Why Aren't My App Installs Converting Into Trials?

Installs are the easiest number to buy and the least informative one to own. The step that decides whether a subscription app has a business is the next one, where an install becomes a trial, and that is also the step where most launch budgets quietly disappear.

RevenueCat’s State of Subscription Apps 2026 puts median install-to-trial conversion at 9.1% in Business apps and 4.4% in Gaming, with the top quartile of Business apps above 16.2%. Category explains part of that spread. Within one category, the gap between median and top quartile is mostly execution.

Why don’t installs turn into trials?

Installs fail to become trials because the ad and the paywall are selling different things to the same person. The ad promises a specific outcome, the user installs expecting it, and the paywall then asks them to pay for a general product they have to translate back into the thing they wanted.

That gap is structural in how most app marketing is organised. Creative is produced by whoever runs acquisition and is judged on installs and cost per install. The paywall is owned by whoever runs monetisation and is judged on conversion of whatever traffic arrives. Both teams can hit their targets while the seam between them loses most of the money.

The second common cause is distance. Every screen between the tap and the paywall is a chance to leave, and the screens that ask for something before giving anything, a permission, an email address, a questionnaire, lose the most. An app with twelve onboarding screens has eleven opportunities to lose a user it already paid for.

The third is that the traffic was never qualified. Broad targeting and a scroll-stopping opening buy installs from people with no intent to pay for anything, and cheap installs of that kind look like a win in the dashboard right up to the point where cost per paying user is calculated.

All three produce the same symptom and need different fixes, which is why diagnosis has to come before changes.

What is a normal install-to-trial rate?

A normal install-to-trial rate depends on category more than on anything you control, so the benchmark to use is your own category’s median. RevenueCat’s 2026 figures give a usable spread across the categories most consumer apps fall into.

Category	Median install-to-trial
Business	9.1%
Health and Fitness	6.9%
Education	6.5%
Utilities	6.5%
Gaming	4.4%

According to the same RevenueCat report, the Business top quartile sits above 16.2%, roughly 1.8 times that category’s median, which is the gap worth paying attention to. Categories differ because of intent and willingness to pay, and within a category the difference between median and top quartile is funnel design.

Channel matters as well as category. According to Adapty’s 2026 Apple Ads benchmarks, install-to-trial on Apple Ads traffic runs at 3.66%, around 2.5 times what the same dataset measured on other paid channels, because store search traffic arrives with intent that paid social traffic has to be given.

So read your own number three ways before concluding anything: against your category’s median, against your own trend over the last eight weeks, and separately for each traffic source. A blended figure averaging high-intent search traffic with cold social traffic hides the only comparison that would tell you what to fix.

What does RevenueCat’s State of Subscription Apps 2026 count as an install-to-trial conversion?

RevenueCat’s State of Subscription Apps 2026 counts an install-to-trial conversion as an install that starts a free trial within 30 days of the download date, and it calculates the metric only for apps running a trial-based acquisition strategy. Both halves of that definition matter before you compare your own number against the medians above.

A 30-day window is wider than most internal dashboards use, so a figure measured on a 7-day window will read lower than the benchmark for reasons that have nothing to do with the funnel.

And an app with a hard paywall is not in the panel at all, which means the benchmark is not the right comparison for one.

Price is the variable in the report that separates apps more sharply than most teams expect. RevenueCat’s 2026 report puts the median download-to-trial rate at 8.9% for higher-priced apps, 5.4% in the middle tier and 4.4% for lower-priced ones, roughly a two times spread driven by who bothers to download an expensive app in the first place.

Price tier	Median download-to-trial, 2026
Higher-priced	8.9%
Mid-priced	5.4%
Lower-priced	4.4%

One number in circulation is from the wrong edition. The pairing of 8.9% against 4.3% appears in several 2026 summaries, but 4.3% is the low-priced median from RevenueCat’s 2025 report, where the high-priced figure was 9.8%. The 2026 low-priced median is 4.4%, per the report itself, so a summary quoting 4.3% alongside a 2026 high-priced figure has mixed two years together.

The practical consequence is that a cheap app with a low install-to-trial rate may be performing normally, and the fix is in pricing or positioning rather than in the paywall. Raising price raises the intent of the person who downloads, which is the mechanism behind that ladder, and how to price a subscription app works through what that costs in volume.

What does click-to-install time tell you about installs that never convert?

Click-to-install time tells you whether the installs you are measuring came from people who chose your app.

It is the gap between an ad click and the first launch of the app, per Adjust’s definition of the metric, and it is the cheapest way to find out that part of your install volume was never a person making a decision at all.

That matters on this page because installs of that kind convert to trials at zero, by construction, and they drag the blended rate down while looking like ordinary traffic in the dashboard.

Two patterns are worth reading in the distribution rather than the average.

A cluster of very short times, within a few seconds of the click, is the signature of click injection, where a click is fired at the moment an install is already happening so the fraudster is credited for it: thresholds published by fraud vendors sit around 3 to 10 seconds, per TrafficGuard’s guide to detecting fraud from click-to-install time.

A flat distribution with no concentration anywhere is the signature of click spamming, where clicks are fired at volume in the hope of catching organic installs. Genuine traffic looks like neither: most installs land inside the first hour with a tail running out over the following day or two.

What the distribution looks like	What it usually means	What to do about it
Heavy cluster under 10 seconds	click injection on that source	exclude the source and reconcile with store installs
Flat with no peak	click spamming, organic installs being claimed	compare the source’s installs against your organic baseline
Peak inside the first hour, tail over a day or two	ordinary user behaviour	read install-to-trial as real

So pull click-to-install time by source before concluding that a paywall or an onboarding flow is at fault.

