Install-to-trial conversion is the share of installs that go on to start a free trial. In RevenueCat's 2026 dataset the median for apps that use trials runs from 4.0% in Media and Entertainment to 9.1% in Business, with most non-gaming categories between 5% and 6%, and from 7.1% in North America down to between 3.0% and 3.7% in the lower converting regions. Adapty's 2026 dataset reports a global average of 10.9% and 14.5% for North America, using a different method. Because the large majority of trials start on the day of install, this number is effectively a verdict on the first session: what the ad promised, what onboarding delivered, and what the first paywall asked for.
That last point is why install-to-trial deserves more attention from paid acquisition teams than it gets. It is the first conversion after the install, it resolves within hours rather than weeks, and it is the stage where an ad and a product meet for the first time. When it is weak, the instinct is to blame the ads or to blame the paywall, and the instinct is wrong about half the time in each direction. What follows is the 2026 benchmark data with its methods stated, an explanation of why the two leading reports disagree by a factor of two on the same market, and a diagnostic for working out which side of the install your problem actually sits on.
What install-to-trial measures, and what it does not
Install-to-trial divides free trial starts by installs. The denominator is installs, which is the one thing that separates it from the rates it gets confused with.
- Install-to-trial, which RevenueCat calls download-to-trial, is the share of installs that start a free trial. RevenueCat counts trials started within 30 days of the download date, and calculates the metric only for apps that use a trial-based acquisition strategy.
- Trial-to-paid is the share of trial starts that become a paid subscription. Its denominator is trial starts, and we cover it in detail in the 2026 trial-to-paid benchmarks.
- Download-to-paid is the share of installs that produce at least one paid subscription, which RevenueCat measures within 35 days of install. It is the product of the two rates above, plus any direct purchases that skip the trial.
Two practical cautions before comparing your own figure with anything. First, check what your stack counts as an install. A store download, a first open and a first SDK initialisation are three different numbers, and some installs are never opened at all, so the same app can report noticeably different install-to-trial rates depending on which tool is doing the dividing. Second, the rate only means something for apps that put a trial in front of most users. An app that sells mostly through direct purchase will show a low install-to-trial rate because it is not trying to generate trials, which is why RevenueCat excludes non-trial apps from the metric.
How this stage fits into the whole funnel, from install through to renewal, is set out in our subscription app marketing strategy guide.
Install-to-trial benchmarks by category, 2026
These are medians across apps in RevenueCat's State of Subscription Apps 2026, which covers more than 115,000 apps, over $16 billion in revenue and more than a billion transactions. All are D30 download-to-trial rates. Only the figures RevenueCat states in the report text are included, and the top quartile threshold is shown where it is published.
| Category | Median install-to-trial (D30) | Top quartile |
|---|---|---|
| Business | 9.1% | Above 16.2% |
| Health & Fitness | 6.9% | Not published |
| Education | 6.5% | Not published |
| Utilities | 6.5% | Not published |
| Gaming | 4.4% | Not published |
| Travel | 4.1% | Not published |
| Media & Entertainment | 4.0% | Not published |
Business leads at more than twice Gaming's rate, and RevenueCat notes that the best Health and Fitness apps convert more than 23% of installs into trials, which is more than three times the category median. That spread inside a single category is the more useful fact. It says the ceiling is set by the app, not by the category.
The comparison worth making is with the other end of the trial. Travel sits near the bottom of this table at 4.1%, and in the same report it has the highest trial-to-paid median of any category at 43.5%. Health and Fitness is near the top of both. We would not read a rule into that, because the report cannot tell us why, but it is a reminder that a low rate at one stage can sit beside a high rate at the next. A category whose users only start a trial when they have a concrete need will filter harder at the trial gate and convert more of the trials it does get. The number that pays for media is the product of the two, never either one alone. For the cost side of the same vertical, the health and fitness CPI benchmarks are the companion read.
