Subscription economics: churn, cohorts and the real payback

A subscription programme is a promise about future revenue, and the only honest way to value that promise is a cohort curve. Average lifetime value calculated from an average churn rate will overstate the programme, because churn is not constant. It is highest in the first two cycles and falls steeply after that, so a single average blends two very different populations and produces a number that flatters the plan.

This is how to model a subscription properly, which numbers to watch monthly, and where the levers actually are.

Build the curve before the average

Take every subscriber who started in a given month. Count how many are still active after one cycle, two cycles, three, and so on. Do that for each start month and stack the rows. The result is a retention curve, and the shape of it tells you almost everything.

Two stores can report the same average churn and have entirely different businesses. One loses half its subscribers after the first delivery and keeps the rest for two years. The other loses a tenth each month, steadily. The first has a product-fit problem in the first box and a strong programme after it. The second has a slow leak that compounds. The average hides both.

Read the curve at three points:

• Cycle one to two. The largest drop on almost every programme. This is the trial-versus-commitment moment.
• Cycle three to six. Where a habit either forms or does not.
• Beyond twelve. The tail, which is where most of the profit sits and which almost nobody measures because their reporting only goes back a year.

The four numbers to hold monthly

Active subscribers. The count at month end, with gross additions and cancellations shown separately. A flat total can hide heavy churn matched by heavy acquisition, and those two situations need opposite responses.

Revenue per active subscriber. Total subscription revenue in the month divided by average active subscribers. This drifts as discounts, add-ons and plan changes accumulate. When it falls while subscriber count rises, the programme is growing in the wrong direction.

Contribution per active subscriber. Revenue per subscriber minus cost of goods, packaging, fulfilment, payment fees and the ongoing discount. This is the number the payback period is calculated against, and getting it wrong by a few points changes what you can afford to pay for a subscriber.

Involuntary churn share. The proportion of cancellations caused by a failed payment rather than a decision. This is usually the single largest recoverable segment and the cheapest to fix.

Voluntary and involuntary churn are different problems

Split them in reporting, always. Mixing them makes the programme look worse than it is and points the work in the wrong direction.

Involuntary churn comes from expired cards, insufficient funds, changed card numbers and bank declines. The remedies are mechanical: retry logic that spaces attempts sensibly rather than hammering the same declined card, card account updater services offered by payment providers, a pre-billing notice three days ahead with a link to update payment details, and a dunning email sequence that is polite rather than alarming. A programme with no dunning sequence is losing subscribers who never intended to leave.

Voluntary churn comes from a decision, and the reason matters. Capture it at cancellation with a short list of options: too much product, too expensive, not using it, quality, found an alternative, temporary. Each reason maps to a different intervention, and offering a pause is the single most useful option on the page. A paused subscriber who returns in two months is worth far more than a cancelled one you win back with a discount, and the pause costs nothing.

Do not make cancelling hard. Beyond being required in several jurisdictions, a difficult cancellation converts a neutral departure into a complaint, and the complaint outlives the saved cycle. Subscription churn rate sets out how to define the measure so it stays comparable month to month.

The first cycle is the whole programme

Almost every subscription that works well does the same thing: it treats the first delivery as an onboarding event rather than an order.

That means the post-purchase sequence explains what arrives, when, and what happens next. It sets the expectation about the second charge before it happens, because a surprise second charge is a cancellation and sometimes a chargeback. It gives the subscriber control early, showing where to change frequency, skip, swap products or pause, because a person who knows they can pause is less likely to cancel outright. And it asks a question after the first delivery that reveals fit, so the second box can be adjusted.

The mechanics of that sequence are the same discipline as any post-purchase flow, with one difference: every message should reduce the chance of a surprise. Surprise is what cancels subscriptions.

Get the cadence right, or nothing else matters

The most common cause of cancellation on a consumable product is accumulation. The customer has three unopened units and cancels because stopping is easier than adjusting.

Set the default interval from actual consumption, not from a preference for monthly billing. A skincare product that lasts seven weeks should not default to four. Where usage varies by customer, ask at signup, and make the interval genuinely easy to change afterwards. Then watch for the signal: a subscriber who skips twice in a row is telling you the interval is wrong, and an automatic prompt to lengthen it saves the subscription.

Products whose consumption rate you cannot state confidently are usually better served by a reminder to reorder than by an automatic charge. The design of that alternative is in replenishment flow design.

A worked example, illustrative only

Say a programme charges 40 dollars per cycle at a 55 percent contribution margin, so each cycle contributes 22 dollars. Acquisition costs 60 dollars per subscriber, and the first cycle carries a 20 percent welcome discount, so cycle one contributes about 14 dollars.

Cumulative contribution runs 14, 36, 58, 80 across the first four cycles. Payback lands between cycle three and cycle four. Now apply retention: if 60 percent of subscribers reach cycle three and 45 percent reach cycle four, the expected contribution per acquired subscriber across four cycles is well below 80 dollars, and whether the cohort has repaid its acquisition cost depends entirely on that curve rather than on the per-cycle margin.

Two moves change the answer. Lifting cycle one to two retention by a few points raises every subsequent number, because retention compounds through the curve. Reducing the welcome discount raises cycle one contribution immediately but may lower the number who start. Both are testable, and they pull in opposite directions, which is why they should be tested one at a time.

Every figure above is invented to show the structure. Run it with your own margin, your own acquisition cost and your own curve.

What to test, in order of expected return

1. Dunning and card updating. Recovering failed payments requires no persuasion and no discount.
2. Cadence defaults and the skip experience. Fixes the largest voluntary churn reason on consumables.
3. The first two cycles of communication. Expectation setting before the second charge.
4. A pause offer at cancellation. Converts departures into deferrals.
5. Plan structure. Bigger sizes at a longer interval often suit the customer better than a monthly small size, and reduce fulfilment cost per unit of product.
6. Incentive design. Ongoing discount, free shipping, member-only products or early access all cost different amounts and pull different people. The ongoing discount is the most expensive and the easiest to reach for.

What to report to whoever funds it

One page, monthly: active subscribers with additions and cancellations, contribution per active subscriber, the retention curve for the last twelve start cohorts, involuntary churn share, and cumulative contribution against acquisition cost by cohort. That last chart is the one that answers whether the programme is an asset or a discount scheme.

Subscription revenue also changes how the rest of the store should be measured, because a subscriber acquired this month is revenue booked over the next year. Acquisition targets set against first-order return will systematically underfund a programme that works. How do you increase repeat purchase rate covers the non-subscription side of the same question, and the CAC payback calculator will show how long a given cohort takes to repay at your numbers.

---

Sources

• Shopify Help Center: Subscriptions
• Shopify.dev: Subscriptions purchase options