Subscription Business Model: 6 Types and When Each One Works.

This post assumes you have already decided to run subscriptions. If you have not, start with the subscription business model guide and come back.

Every article on subscription business models hands you a menu of four or five types and closes with "pick the one that fits your brand." Fit is not a feeling. A curated box needs a gross margin most Shopify stores do not have, and prepaid needs a balance sheet that can absorb a refund on goods you have not shipped.

Repeat buyers generate 44% of total store revenue while making up only 21% of customers, according to Rivo's 2026 Shopify retention benchmark. A subscription programme is a bet on that 44%. The merchant who picks the curated box because it is the most fun to imagine puts the bet behind a model his margin cannot fund. He finds out in month nine, with the inventory bought.

The Five Numbers That Decide Your Subscription Model

You already have all five, and none needs a new tool to find.

• Gross margin on the one-time price. Revenue minus COGS on the product you would put on subscription, not your blended store margin. Shopify's product cost field holds it per SKU.
• Average order value. The AOV of the products you would subscribe, not the store average. One $200 gift order flatters a $30 replenishment business badly, and high AOV and high volume are two different businesses.
• Natural reorder interval. The median gap in days between one customer's repeat purchases of the same item. No cadence setting in any app overrides it.
• Subscribable SKU count. Products a customer would plausibly want again on a schedule, not the whole catalogue.
• Churn shape. Where the cancellations cluster, rather than the rate. A percentage with no period attached cannot be acted on, which is the argument behind Joy's churn benchmarks.

Four of those five feed one formula:

Margin floor = 30 points of contribution + the recurring discount + the app and payment cost + the per-cycle handling cost + the COGS variance buffer.

The 30 points is the floor Joy already publishes: below 30% gross margin you struggle to cover acquisition costs. The recurring discount, call it 10 points, is the median subscription discount across all three performance cohorts in Joy's 2026 dataset of 12,969 Shopify stores.

Eightx's 2026 DTC benchmark puts median DTC gross margin at 56.6%, with the 25th to 75th percentile running 45.6 to 63.8%. Beauty and haircare anchor the top of that range, food and personal care the bottom.

These five numbers eliminate rather than rank. A model is in or it is out.

The Six Subscription Business Model Types at a Glance

Read down the margin floor column first. It eliminates faster than the other four combined.

Every floor in that table was computed, not surveyed. Nobody publishes a margin floor by subscription model, because it depends on your own discount and handling cost. Each section below shows its stack.

Two lines in every stack are assumptions rather than data: per-cycle handling and COGS variance. A third, the 0.75 points of app cost, is Joy Subscriptions' rate on Starter. Substitute your own numbers where they do not match.

Replenishment (Subscribe & Save): The Default, and Why It Usually Wins

Replenishment carries the lowest margin floor of the six, so it is the model you run unless a number disqualifies it. COGS is fixed because every shipment holds the same thing, and nobody has to kit anything.

The stack is 30 points of contribution and 10 points of recurring discount. Add 0.75 for the app and roughly a point of per-cycle handling, and it comes to 41.75. The table says 40% because the discount is the one line you can shrink. At a 37% gross margin a 5-point discount still clears: 30 + 5 + 0.75 + 1 = 36.75. That margin funds Subscribe & Save and nothing else on this page.

• Margin floor: around 40%.
• AOV range: $25 and up. Below that, shipping and the discount together outrun the margin.
• Reorder interval: 14 to 90 days, and out past roughly 120.
• Subscribable SKUs: one is enough.
• Churn shape: flat and consumption-linked. Cancellations track running out of need, not disappointment.

The disqualifier is the interval, and Joy's cadence data shows the wall. Across the 2026 dataset, 6,018 stores offer a monthly cadence, 3,099 offer two-monthly and 1,305 offer three-monthly. Past a quarter it collapses: 364 stores at six months, 290 at twelve. The gate shows up in what merchants configure, not in what anyone recommends.

Recurly's July 2026 churn benchmarks put median annual ecommerce churn at 4.25%, split 2.87 voluntary and 1.38 involuntary. Nearly a third of it is a failed card rather than a decision, which matters most here: the replenishment subscriber who still needs the product never chose to leave.