Install-to-trial benchmarks by market
| Region | Median install-to-trial (D30) | Top quartile |
|---|---|---|
| North America | 7.1% | Above 15.0% |
| Asia-Pacific | 5.7% | Above 13.5% |
| Western Europe | 5.0% | Not published |
| India and South East Asia, Latin America, Middle East and Africa, rest of world | 3.0% to 3.7% | Not published |
RevenueCat groups the last four regions together and gives a median range of 3.0% to 3.7% across them, so individual figures for India and South East Asia or Latin America are not published. North America's median is roughly double most of the rest of the world.
For paid acquisition this compounds with the next stage, and that is the part most geo expansion plans underestimate. The trial-to-paid medians in the same report are 34.2% for North America and 15.2% for India and South East Asia. Multiply each region's two rates together as a rough, indicative calculation, and a North American install produces a paying subscriber at around 2.4%, against somewhere between 0.5% and 0.6% for an install from the lower converting regions. That is not a precise figure, because two medians multiplied together do not give the median of the product, but the order of magnitude holds: before any difference in price, a North American install is worth four to five times as much to a typical trial-based app. A cheaper market is only cheaper if its CPI is lower by more than that. We make the same argument about platforms in iOS versus Android CPI.
Install-to-trial by price point and by AI apps
RevenueCat publishes two further segmentations that are worth knowing, because both run against intuition.
| Segment | Median install-to-trial | Top quartile |
|---|---|---|
| High-priced apps | 8.9% | Above 16.5% |
| Mid-priced apps | 5.4% | Above 11.6% |
| Low-priced apps | 4.4% | Above 10.3% |
| AI apps | 8.5% | Above 15% |
| Non-AI apps | 5.6% | Above 12% |
Price tier here is defined relative to the rest of the panel: whether an app's average price point is below average, average or above average. Higher-priced apps start trials at roughly twice the rate of low-priced ones, and the top 10% of high-priced apps reach 27.0%. It would be easy to read that as permission to raise prices. It is not. Apps that can sustain a high price tend to be the ones delivering a clear, specific result, and a clear, specific result is also what makes a trial worth starting. The data cannot separate the price from the kind of app that charges it.
AI apps start trials at a median of 8.5% against 5.6% for everything else. One small note on that second figure, because it travels: a 5.6% number circulates as RevenueCat's global download-to-trial median. The report does not state a global median anywhere. 5.6% is the median for non-AI apps, which is a different population.
Why RevenueCat and Adapty disagree by a factor of two
Adapty's State of In-App Subscriptions 2026 covers more than 16,000 apps and $3 billion in subscription revenue. It reports North American install-to-trial at 14.5%, against 7.6% to 10.2% in other regions. RevenueCat's North American median is 7.1%. Both are credible datasets, and neither is wrong. Four things separate them.
An average of per-app rates against a median
Adapty states its method plainly: install-to-trial is calculated for each app as trials divided by installs, then the per-app rates are averaged across the segment. RevenueCat publishes the median app. On a distribution this skewed, the difference is large. RevenueCat's own figures put the top quartile at roughly twice the median in every segment above, which is the shape where a handful of very high converting apps drag an average well above the typical app. This is probably the biggest single reason for the gap, although neither report publishes enough to say how much of it is explained.
Different panels, different stores
Adapty says most of its data comes from the Apple App Store, with Google Play included where the report compares stores, and that the analysis covers 2025. RevenueCat's panel of more than 115,000 apps includes both stores. Each is a sample of one vendor's customer base rather than of the market, and the two customer bases are not the same shape.
Which apps are counted
RevenueCat says explicitly that it calculates download-to-trial only for apps using a trial-based acquisition strategy. Adapty's page does not say whether apps that offer no trial are excluded from the denominator.
The window, which matters least
RevenueCat counts trials within 30 days of download and Adapty does not publish a window. Normally that would be the first suspect. Here it is the last, because so few trials start after the first few days that a 7 day window and a 30 day window would produce almost the same number.
One more thing is worth knowing if you use Adapty's figures. Adapty currently publishes two global install-to-trial averages. Its summary report page gives 10.9%. The interactive report's default all-categories, global view shows 11.2%, and Adapty's own Health and Fitness benchmark article quotes 11.2% as the global figure. The difference is small and neither page explains it. Quote whichever you use with its page, and do not treat the other as a correction.
