Which Converts Better: Freemium, Free Trial or Hard Paywall?

The question gets asked as a three-way choice and it is really a two-by-two. Freemium and hard gating are the two models, because they answer the question of whether a person can use the product without paying. A free trial is a modifier you bolt onto either one.

Getting that straight matters because the published benchmarks split on the models, not on the trial. RevenueCat’s State of Subscription Apps 2026, built on more than 115,000 apps and $16 billion in tracked revenue, reports median day-35 download-to-paid conversion at 10.7% for hard paywall apps and 2.1% for freemium. Roughly five to one, on the same traffic.

What do freemium, free trial and hard paywall actually mean?

A hard paywall means no meaningful use of the product before payment or trial signup. A freemium app gives away a usable tier and sells an upgrade. A soft paywall sits inside freemium: it shows the offer, the user dismisses it, and the free tier continues.

A free trial is a payment delay rather than a model. You can run a hard paywall with a seven-day trial, or a freemium tier with a trial on the premium features, and the published data treats those as variants of the two models rather than as a third one. That is why “freemium vs free trial” is a category error that keeps getting repeated in listicles.

The distinction that actually predicts the numbers is whether the product does anything for a non-payer. Everything else follows from it. A hard-gated app has one funnel and every user in it is a prospect. A freemium app has two populations with different economics, and the mistake teams make is reporting them as one.

One more definition worth pinning down, because it causes most of the confusion between published benchmarks. Download-to-paid measures payers as a share of everyone who installed. Paywall-view-to-payment measures them as a share of people who reached a paywall screen. The second number is always larger and the two are not interchangeable, which is covered in more detail in what a good app paywall conversion rate looks like.

What does RevenueCat’s State of Subscription Apps 2026 say?

RevenueCat’s State of Subscription Apps 2026 says hard gating converts roughly five times as many downloads as freemium, and that the money in subscription apps is concentrated in very few of them.

Per the report’s own description of its dataset, it is built on more than 115,000 apps, more than $16 billion in tracked revenue and more than a billion transactions across iOS, Android and web, and it reads mostly on 2025 performance.

Three of its figures bear directly on the model choice, per the report. Median day-35 download-to-paid conversion is 10.7% for hard paywall apps against 2.1% for freemium.

Median revenue per install is $2.32 at day 14 for hard paywall apps against $0.27 for freemium, widening to $3.09 against $0.38 by day 60, so the gap is already visible a fortnight after install rather than only at the end of a long window.

On trials, per the same report, median trial-to-paid runs 25.5% for trials of four days or fewer and 42.5% for trials of 17 to 32 days.

The figures most summaries leave out describe the sample, and they change how the medians above should be read. Per RevenueCat’s own summary of the 2026 report, the median app earns about $72 a month one year after launch and the top quartile earns above $429.

According to the same summary, the median app grew monthly recurring revenue 5.3% year on year while the top decile grew 306% or more and the bottom quarter shrank by about a third, and apps launched before 2020 still take 69% of all subscription revenue.

What the 2026 report measures	Figure
Apps in the dataset	more than 115,000
Revenue tracked	more than $16bn, across more than 1bn transactions
Median monthly revenue, one year after launch	about $72
Top quartile, same point	above $429
Median year-on-year MRR growth	5.3%
Top-decile year-on-year MRR growth	306% or more
Revenue earned by apps launched before 2020	69% of the total
Median day-35 download-to-paid, hard paywall against freemium	10.7% against 2.1%

Every figure in that table is RevenueCat’s, from the State of Subscription Apps 2026 and its accompanying summary.

Put those two halves together and the hard-paywall median stops being a target. It is the middle of a pool where most apps are small and growth sits with a handful at the top, so a gate cannot move an app from the median to the top decile on its own.

Treat the number as a position to locate yourself against. One further caveat is ours rather than the report’s: the dataset is apps that use RevenueCat, which reads the independent end of the market better than it reads the largest publishers.

What is the difference between a hard paywall and a soft paywall?

A hard paywall blocks the product until the user pays or starts a trial. A soft paywall shows the same offer and lets the user dismiss it and carry on with limited access, which makes it a freemium variant rather than a third model. The whole difference is one dismiss button, and it decides who ever sees the offer.