Curated Box: The Highest Margin Floor of the Six

The model merchants want most needs the most margin. A curated box sells a selection you choose, and every part of that promise costs margin.

The stack starts at 30 points of contribution and adds about 25 for the perceived-value gap. A box has to look worth more than its price, and that gap is a discount by another name.

Add 0.75 for the app and about 10 points of kitting, because someone picks, packs and prints an insert for every box, every cycle. That comes to 65.75 before you buffer a point for COGS variance, and a box's COGS varies every cycle by definition. So 65% is the floor, not the target.

• Margin floor: around 65%, the highest of the six.
• AOV range: $35 to $75. Under $35 the kitting labour dominates. Over $75 the recurring charge gets re-examined every cycle.
• Reorder interval: a fixed 30 days. The cadence is the product, not the consumption rate.
• Subscribable SKUs: none of your own. A box needs a rotating sourcing pipeline, not a catalogue.
• Churn shape: front-loaded. Novelty is the product, so boxes two and three carry the decision.

Hold that 65% against the benchmark. Eightx puts the 75th percentile of public DTC gross margin at 63.8%, so a 65% floor sits above three quarters of the brands it measured. Every company it lists at 65% or better sells beauty or haircare. Beauty boxes work because beauty margins clear the floor. Pantry boxes fail because food and beverage, in Eightx's words, rarely exceeds 40% at scale, and no amount of brand work moves a 35% margin to 65%. If your margin sits near the 56.6% DTC median, picking a better box niche will not rescue the model.

A box subscriber re-decides every cycle until the habit sets, which is why the first three cycles are where reducing churn pays back.

Build-a-Box: When Catalogue Size Does the Retention Work

Build-a-box is gated by catalogue depth rather than margin. The customer assembles their own box, so what you sell is choice, and choice runs out. Below roughly 12 subscribable SKUs the interesting combinations are gone by cycle three, and the model's only advantage goes with them.

The stack sits between the other two. Start with 30 points of contribution, 10 points of discount and 0.75 for the app. Add about 5 points of variable pick cost, because every box is a different set of lines. Then about 3 points of COGS variance. That comes to 48.75, and the floor is 50%, because merchants underestimate variable pick cost.

• Margin floor: around 50%.
• AOV range: $45 and up, so that building the box is worth the customer's time.
• Reorder interval: 30 to 60 days.
• Subscribable SKUs: 12 or more. This is the gate.
• Churn shape: front-loaded, then flat. Choice fatigue arrives when the catalogue runs dry around cycle three.

Joy's own data points the same way. Across the 2026 dataset, stores in the bottom half of performers carry an average of 14 products on subscription, the middle 40% carry 19, and the top 10% carry 105. That is an association inside Joy's merchant base rather than an industry law, and the 105 is a consequence of scale as much as a cause of it. Still, stores where choice is the product have a catalogue behind it.

Joy Subscriptions ships both box shapes, fixed-quantity and dynamic, on the Starter plan and above. Fixed bundles and build-a-box are the two mechanics underneath.

Prepaid: The Model Fewer Than 1 in 10 Stores Run — and Better Than 1 in 4 Top Performers Do

Prepaid is a commitment structure, not a price. The customer pays for several cycles up front, so the question is never what you charge. It is who carries the risk, and when the cash arrives.

The adoption gap is the argument for prepaid. In Joy's 2026 dataset, 9.7% of the 7,479 classified Shopify stores offer a prepaid option at all. Among the top 10% of performers, 27% do. Among the bottom 50%, 13% do. Only 1.8% of those stores, 137 of them, lean on prepaid hard enough to be classified as a prepaid-heavy archetype in the same dataset.

Better than one in four top performers run prepaid. Fewer than one in ten stores overall do. That is an association, not proof that prepaid causes performance, and the gap between the top decile and the bottom half is better than two to one.