The practical rule is the one we set out for trial-to-paid: pick one source for a funnel model and stay inside it. An Adapty install-to-trial multiplied by a RevenueCat trial-to-paid produces a number that describes nobody's app.
Install-to-trial is decided in the first session
The single most important fact about this metric is when it happens. RevenueCat's 2026 data shows the share of trial starts occurring on Day 0 ranging from 78% in Productivity to 89.9% in Business, with Health and Fitness at 82.1% and Gaming at 81.5%. After Day 3, trial starts fall below 5% of the total in every category. RevenueCat's own summary is that users who do not try immediately rarely try at all. Adapty's separate dataset puts the Day 0 share at 89.4%. These are two different panels and should not be read as confirming each other's exact figure, but they point firmly in the same direction.
Three consequences follow for anyone buying installs.
- The first few minutes do nearly all of the work. A lifecycle email on day three is competing for a small remainder. The onboarding flow and the first paywall are the only places this rate is really won.
- It is the fastest honest signal you have. Paid conversions are much slower. RevenueCat finds 50.6% of paid conversions on install day and 19.2% in week six or later, both as a share of eventual conversions rather than of installs. Trial starts resolve within a day. That makes install-to-trial the earliest downstream read on a new creative concept, well before a Day 7 ROAS figure means anything.
- The ad and the first screen are one experience. The user who taps an ad and starts a trial ten minutes later has not separated the two in their head, so neither should the team measuring them.
Is the ad wrong, or is onboarding and the paywall the bottleneck?
This is the question a weak install-to-trial rate actually raises, and it is usually argued rather than tested. The creative team points at onboarding, the product team points at traffic quality, and both are reasoning from an aggregate number that cannot settle it. Four cuts of your own data will settle most cases.
Split the rate by creative concept
If every concept lands on the same onboarding and the same paywall, then the product side is held constant and any spread in install-to-trial between concepts is being created by the ads. A wide spread means some concepts are buying installs the first session does not pay off. A narrow spread at a low level means the ads are not the variable. Group by concept rather than by individual ad, because a concept is the promise and the individual ads are just executions of it. On iOS, attribution is aggregated and delayed, so read concept-level splits on Android or through campaign structure rather than expecting clean per-ad data. Our note on SKAdNetwork postback delays covers why.
Compare paid installs with organic installs
Organic installs went through the same onboarding without seeing your ads. If organic install-to-trial is healthy and paid is weak, the first session works for people who arrived wanting the app, and the problem is who the ads are reaching or what they are telling them. If both are weak, stop blaming the traffic.
Find where in the first session people leave
Instrument the steps between install and trial: first open, onboarding started, onboarding completed, paywall viewed, trial started. Users who never reach the paywall point at onboarding. Users who reach it and do not start a trial point at the offer. Those need different fixes, and the offer is often cheaper to fix than people expect. In a contribution to RevenueCat's report, the team at Mojo describes anchoring their default yearly plan to its monthly equivalent in parts of Latin America. Trial starts rose 30% with no impact on trial-to-paid, and the price did not change. Framing is a lever before price is.
Check trial-to-paid by the same cut before calling a winner
A concept that produces lots of trials can still be the worst ad in the account if those trials do not pay. Over-promising is the most reliable way to raise install-to-trial, and trial-to-paid is where it gets found out. Wait for the cohort to resolve and rank concepts on install-to-paid.