That is also why the two are benchmarked on different denominators, and why quoting one figure at the other produces nonsense. A hard paywall appears to everyone who opens the app, so its conversion is measured against downloads. A soft paywall appears to someone who walked into a gated feature, so its conversion is measured against paywall views, a smaller and much warmer group.

Adapty’s State of In-App Subscriptions 2026, covering $3 billion in revenue across more than 16,000 apps, found soft paywalls converting 4.85% of paywall views to payment against 3.34% for hard paywalls. Quoted alone that reads as a win for soft gating, and it does not contradict RevenueCat’s five-to-one download-to-paid gap, because the two studies count different people.

	Hard paywall	Soft paywall
Can the user dismiss it	no	yes
What the product does before payment	nothing	limited free use continues
Who sees the offer	everyone who opens the app	people who reached a gated feature
Benchmark denominator	downloads	paywall views
Conversion on that denominator	3.34%	4.85%
Which model it belongs to	hard gating	freemium

Conversion figures in that table are Adapty’s, from its State of In-App Subscriptions 2026.

A soft paywall earns its place when the product has something to show in the first session but needs more than one session to prove itself, because it lets the user keep going and asks again later.

What it costs you is the measurement: once the free tier runs, download-to-paid and paywall-view-to-payment diverge and you need both to see what is happening. If you run one, read the free tier as a design problem rather than a default, which is the subject of the freemium section below.

Which model converts the most installs into payers?

Hard paywalls convert the most installs, at a median 10.7% of downloads paid by day 35 against 2.1% for freemium, according to RevenueCat’s 2026 dataset. The top decile of hard paywall apps reaches 38.7%, which is higher than most teams believe is possible from cold paid traffic.

	Hard paywall	Freemium
Free access before paying	none	usable free tier
Download-to-paid, day 35 (median)	10.7%	2.1%
Revenue per install, day 14	$2.32	$0.27
Revenue per install, day 60	$3.09	$0.38
First-year LTV per payer (median)	$49.30	$24.24
Refund rate	5.8%	3.4%

Conversion and revenue-per-install figures per RevenueCat’s State of Subscription Apps 2026; lifetime value and refund figures per Airbridge’s analysis of RevenueCat’s 2025 report, which drew on a smaller sample of more than 75,000 apps.

The revenue gap is wider than the conversion gap, and that is the part worth sitting with. RevenueCat puts revenue per install at day 60 at $3.09 for hard paywall apps against $0.38 for freemium, about eight to one.

Conversion explains five of that and the rest comes from who converts. Airbridge’s analysis of RevenueCat’s 2025 report, covering more than 75,000 apps, puts median first-year lifetime value per payer at $49.30 for hard-gated apps against $24.24 for freemium. A hard paywall payer is worth roughly twice a freemium payer as well as being five times more likely to exist.

Stacked together those two effects produce the roughly elevenfold revenue-per-install difference Airbridge reports, $5.96 against $0.53, which is a wider gap than RevenueCat’s own day-60 figures because it is built from the older sample. For anyone buying installs, that is the number that decides whether a campaign can pay back, because it is the only figure on this page denominated in the same unit as a cost per install.

Two honest caveats. These are medians across a very wide pool of apps, and category does more work than model in plenty of cases. And selection runs through all of it: the teams that choose hard paywalls tend to have products with obvious immediate value, so some of the gap is the kind of product rather than the gate in front of it.

Is the five-times gap an A/B test or a comparison between different apps?

The five-times gap is a comparison between different apps, not an A/B test. Neither of the large datasets on the question splits one app’s traffic between a hard paywall and a freemium tier, and the 2026 write-ups of both report nothing that does.

RevenueCat’s medians describe apps that had already chosen a model, so the difference between the two groups carries every other difference between them, including the kind of product that makes a founder confident enough to charge before anyone has used it.

A second dataset measures the same question from the other end and points the other way. Adapty’s State of In-App Subscriptions 2026, drawn from more than 16,000 apps, puts soft paywalls at 4.85% of paywall views converting to payment against 3.34% for hard paywalls, close to 50% better, as reported in NeoAds’ write-up of the two reports side by side.

