The paywall
Hard gate versus soft gate, what a converting paywall contains, what Apple's review guidelines require it to say, and the published conversion figures for each choice.
Where the paywall goes
The single largest documented effect in this playbook is placement. Apps with a hard paywall — the offer at the end of onboarding, nothing free behind it — convert Download → paid at day 35: 12.1% (median, hard paywall) RevenueCat 2025 against 2.2% (median, freemium) RevenueCat 2025 for freemium. Five to one. The cost is refunds: 5.8% (hard-paywall apps) RevenueCat 2025 against 3.4% (freemium apps) RevenueCat 2025, because some of the people a hard gate converts had not yet decided.
Timing explains most of it. Trials that start on install day: 82% (share of all trial starts) RevenueCat 2025. Intent peaks at install and decays; a paywall met on day three, in Settings, is an offer to someone who has already decided the free version is enough.
| Metric | Figure | Source |
|---|---|---|
| Download → paid at day 35 | 12.1%median, hard paywall | RevenueCat 2025 |
| Download → paid at day 35 | 2.2%median, freemium | RevenueCat 2025 |
| Refund rate | 5.8%hard-paywall apps | RevenueCat 2025 |
| Refund rate | 3.4%freemium apps | RevenueCat 2025 |
| Trials that start on install day | 82%share of all trial starts | RevenueCat 2025 |
RevenueCat 2025: RevenueCat sells subscription infrastructure, so the sample is apps that chose that tooling — better-resourced than the App Store as a whole. Medians and percentiles are reported; exact app counts per figure are not.
Soft gate: when freemium is right
Freemium wins when the free product is itself the acquisition channel — content people share, a utility people recommend — and the paid layer is depth. Then the paywall is a feature gate: the one thing people came for, behind the offer, shown at the moment they reach for it. Gating something nobody reaches for is freemium's most common failure.
Multi-page versus single-page
A multi-page paywall walks through two or three screens of value before the offer. On Onboarding paywall conversion, multi-page vs single-page: 12.41% vs 9.07% (trials or purchases ÷ paywall views; +37% relative) Superwall 2026, and only 24% Superwall 2026 of onboarding paywalls are built that way. It is the best-documented layout finding available — and it is observational. Apps that build multi-page paywalls are apps with a monetisation team; the gap is partly them.
Superwall 2026: Observational, not randomised: apps that build multi-page paywalls are self-selected and likely better at everything else too. Superwall sells paywall testing. The metric is trials or purchases divided by paywall views at onboarding placements.
Measure it before believing it
Sendpaywall_viewed with placement and a variant property, run both for the same weeks, and compare paywall_viewed → trial_started per variant in a custom funnel. Fifty people per variant before drawing a conclusion.How many plans, and which one is selected
One plan converts worst in every published comparison. Two plans against one: +61% (relative; 15 apps) Superwall 2022; three against two: +44% (relative; 15 apps) Superwall 2022. The sample behind those percentages is small and old, so take the direction, not the digits.
Superwall 2022: Fifteen apps is a small sample for a distribution claim, the page's own headline statistics do not reconcile arithmetically, and it is four years old. Directionally consistent with practitioner experience; do not treat the percentages as precise.
- Two or three plans, one pre-selected. The pre-selected plan is the one you want people on — usually annual — framed per week or per month so its price reads smaller than the monthly plan's.
- Show the monthly price beside the annual one. The comparison is the argument; a lone annual price is a large number with nothing to be smaller than.
- A trial toggle ("Not sure yet? Enable free trial") lets people who would not pay today start a trial instead of leaving. It is the most-copied paywall pattern of the last three years and, like the rest, unmeasured in public.
What a converting paywall contains
Practitioner consensus, across every published teardown, on the elements in order of importance:
- An outcome headline, ideally quoting the user's own onboarding answer. \"Reach your goal weight by March\" outperforms \"Unlock Premium\" in every account — nobody has published a controlled measurement.
- Three to five benefits as checkmarks, each an outcome, none a feature name.
- The plans, as above, with the trial length, the price after it and the charge date written on the paywall — not only on the system sheet. A surprise on the sheet is a cancelled purchase.
- Social proof — a rating, a count, one quote — nearest the button.
- One button, with the action on it (\"Start my 7-day free trial\"), and beneath it the small print App Review requires.
What App Review requires
Apple 2026 §3.1.2: the paywall must state the subscription's length and price, what renews and at what price after any trial, and must carry links to terms of use and privacy policy and a way to restore purchases. A paywall that hides the post-trial price is both a rejection and a refund generator. Apple also forbids implying that the app is unusable without paying when it is not, and any UI that makes the close control hard to find is a review risk — a hard paywall with no close is allowed; a close hidden behind a delay is a complaint.
After the paywall: the abandoned purchase
About Paywalled users who abandon a started purchase: ~20% Superwall 2024. A follow-up offer to exactly those people — a different plan, a trial where there was none — is credited with Revenue attributed to transaction-abandon paywalls: 17% (share, not lift) Superwall 2024 in the one published case study. That figure is a share of revenue, not measured lift, and the study is not a controlled one; the pattern is cheap to try and easy to measure with a custom funnel from purchase_failed to purchase_completed.
Superwall 2024: Not an A/B test: the 'control' and 'variant' are different populations (all installs versus users who abandoned a purchase). The 17% is a share of revenue attributed to the pattern, not measured incremental lift.
Measure it in Broctic
The default subscription funnel already contains onboarding_completed → paywall_viewed → trial_started → subscription_started, benchmarked against the figures on this page. The store side — downloads that start a trial, trials that pay, refund rate — is computed from App Store Connect without any events at all (how verdicts are made).