| What you see | What it usually means | First move |
|---|---|---|
| Install-to-trial varies widely between concepts, organic installs convert normally | An ad problem. Some concepts are buying installs the first session does not pay off. | Rewrite or retire those concepts, and make the first onboarding screen continue the promise the ad made. |
| Uniformly low across every concept, organic just as low | An onboarding or paywall problem. The ads are not the variable. | Fix the first session before adding spend. More traffic will only buy more of the same result. |
| Users reach the paywall at a normal rate, few start a trial | An offer problem: price presentation, plan framing or trust. | Test how the offer is framed before you test the price itself. |
| Users leave before they ever see the paywall | An onboarding problem: too long, too generic, or value never shown. | Shorten the path to the first moment of value and cut screens that ask without giving. |
| A concept wins on install-to-trial and loses on trial-to-paid | Curiosity trials. The ad over-promised and the trial found out. | Judge concepts on install-to-paid once the cohort resolves, not on trial starts. |
| The rate drifts down as spend rises on the same concepts | Normal audience broadening, not a fault. | Recalculate the CPI you can afford at the new rate rather than assuming the old one holds. |
The first row is the one we see most often. A concept built around the outcome a user wants gets the tap, then the app opens on a generic welcome screen and asks for permissions, and the thread is lost. The fix is rarely a new ad. It is making the first onboarding screen say the same thing the ad said, in the same words, which is the message match point we make about the cost side in how to lower CPI without wrecking the funnel. It also helps to know which stage of awareness a concept is speaking to, because an ad that meets someone who has never heard of the problem needs an onboarding flow that does more explaining. That framework is in our piece on awareness levels in Meta ads.
What install-to-trial is worth in media terms
The chain is the same one that governs every stage of this funnel. Allowable CPI equals your target subscriber acquisition cost multiplied by install-to-trial multiplied by trial-to-paid. Install-to-trial sits directly in the multiplication, so a proportional change in the rate is a proportional change in the CPI you can afford.
The figures below are hypothetical and are not client data. Take a North American app spending £10,000 a month on Meta at a £2.00 CPI, so 5,000 installs, converting trials to paid at the regional median of 34.2%, with a £60 target cost per subscriber. Hold all of that constant and vary only install-to-trial: a weak rate, the regional median of 7.1%, and the regional top quartile threshold of 15.0%.
| Install-to-trial | Trials from 5,000 installs | Cost per trial | Subscribers | Subscriber acquisition cost | Allowable CPI at £60 target |
|---|---|---|---|---|---|
| 3.5% | 175 | £57.14 | 59.9 | £167.08 | £0.72 |
| 7.1% | 355 | £28.17 | 121.4 | £82.37 | £1.46 |
| 15.0% | 750 | £13.33 | 256.5 | £38.99 | £3.08 |
At the median, this app pays £82.37 for a subscriber it wanted at £60, and can only afford a £1.46 CPI while paying £2.00. At the top quartile threshold, with identical media, it acquires subscribers at £38.99 and could pay £3.08 an install. At 3.5% it is not a business. Nothing in the ad account differs between the three rows.
Two things are worth drawing out. First, the published spread in install-to-trial is wider than the spread in trial-to-paid. Top quartile thresholds sit at roughly twice the median here, against roughly one and a half times for trial-to-paid, so for most apps this is the bigger lever of the two. Second, it is the lever the media buyer and the product team share. A better concept and a better first screen both move it, which is why the cost of a trial belongs in the same conversation as the cost of an install, as we argue in CPI vs CPA vs CAC for subscription apps. The full calculation for the cost side is in our guide to subscription app CAC, and the install costs themselves are in the 2026 mobile app CPI benchmarks.
A note on optimisation. Start Trial is one of Meta's standard app events, and because trial starts land within hours of the install it gives the algorithm a faster and denser signal than a subscription does. That is a real advantage. It also means a campaign optimised for trials will find you people who start trials, including people who never intended to pay, which is exactly the failure in the fifth row of the table above. If you optimise for trial starts, read trial-to-paid by campaign alongside it.
What this looks like in an account
Most of the subscription apps we work with arrive describing a cost problem, and a fair number of them turn out to have a first-session problem that the cost was hiding. Our Steps & Beasts case study is the clearest example on our site: heavy creative testing ran alongside onboarding changes that lifted install-to-subscriber conversion, and revenue rose 145% with active subscriptions up 118%. The creative found the people, and the onboarding stopped losing them in the first session.
The Oli Help case study shows the other half: an ICP-led creative strategy aimed at the parents the app was built for, which grew trials by 1,750% and revenue by 394%. To be precise about what that figure is, it is growth in the number of trials, not a change in the install-to-trial rate, and we would not present it as a benchmark. What it does show is that reaching the right people is itself a trial-rate decision, made in the ad account before anyone opens the app.
Frequently asked questions
What is a good install-to-trial conversion rate for a subscription app?
