Sell a subscription when your app delivers value again and again, like content, a tool people use every week, or an AI feature with ongoing costs.
Sell one-time in-app purchases when the value is finite: a permanent unlock (non-consumable) or something that gets used up, like credits or coins (consumable).
Many apps now mix the two, but the tests in our database show that adding one-time options next to a subscription can cost you subscription revenue, so test the mix rather than assume it helps.
Below we define each in-app purchase type, walk through the revenue and retention trade-offs, look at hybrid credit-plus-subscription models, and close with what real experiments from our app paywall tests show.
In-app purchase types: consumable, non-consumable and auto-renewing
An in-app purchase (IAP) is anything a user buys inside an app through the store's billing system rather than when downloading the app.
Apple defines four types, and Google Play groups the same ideas into one-time products and subscriptions.
| Type | What it is | Typical examples | Bought how often |
|---|---|---|---|
| Consumable | Used up, then bought again | Coins, AI credits, extra lives, boosts | Repeatedly |
| Non-consumable | Bought once, kept forever | Pro unlock, ad removal, a level pack | Once |
| Auto-renewable subscription | Recurring charge until the user cancels | Weekly, monthly or yearly premium plans | Every billing period |
| Non-renewing subscription (Apple) | Access for a fixed period, no auto-renewal | A season pass, a 3-month course | User renews by hand |
Consumable vs non-consumable in-app purchases
The difference is whether the item survives. A consumable is spent and can be bought again; a non-consumable stays tied to the user's account and can only be bought once, according to Google's billing docs.
That has practical effects. Apple requires a restore mechanism for restorable purchases, so a non-consumable must come back when someone reinstalls or switches devices.
Apple's guideline 3.1.1 also says credits or in-game currencies bought through IAP may not expire.
If you sell AI credits, you can't quietly expire unused balances on iOS.
Family Sharing is another split: Apple lets auto-renewable subscriptions and non-consumables be shared with up to five family members. Consumables can't be shared.
Apple in-app purchase rules in short
Under guideline 3.1.1, if you unlock features or content inside an iOS app, you must use in-app purchase.
The big exception is the United States storefront, where apps may now include buttons and external links to other ways to pay.
That opened up web checkout tests, which we cover below and on our web-to-app page.
Revenue and retention trade-offs
The model you pick changes three things: how much you earn per buyer, how predictable that revenue is, and how much the store keeps.
Store fees favor long-lived subscriptions
On Apple, you get 70% of each subscription payment during a subscriber's first year, rising to 85% after they accumulate one year of paid service.
Free trial days don't count toward that year.
Developers in the App Store Small Business Program, for those under $1 million in prior-year proceeds, get 85% on everything from day one.
Google Play changed its fees in 2026. From June 30, 2026 in the US, UK and EEA, auto-renewing subscriptions carry a 10% service fee plus a 5% billing fee when you use Google Play Billing.
One-time purchases have tiered rates that depend on earnings and on when the user first installed the app; check Google's fee page for your case.
What each model is good and bad at
| Model | Strengths | Weaknesses | Fits |
|---|---|---|---|
| Subscription | Recurring, predictable revenue; higher lifetime value when retention holds | Churn; users resist another recurring bill; you must keep shipping value | Content, fitness, productivity, AI tools with running costs |
| Non-consumable unlock | Simple, honest offer; no churn to manage | Revenue stops after the sale; hard to fund ongoing costs | Utilities, offline tools, ad removal |
| Consumables | Revenue scales with heavy users; low commitment for buyers | Spiky, less predictable revenue; can feel like nickel-and-diming | Games, AI generation, pay-per-use features |
Subscriptions only win if people stay. If most subscribers churn in month one, a one-time unlock may earn as much with less friction.
Model it with our subscription LTV calculator and compare against subscription churn benchmarks before you commit.
Also think about how much you give away for free.
Moving between free and paid tiers changes results as much as the purchase type does.
Opal, a screen time app, replaced its hard paywall with a generous free tier and saw pay penetration fall from 20% to 9% while daily users passed one million.
Opal still judged it a long-term win; it was a before-and-after change, not a split test.
Our freemium guide and freemium limits page go deeper.
