The short version: hard paywalls win on conversion rate, freemium wins on audience size, and trials sit in between, with trial length mattering more than most teams think. The right model depends on one question: how fast can your app prove its value? The faster the proof, the harder your paywall can be.
What the data says
RevenueCat's State of Subscription Apps 2026, drawing on over 115,000 apps and more than $16 billion in tracked revenue, gives the clearest published picture of how the three models perform:
- Hard paywalls convert about 5x better than freemium: a 10.7% vs 2.1% median download-to-paid rate.
- Longer trials convert better. Trials of 17 to 32 days convert at a 42.5% median vs 25.5% for trials under four days.
- Yet short trials are gaining share. The share of apps running trials of four days or less rose year over year, trading conversion rate for faster billing.
- Trial cancellations are front-loaded: 55% of 3-day trial cancellations happen on Day 0, the day the trial starts.
Read those numbers carefully, because the obvious conclusion (everyone should run a hard paywall with a month-long trial) is wrong. Median conversion rate says nothing about volume. A hard paywall converts a much smaller funnel: everyone who will not pay bounces at the door instead of becoming a free user you might convert later, or who might fill your leaderboards, marketplaces, and referral loops.
The decision framework: speed of value proof
Ask how many sessions your app needs before a user would honestly say "this works for me."
- Value proven in the first session (photo editors, scanners, single-purpose utilities): hard paywall or a short trial. The user has seen everything they need; waiting only loses them.
- Value proven over days (fitness plans, learning streaks, budgeting): a trial long enough to contain the proof. If your habit loop takes ten days to feel real, a 3-day trial asks users to pay on faith. This is where the long-trial conversion advantage comes from.
- Value that grows with the network or library (social features, marketplaces, content catalogs): freemium, because free users are an input to the product, not just unconverted revenue.
Notice the framework has no entry for "we chose freemium because asking for money felt rude." That is the most common real reason, and the 5x gap is its price tag.
What your category expects
Users arrive with expectations set by every other app in your category. Directional estimates from a July 2026 snapshot of PaywallPro's public Open Paywall Gallery dataset of 910 top iOS subscription apps show trial-led paywalls ranging from about 77% of Business apps down to about 24% of Entertainment apps. Deviating from your category norm is a legitimate test, but it should be a deliberate one. The full breakdown is in our guide to paywall placement.
The hybrid patterns worth knowing
- Reverse trial. Start every user with full premium access for a set period, then drop them to a free tier. They experience the ceiling before the floor, and the downgrade itself becomes an upgrade prompt. It combines freemium reach with trial urgency.
- Freemium with a trial-led paywall. The free tier exists, but the onboarding paywall still leads with a trial offer. The common pattern behind most "77% trial-led" business apps.
- Hard paywall with a dismiss-to-limited mode. Looks like a hard paywall, converts like one, but a small close affordance drops users into a heavily limited preview instead of losing them entirely.
Price structure is part of the model
The trial question and the price-menu question travel together, and the menu also varies by category far more than teams assume. In the same third-party dataset snapshot, weekly plans appear on about 68% of Utilities paywalls and 52% of Photo & Video paywalls, but on essentially 0% of Finance paywalls and 5% of Sports paywalls. Median observed US prices cluster around $6.99 to $9.99 per week, $12 to $15 per month, and $50 to $80 per year, with Finance topping the annual range at around $119.
The pattern behind the numbers: weekly pricing thrives where the job is bursty or short-horizon (edit these photos, get through this trip) and disappears where the product is a long-term relationship with trust at stake. Weekly plans also pair naturally with hard paywalls and no trial, because the plan itself is the low-commitment entry. If you sell a long-horizon outcome, the conventional annual-with-trial anchor remains conventional because it works. As always, these are directional third-party estimates: treat them as the market's prior, not your answer.
Whatever you choose, make it cheap to change
Monetization models are not one-way doors, but for most teams they behave like one because every paywall change rides an app release. The teams that find the right model are the ones that can test trial lengths, framing, and placement weekly. Setgreet ships paywall and upgrade screens as native flows you can restructure and A/B test without a release, alongside the billing stack you already run. Setgreet charges a flat MAU price, never a percentage of your revenue.
Frequently asked questions
What is a good trial length for a subscription app?
Long enough to contain your value proof. Published medians favor longer trials on conversion (42.5% for 17-32 days vs 25.5% under four days, per RevenueCat), but a long trial on an app whose value is obvious in one session just delays revenue. Match the trial to the proof, then test.
Should I switch from freemium to a hard paywall?
First check whether your free tier feeds the product (network effects, content, referrals) or just houses users who will never pay. If it is the latter, test a hard paywall on a fraction of new installs before committing; the conversion lift is real, but so is the volume drop, and only your numbers can say which wins.
Why do users cancel trials immediately?
Mostly self-protection against forgotten renewals, and it is common: 55% of 3-day trial cancellations happen on the start day. Early cancellation does not mean the user is lost; it means the trial period itself has to convince them to come back and resubscribe.