If one source shows a spike of installs arriving seconds after their clicks, its install-to-trial rate is measuring attribution rather than behaviour, and removing it from the blended number usually moves the headline rate more than any paywall test will.

The definitions vary between attribution providers, with some measuring to the first open and others to when the install is registered, so read your own provider’s definition before setting a threshold on it.

Does the ad promise have to match the paywall promise?

The ad promise and the paywall promise have to describe the same outcome in recognisably the same language, and the measured value of doing that is large. Adapty’s 2026 Apple Ads benchmarks, built on more than 8,000 apps running over a million ad groups across 90 countries, found that tailored paywalls lifted install-to-trial by 41% and trial-to-paid by 24%.

The effect compounds rather than fading. The same Adapty dataset found that the ROAS advantage of a tailored paywall over a default one ran at 48.4% on day zero and widened to 82.5% by day 92, because a user who was promised the right thing is also a user who keeps paying for it.

Matching does not mean cloning the ad. It means the outcome named in the creative appears in the paywall headline, the feature the ad demonstrated is the first one listed, and the words the creator used survive into the offer. A user should not have to work out that the two screens are about the same product.

Where several distinct promises are driving installs, one paywall cannot serve them all. That is the case for separate paywalls by campaign or by acquisition source, which is an ordinary capability in subscription platforms now and is usually left unused.

The seam runs the other way too. If the paywall can only sell one outcome credibly, that outcome is the one the ads should promise, and creative that performs on installs while promising something else is costing you money rather than making it.

How do you find the mismatch in your own funnel?

Find it by laying out the drop-off at every step in order and looking for the largest single fall. Impression to tap, tap to store listing view, listing view to install, install to onboarding completion, onboarding to paywall view, paywall view to trial start. One of those will be much worse than the rest.

Each location implies a different cause. A weak listing view to install rate is a store listing problem. A weak install to paywall view rate is onboarding. A healthy paywall view rate with a weak trial start rate is the paywall or the offer. Install-to-trial on its own cannot distinguish between them, which is why the blended number is a poor diagnostic.

Then cut the same funnel by creative and by audience. If one creative delivers installs that convert at three times the rate of another at comparable volume, the product is fine and the losing creative is promising something your paywall does not deliver.

Watch the step where users first see a price, not just the step where they start a trial. A paywall that is viewed and abandoned is a message problem.

A paywall that is rarely viewed at all is an onboarding problem dressed up as a pricing one. Check what your tool counts as a view before you read either, because the major paywall tools count impressions rather than people, which is worked through in what the denominator is in a paywall conversion rate.

Do this weekly during a launch and write down what changed. Most of the diagnosis is noticing that last week’s improvement in installs came with a fall in trials, which blended reporting will average away.

What should you change first?

Change the largest drop-off first, and change one thing per cycle. The temptation when conversion is poor is to rewrite the paywall, redesign onboarding and refresh the creative at once, which guarantees you cannot attribute the result.

If onboarding is where users disappear, cut screens rather than redesign them, and move every request for a permission or an account to after the first moment the app does something useful. Our piece on how many onboarding screens an app should have covers how far that can be cut.

If the paywall is where they disappear, start by rewriting the headline to name the outcome the ad promised, before touching price or structure. Message changes are cheap, reversible and frequently larger in effect than a pricing change, and they do not disturb the economics you already modelled.

If the creative is bringing in people who never convert, kill it even while it is winning on cost per install. The number to optimise is cost per paying user, and a cheap install that never converts raises it.

This is the part of an app launch we take on end to end, because the ad promise, the store listing, the onboarding and the paywall only work when one person is responsible for all four. The work is described on our app launch page.

When is the problem the audience rather than the creative?

The problem is the audience when every creative converts at a similar low rate within one targeting setup, and a different setup converts better on the same creative. That pattern says the message is fine and it is reaching people who were never going to pay.

Broad targeting with a strong opening is the usual way this happens. A scroll-stopping first second buys installs from everyone, including people with no interest in the product, and the install-to-trial rate drops as volume rises. Install volume looks healthy and cost per paying user climbs.

Price sensitivity by market compounds it. According to RevenueCat’s 2026 figures, median day-35 download-to-paid conversion is 2.6% in North America against 1.4% in India and Southeast Asia, so the same app with the same funnel converts at very different rates depending on where the installs come from. Spend distributed across markets without separate targets produces a blended number nobody can act on.

Check the obvious mechanical causes before concluding it is the audience, though. Automated placements sending traffic to surfaces where accidental taps are common, or a campaign optimising to installs rather than to a post-install event, will both produce a low conversion rate that has nothing to do with who was targeted.

Then move the optimisation event down the funnel as soon as volume allows. Optimising for trial starts rather than installs costs more per install and usually lowers cost per paying user, which is the number that matters.

The short version

Install-to-trial is where the money is won or lost, and the usual cause of a bad rate is that the ad and the paywall are selling different things. According to RevenueCat’s 2026 benchmarks, the category medians run between 4.4% in Gaming and 9.1% in Business, with the Business top quartile above 16.2%, so compare yourself to your own category and your own trend before anything else.

Diagnose by step rather than by instinct. Lay out the drop-off from impression to trial start, find the largest single fall, then cut the same funnel by creative and by audience. An install problem, an onboarding problem and a paywall problem all look identical in a blended conversion rate.

Then fix the seam. Adapty’s 2026 benchmarks across 8,000 apps found that tailored paywalls lifted install-to-trial by 41% and trial-to-paid by 24%, with the ROAS gap widening from 48.4% on day zero to 82.5% by day 92. Put the outcome the ad promised in the paywall headline, cut onboarding screens that ask before they give, and judge creative on cost per paying user rather than cost per install.