cancellation?
Offering pause is worth doing, because a large share of cancellations are about accumulated stock rather than dissatisfaction — someone with three unopened boxes wants to stop receiving them, not to stop being a customer. What matters is that pause is offered as a genuine alternative rather than used to obstruct cancellation. Regulators have specifically criticised retention flows where a default pause setting or extra screens make cancelling harder than signing up was.