Both readings are defensible because they are not the same ratio, and the denominator is doing the work.

	RevenueCat 2026	Adapty 2026
Sample	more than 115,000 apps	more than 16,000 apps
What is counted	downloads that pay within 35 days	paywall views that convert to payment
Hard gating	10.7%	3.34%
Freemium or soft gating	2.1%	4.85%
Which model leads	hard gating, by about five times	soft gating, by about 50%

Sample sizes per each report’s own description of its dataset; conversion figures per RevenueCat’s State of Subscription Apps 2026 and per NeoAds’ reading of Adapty’s 2026 report.

A hard paywall shows its paywall to everyone who opens the app, so its paywall-view denominator is close to its install base. A freemium app shows a paywall only to someone who has already used the product and run into a limit, which is a pre-qualified group that any gate would convert better.

Measured per install, hard gating wins by a lot. Measured per paywall view, it loses, because the freemium paywall is being judged on a warmer audience.

Revenue points the same way as the install-denominated reading. Per NeoAds’ summary of the Adapty report, hard paywalls produce roughly 21% higher lifetime value per subscriber, with a median first-year figure of $24 against $20 for soft paywalls, although another Adapty page gives different medians for the same comparison, so take the direction rather than the dollars.

What survives all of that is narrower than the headline. Across large pools of apps, hard gating is associated with more revenue per install, and the effect is big enough that it is unlikely to be an artefact. What does not survive is the sentence most summaries write, which is that switching to a hard paywall will convert five times as many of your own installs. Neither dataset measures that.

How do you test a hard paywall against freemium in your own app?

Testing a hard paywall against freemium in your own app means randomising new installs into the two experiences and reading each cohort to revenue, and the reason almost nobody does it is that the gate is a product decision rather than a paywall variant. A paywall test swaps a screen.

Moving between models changes what the app does before the screen appears, which usually means shipping and maintaining two onboarding paths.

The paywall tools are built for the smaller question, and their published limits shape what you can design. Adapty’s own comparison against Superwall puts RevenueCat’s experiments at four paywall variants with equal traffic weights and a reporting delay of about a day, advertises unlimited variants with custom weights and live results for itself, and credits Superwall with holdout groups.

That is one vendor describing its competitors, so check the current caps in each product’s own documentation before planning around them. All of it is enough to test where the wall sits and what it says, and none of it tests whether the wall should be there at all.

Four design choices decide whether the result means anything.

Decision	Do this	Because
Randomisation unit	assign on install, before either paywall is seen	assigning at paywall view throws away the users a hard gate loses
Primary metric	revenue per install at day 60	the two models move conversion and value per payer in opposite directions
Net or gross	net of refunds	hard gating carries the higher refund rate priced in the next section
Duration	long enough for trials to resolve	a trial-led cohort has not finished converting while the trial is still running

Most apps cannot run this test, and that is a reasonable place to land. At the conversion rates in either report, detecting a difference in revenue per install needs more new installs than a pre-launch app will see in a quarter.

The workable version is to treat the install-denominated direction as the starting assumption, choose the model that fits the product, and then test the things a paywall tool can actually test: the position of the gate, the price on it, and the length of the trial behind it.

What does a hard paywall cost you?

A hard paywall costs you refunds and your entire non-paying audience. Refunds are the measurable half: Airbridge’s read of RevenueCat’s 2025 report found that hard-gated refund rates run 5.8% against 3.4% for freemium, about 70% higher, which is the predictable result of asking people to pay before they have used anything.

Refunds are a line item you can plan for. Read net revenue rather than gross when you compare models, and expect support volume to rise with it, because a refund request usually arrives as a message rather than as a silent cancellation.

The audience cost is harder to price and easier to underestimate. At the 10.7% median conversion rate reported by RevenueCat, a hard-gated app keeps about one user in nine and the rest are gone with no further relationship.

A freemium app with 2.1% conversion keeps the other 97.9% in some form. Which model you can afford depends on the number on the paywall, which is set out in how to price a subscription app.