Hybrid models: credits plus a subscription
A hybrid model sells a subscription and one-time purchases side by side.
The common version in AI apps: a subscription includes a monthly allowance, and credit packs top it up.
Games often pair a battle pass or VIP subscription with coin packs.
The appeal is clear. Credits catch users who won't subscribe and let heavy users pay for heavy use.
The risk is that credit packs give people a cheaper way out of the subscription, and more options on a paywall can mean slower decisions.
When hiding credits helped
Mojo, a video editing app, tested its paywall with and without AI credit packs.
Showing subscriptions only produced +14% new subscription revenue, with no significant rise in paywall drop-off.
The result was clear within a week, per a Superwall case study (vendor-reported).
Note the metric: new subscription revenue, so it doesn't capture any credit revenue that was lost.
When adding a cheaper option helped
The opposite move can also work when the alternative isn't a cheap purchase.
Ryn VPN added a "Continue with Ads" option giving 30 free minutes next to three paid tiers.
Revenue rose 158%, conversion went from 0.08% to 1.2% and ARPPU rose 98%.
This is vendor-reported and compares 28 days after launch with the same period a year earlier, so treat it as directional.
The pattern across both: a second option helps when it doesn't undercut the subscription's value.
Ads are a cost to the user's time; a credit pack can look like a better deal than the plan.
If you add credits, price them so that a regular user is better off subscribing.
Our plans and subscription pricing pages cover how to lay out options.
One-time payments inside a subscription
You can also borrow the low commitment of a one-time purchase without leaving the subscription model.
Apple supports pay-up-front introductory offers, where a subscriber pays once for a set period before the standard renewal price.
In pooled data from 10 health and fitness apps, Vahe Baghdasaryan's team found a $1.99 paid 30-day trial raised D90 ARPU from $2.23 to $2.89 (+30%) and conversion to paying from 10.2% to 12.7%, versus a monthly plan with no trial.
Ahead App saw +7% conversion to trial when it offered a 30-day trial for a small fee as a second prompt.
What the tests show
Few public tests compare a pure one-time model against a subscription in the same app.
But the tests we have answer the questions around that choice: how much to give away, where people pay, and how to package the purchase.
Paid models tend to beat generous free ones on revenue
V1 Sports, a golf swing analysis app, replaced freemium with paid subscriptions and free trials and reported +90% revenue within 12 months, along with a wave of negative reviews.
A subscription app advised by Phil Carter that switched from a hard paywall to freemium saw more than 50% lower subscriber conversion and reverted within weeks.
Both are before-and-after comparisons. See hard vs soft paywalls for the full set.
Where the purchase happens matters
Store IAP converts better than web checkout, but web checkout keeps more of each payment.
RevenueCat ran a test on Dipsea, an audio stories app it owns: sending every subscribe button to the web instead of Apple IAP dropped initial conversion from 27.0% to 18.1%, and net proceeds per user went from about $2.09 to $1.96 at a 30% store fee.
Offering IAP plus a discounted web option also lowered conversion to paying, from 6.3% to 5.9%.
Other teams found the opposite on proceeds. Zumba routed its paywall button to a Stripe web checkout and saw immediate conversions fall 25% but LTV rise 17%.
Superwall's pooled early tests across US iOS apps showed initial conversion down 11.9% and net proceeds up about 19.85% (vendor-reported).
Superwall also estimated that apps on the 15% Small Business rate would lose about 1.3% of proceeds by moving to web, so the math depends on your fee tier.
Superwall
RevenueCatHow to choose
- Match the model to how value arrives. Ongoing value suggests a subscription; a single job done suggests a non-consumable; variable, metered use suggests consumables or a hybrid.
- Check your costs. If each use costs you money (AI inference, content licensing), a one-time unlock can lose money on your best users.
- Start simple. Launch with one model, then test adding a second option. Mojo's result is a warning that more options can cut subscription revenue.
- Measure the right metric. Track revenue per install or ARPU over 30 to 90 days, not just paywall conversion. Several tests above lost on conversion and won on proceeds.
- Run it as a split test. Many results here are before-and-after or vendor-reported. Your own A/B test, set up with our app A/B testing guide, is the only result that applies to your app.