• Margin floor: around 45%. Prepaid buys its commitment with a deeper discount, so put 13 points in, the deepest median discount of any industry in Joy's 2026 dataset: 30 + 13 + 0.75 + 1 = 44.75. Every extra point of prepaid discount moves the floor with it.
• AOV range: $30 and up per cycle, and roughly $100 and up per prepaid ticket. Below that the cash advantage does not repay the refund exposure.
• Reorder interval: any, because the term is what you are selling. Three to twelve cycles.
• Subscribable SKUs: one is enough.
• Churn shape: a cliff at term end. By construction there is no churn inside the term, so all of it lands at renewal.

Say your cycle is $40 and you sell a six-cycle term at the same 13% discount the floor assumes. The customer pays $208.80 on day one instead of $240 spread across six deliveries, and that cash is in your account immediately.

The obligation stays put. If that subscriber quits after cycle two, four cycles are unshipped, and four sixths of the $208.80, or $139.20, is money you are holding against goods you still owe. Prepaid does not reduce churn. It relocates churn, from a decision made every cycle to one decision at the end of the term, and it makes you the bank in between.

That relocation is worth a deeper discount, but only if you can refund the unshipped balance without breaking cash flow. If you cannot, prepaid is out, however good the adoption numbers look. Old Salt Coffee runs a prepaid annual coffee-of-the-month term, which removes the cancel decision for twelve cycles and concentrates the entire renewal risk into one month.

Prepaid ships on every Joy plan, including Free Forever.

Once you have picked the model, the pricing mechanic is a separate decision, and subscription pricing models for Shopify covers that side of it.

Access and Membership: What It Takes, and When It Is the Wrong Tool

A membership sells a benefit rather than a shipment, and that is the entire gate. With no COGS per cycle there is no margin floor. The test is perk cost per member per cycle, held against the recurring fee.

If free shipping costs you $7 an order and your member places two orders a cycle, the perk costs $14 against a $12 recurring fee, so you are paying members to stay. A free-shipping membership fails this way quietly, because the cost lands in a different line of the P&L from the revenue that justifies it.

• Margin floor: not a margin gate. The gate is perk cost against the fee.
• AOV range: a $5 to $50 recurring fee, with no COGS per cycle.
• Reorder interval: none. There is no shipment, only a billing cadence.
• Subscribable SKUs: irrelevant. Catalogue depth does nothing for this model.
• Churn shape: a step function at the first unused period. The member who skips a cycle cancels the next one.

When memberships are the core model, with content access control, member-only areas and loyalty tiers, a purpose-built membership app is the right choice, and Joy's own Subi comparison says so plainly. Subscription apps bill on a cadence; membership apps gate on entitlement.

Your decision is one sentence long. If you cannot name a recurring benefit that is not a shipment, this model is out rather than deferred.

Usage-Based: Why It Almost Never Fits a Shopify Store

Usage-based billing charges for what the customer consumed, after they consumed it. It is the dominant model in cloud infrastructure and it almost never fits a Shopify store, because it needs a meter.

A meter counts consumption in the background and reports it before the invoice goes out. Once a bag of coffee leaves the warehouse, nothing tells you how fast it is being drunk, and a model that cannot measure consumption cannot bill for it.

• Margin floor: not a margin gate. The gate is metering.
• AOV range: not applicable. Priced per unit, billed in arrears.
• Reorder interval: none. Billing runs per period rather than per delivery.
• Subscribable SKUs: irrelevant. This model needs a meter, not a catalogue.
• Churn shape: hidden in usage decay. The subscriber stays on the books while the revenue quietly leaves.

A store selling API calls, print credits, storage or metered service hours has a meter, because the product lives in software. A store selling anything you put in a box does not, so this model is out at the first question.

Run the Elimination Test on Your Own Numbers

Five steps, each with a single threshold that eliminates rather than scores. Pull your figures from your subscription analytics first, then work down the list in order.