For an app that uses a free trial to acquire subscribers, RevenueCat's 2026 data puts the median share of installs starting a trial within 30 days at between 4.0% and 9.1% depending on category, with most non-gaming categories between 5% and 6%. By market the median is 7.1% in North America, 5.7% in Asia-Pacific, 5.0% in Western Europe and between 3.0% and 3.7% in the remaining regions. Top quartile apps typically clear roughly double their slice's median. Compare yourself against the slice that matches your category and market mix rather than a single global figure.
How is install-to-trial conversion rate calculated?
Divide the number of free trial starts by the number of installs in the same cohort. RevenueCat calls this download-to-trial and counts trials started within 30 days of the download date, and only calculates it for apps that use a trial-based acquisition strategy. Adapty computes install-to-trial as trials divided by installs for each app, then averages those per-app rates across the segment. State the window and the definition of an install whenever you quote your own figure, because both change the number.
Why is Adapty's install-to-trial benchmark higher than RevenueCat's?
Adapty reports 14.5% for North America and RevenueCat reports a 7.1% median for the same region, so the gap is roughly a factor of two. The two most likely reasons are the statistic and the panel. Adapty publishes an average of per-app rates and RevenueCat publishes a median, and on a distribution where the top quartile sits at around twice the median an average lands well above the typical app. Adapty's data also comes mostly from the Apple App Store and covers 2025, while RevenueCat's panel of more than 115,000 apps includes Google Play. Neither is wrong. They should never be mixed in one funnel.
When do users start free trials in a subscription app?
Almost always on the day they install. RevenueCat's 2026 data shows between 78% and 89.9% of trial starts happening on Day 0 depending on category, with trial starts falling below 5% of the total after Day 3 in every category. Adapty's separate dataset puts the Day 0 share at 89.4%. In practice install-to-trial is a verdict on the first session: the promise in the ad, the onboarding flow and the first paywall.
Is a low install-to-trial rate an ad problem or an onboarding problem?
Split the rate before you decide. If install-to-trial varies widely between creative concepts that all land on the same onboarding, the gap is being created by the ads, usually because some concepts make a promise the first screens do not pay off. If the rate is uniformly low across every concept and organic installs convert just as poorly, the bottleneck is onboarding or the paywall. If users reach the paywall at a normal rate but few start a trial, look at the offer and how it is framed.
How much does install-to-trial affect the CPI I can afford?
Proportionally. Allowable CPI equals your target subscriber acquisition cost multiplied by install-to-trial multiplied by trial-to-paid. On a hypothetical app with a £60 target cost per subscriber and a 34.2% trial-to-paid rate, moving install-to-trial from 7.1% to 15.0% raises the CPI it can afford from £1.46 to £3.08. Because the published spread in install-to-trial is wider than the spread in trial-to-paid, it is often the bigger lever of the two.
Should I compare my paid installs against these benchmarks?
Carefully. Neither report splits install-to-trial by acquisition source, so the benchmarks blend paid, organic and referral installs. Someone who searched the App Store for your category arrives with more intent than someone who was interrupted by an ad, so a heavily paid install base will usually sit below a blended benchmark without anything being broken. Track your paid cohorts against your own paid history first and use the published medians as a sanity check.
The short version
Install-to-trial is the share of installs that start a free trial, and the typical trial-based subscription app lands between 5% and 7%, with category medians from 4.0% to 9.1% and a top quartile at roughly twice the median. Adapty's higher figures are an average of per-app rates from a mostly App Store panel, not a contradiction, and should never be mixed with RevenueCat's in one model. Because the large majority of trials start on the day of install, the rate is decided in the first session, which makes it both the fastest signal a paid acquisition team has and the one most often misdiagnosed. Split it by concept, compare paid with organic, find where the first session loses people, and check trial-to-paid before you call a winner.
If you already know your install-to-trial rate by concept, you are ahead of most accounts. If you do not, that is usually the first thing we build. Our consumer apps page sets out how we run subscription acquisition on Meta, the creative refresh calculator gives a rough read on how many concepts your spend level needs to test, and if your app is already spending, our pricing is published.