Whether that is an asset or an expense depends entirely on the product. For a photo editor it is mostly an expense, since free users cost server time and give nothing back. For anything with a social graph, a marketplace or user-supplied content, the free base is the product.

When is freemium the right choice anyway?

Freemium is right when free users do work that paying users cannot do. Referrals, network effects, a content library the users themselves supply, and any category where the product improves as more people use it all justify carrying a large non-paying base, and none of those jobs shows up in a download-to-paid benchmark.

The test is whether you can name the job. “Free users might convert later” is not a job, it is a hope, and RevenueCat’s medians say you are buying that hope with roughly four fifths of your conversion rate. “Free users invite their team, and teams convert at four times the rate of individuals” is a job, and it is checkable against your own data inside a quarter.

The second case for freemium is a product whose value takes real time to become apparent. Anything built on accumulated history, whether that is a habit tracker, a finance app or a journal, has very little to show a new user in session one. Gating it hard means selling a promise, and a promise converts worse than a demonstration.

Where freemium goes wrong is almost never the decision and almost always the design of the free tier. A free tier generous enough that nobody needs to upgrade produces the 2.1% median in RevenueCat’s study and none of the offsetting benefits. The free tier has to be complete enough to prove the product works and narrow enough that the main use case runs into a wall.

How long should the free trial be?

Longer than most apps set it. RevenueCat’s 2026 study found that trial-to-paid conversion runs 25.5% for trials of four days or fewer and 42.5% for trials of 17 to 32 days, so a three-day trial gives up roughly 17 points of conversion against a month-long one.

The intuition that shorter trials create urgency does not survive the data. What a short trial mostly creates is a decision made before the user has built anything worth keeping, and for products that depend on accumulated history, a four-day trial can be functionally the same as no trial at all.

The cost of a long trial is cash flow and the risk of trial abuse, both real and both manageable. What it buys is a user who has done enough with the product that cancelling has a switching cost, which is why the curve keeps climbing all the way out to a month.

Trial start timing is not something you get to choose in the same way. Most trials begin immediately, and moving a subscription decision later in the lifecycle usually means moving it out of the window where users actually make it. The sequencing question is covered in why app onboarding is so important.

Which model should a new app launch with?

Launch with a hard paywall and a trial unless you can name what free users are for. A gated launch gives you revenue per install within weeks, which is the only figure that tells you whether paid acquisition can work, and it is the easier decision to reverse: opening up a gated product is straightforward, while closing one people already use free costs you goodwill and reviews.

Run the model choice as one decision with the creative, not after it. The ad promises something specific and the paywall has to sell that same thing, and a mismatch between them shows up as high install volume with a dead funnel. Keeping the ad promise and the paywall promise as a single experiment is the part we own end to end in an app launch.

The exception is the product that does not work without a user base. Marketplaces, social products and anything with network effects have to buy liquidity before they can charge for it, and a hard paywall at launch means never reaching the point where the product is worth paying for.

Whatever you launch with, instrument both halves. Track download-to-paid and paywall-view-to-payment separately, segment by acquisition source, and read net revenue after refunds. Those three habits will tell you within a quarter whether the model was the constraint or whether something else was.

The short version

Hard paywalls convert about five times as many installs as freemium, at a median 10.7% against 2.1% by day 35 per RevenueCat’s State of Subscription Apps 2026, and they earn about eight times as much per install by day 60, at $3.09 against $0.38. Per payer, first-year value runs $49.30 against $24.24 on Airbridge’s read of RevenueCat’s 2025 report.

You pay for that with refunds at 5.8% against 3.4% per Airbridge, and with the non-paying audience you no longer have. Freemium earns its lower conversion only when free users refer, supply content or create network effects, and if you cannot name that job the free tier is costing you most of your revenue.

A free trial is a modifier on either model rather than a third option, and length is the lever, per RevenueCat: 42.5% trial-to-paid at 17 to 32 days against 25.5% at four days or fewer. Default to gating with a trial, keep the ad promise and the paywall promise identical, and read net revenue per install as the number that decides whether any of it pays back.