1. Compute gross margin on the one-time price of what you would put on subscription, then read the band against the floors in the table above. Below 40%, only replenishment survives, and only at a discount under 10%. From 40 to 44%, replenishment only, because prepaid, build-a-box and the curated box are all out on margin. From 45 to 49%, replenishment and prepaid. From 50 to 64%, replenishment, prepaid and build-a-box, and the curated box is out. At 65% and above, every model is still in play. Membership and usage-based are not margin-gated, so they survive this step and go at step 5 and at the metering question instead.
2. Measure the natural reorder interval in days, as the median gap between one customer's repeat purchases. Past 120 days, replenishment and prepaid are both out. The product is not consumed fast enough for either.
3. Count subscribable SKUs, meaning products a customer could actually want on a recurring basis, not the whole catalogue. Below 12, build-a-box is out.
4. Divide per-cycle handling cost by AOV. Handling means picking, kitting, packing and inserts: everything you pay every cycle that is not the goods. Above 15%, the curated box and build-a-box are both out. The labour has eaten the model.
5. Answer two yes/no questions. Can you refund an unshipped prepaid balance without breaking cash flow? Can you name a recurring benefit that is not a shipment, in one sentence? A no on the first eliminates prepaid. A no on the second eliminates membership.

Usage-based is already out for almost every reader, at the metering question in the section above, so it does not need a step of its own.

The same test, run on four store shapes, gives four different answers:

Replenishment survives in three of the four, which is what the lowest floor of the six buys you. What separates the examples is what clears alongside it, and three of the four clear more than one model. The test narrows the field. It does not hand you a single answer.

The beauty brand is the awkward case, because three models survive and nothing in the five steps ranks them. Pick on the interval. At a 90-day reorder gap there is barely anything to replenish, and build-a-box would hand a discovery customer the job of choosing, which is the job she is paying the brand to do. The curated box sets its own cadence, so a slow consumption rate stops being a problem at all. The pet retailer inverts every one of those: a 30-day gap and 60 SKUs, so choice is the mechanic and build-a-box is the box to build.

Read example D twice. The ceramics studio should run a waitlist and a restock alert, not a subscription. A test is allowed to return nothing, and a framework that always produces an answer is not screening anything. It is recommending.

Whichever model survives, the number that tells you the call was right is lifetime value per subscriber, read after three cycles rather than after three days.

Pick the Model Your Margin Can Fund

Your subscription model is determined, not chosen, and the arithmetic that determines it takes about twenty minutes.

The wrong turn is always the curated box, and the execution is rarely what goes wrong. A 48% gross margin was never going to carry a 65% floor, however good the sourcing calendar. Run the five steps before you build anything, then set up the model that survived.

Joy Subscriptions' Free Forever plan covers up to 50 active subscriptions at no cost and no transaction fee. That is enough to launch a replenishment or prepaid programme. You find out whether the numbers hold in your store rather than on a page. The box models, fixed and dynamic, sit on Starter at $49/month plus 0.75%. Compare the plans against whichever model survived.

Frequently Asked Questions

What are examples of subscription business models in ecommerce?

Replenishment is a coffee roaster shipping the same bag every 30 days. A curated box is a beauty brand sending a different selection each cycle. Build-a-box is a pet retailer letting the customer pick six items from sixty. Prepaid is a six-month tea term paid up front. Membership is a free-shipping club with no shipment attached. Usage-based is a print shop billing per credit used.

For the definition itself, the subscription business model guide and the glossary entry cover that ground.

What is the difference between a membership and a subscription business model?

A subscription delivers a thing on a cadence. A membership delivers access, with no shipment attached. That difference decides the economics: a subscription carries COGS every cycle and needs a margin floor, while a membership carries a perk cost per member and needs the recurring fee to clear it.

Which subscription business model is the most profitable?

Most profitable is the wrong question. Replenishment carries the lowest margin floor, around 40%, so it is viable for the widest range of stores. The curated box carries the highest floor, around 65%, and the highest ceiling for the stores that clear it. Ask instead which model your margin can fund.

How much gross margin do I need to run a subscription box?

Around 65% on the one-time price. The stack is 30 points of contribution, about 25 for the perceived-value gap, 0.75 for the app and about 10 for kitting and packaging. That floor sits above the 63.8% Eightx reports as the 75th percentile of public DTC gross margin, and level with the beauty and haircare brands that anchor the top of its table.

Which subscription model works for a store with fewer than 10 products?

Replenishment or prepaid. Build-a-box is out below 12 subscribable SKUs, because choice runs out by cycle three. A curated box does not need your catalogue at all, since its items are sourced rather than picked from your own products, but it still needs the 65% margin floor. One hero SKU is enough for replenishment.