Table of Contents
App paywall optimization is the practice of tuning the screen where a user decides to pay: the model (hard, freemium or hybrid), the trial length, the price points, the plan mix and the timing of the ask, so that revenue per install rises without collapsing top of funnel volume. Most teams file it under product. Admiral Media files it under media buying, because the paywall sets the revenue per install that every bid, every target ROAS and every payback window is calculated against.
Put plainly: your paywall decides what you can afford to pay for a user. Change the paywall and you have changed your bid ceiling, your channel mix and your scaling headroom, before anyone touches an ad account. That is why the Admiral Media team treats paywall design as an input to the media plan rather than a downstream detail.
This guide covers what the current public data says about paywall models and trial lengths, how Admiral Media connects paywall revenue back into platform bidding, and what the Admiral Media team has seen across subscription clients including ChatPDF, NeuroNation and Fastic.
Why does a performance agency care about your paywall?
Because the paywall is the only screen that converts attention into money, and every number a media buyer optimises toward is derived from it. An agency that ignores the paywall is optimising a ratio while somebody else controls the numerator.
Work through the arithmetic. If a campaign delivers installs at 3 euros and the paywall produces 1.50 euros of day 60 revenue per install, that campaign is a loss you have to justify with a long payback story. Hold the same traffic, same creative, same bid, and lift day 60 revenue per install to 3.50 euros, and the same campaign is now profitable inside two months. Nothing changed in the auction. Everything changed in what you can pay.
Admiral Media has managed more than 500 million euros in ad spend across 150+ brands, and in subscription accounts the pattern repeats: the constraint on scale is almost never the media team’s ability to find volume. It is the revenue per install the product hands them to bid with. When NeuroNation’s co-founder Jakob Futorjanski described the work with Admiral Media, the line that mattered was not about ad accounts: the team “got a lot of useful feedback to optimise our store listing pages and onboarding flows”. That feedback loop, from auction back into product, is the point.
There is a second reason, and it is technical. Modern bidding on Google App Campaigns and Meta Advantage+ App Campaigns is value based. Target ROAS bidding needs revenue signal to learn from. Under SKAdNetwork and AdAttributionKit, iOS gives you a small, coarse, time boxed window in which to encode value. The paywall determines what happens inside that window. If the purchase event lands after the measurement window has closed, the bidding algorithm never sees it, no matter how good the paywall was. Paywall design and measurement design are the same problem wearing different hats.
Hard paywall or freemium: which converts better in 2026?
Hard paywalls convert substantially better on trial to paid, and the gap is now large enough that it should be an explicit business decision rather than a default. According to RevenueCat’s State of Subscription Apps 2026, drawn from over 115,000 apps representing more than 16 billion dollars in revenue, apps with a hard paywall show a median day 35 trial to paid conversion rate of 10.7 percent, against 2.1 percent for freemium apps. That is roughly a five times advantage.
The revenue gap is wider still. RevenueCat reports that hard paywall apps generate 3.09 dollars of revenue per install at day 60, compared with 0.38 dollars for freemium apps, an eight times difference. For a media buyer, that gap is the whole game: two apps bidding in the same auction, with the same creative, can have bid ceilings that differ by an order of magnitude purely because of what happens after the install.
Two caveats matter before anyone rips out a freemium tier on Monday morning. First, the same RevenueCat dataset shows retention barely differs: freemium apps retain 28 percent of yearly subscribers after one year, hard paywall apps 27 percent. The hard paywall advantage is front loaded, not compounding. Second, hard paywall conversion is drifting down, from 12.1 percent in 2025 to 10.7 percent in 2026, which suggests users are getting more resistant to the immediate ask rather than less.
There is also a top of funnel cost that the conversion headline hides. Business of Apps, in its App Subscription Trial Benchmarks (2026), reports that with a hard paywall 78 percent of users who start a trial do so in the first week after downloading. Decisions get made fast. Users who were never going to decide fast simply leave, and they leave having cost you a full install price.
In Admiral Media’s experience running subscription accounts, the honest framing is not “hard paywall good, freemium bad”. It is a trade between speed of cash and size of addressable audience, and the right answer depends on how expensive your traffic is and how patient your balance sheet can be.
| Paywall model | What the user sees | Effect on paid media | Fits best when |
|---|---|---|---|
| Hard paywall | No core value without a subscription or trial start | Highest early revenue per install, so the highest sustainable CPI and the fastest payback; smaller addressable pool | Traffic is expensive, the value proposition is obvious in one screen, and cash needs to come back inside 60 days |
| Freemium | Meaningful free tier, upgrade prompted later | Lowest early revenue per install, so bids must stay low; conversion continues well past week 6, which stresses attribution windows | Organic and viral loops are real, retention is strong, and the business can fund a long payback |
| Hybrid or soft paywall | Paywall appears after a defined value moment | Middle ground on revenue per install; needs event level instrumentation so bidding can see the value moment | The aha moment is identifiable and reachable inside the first session |
| Web paywall before install | Checkout completes on web, app is delivered after purchase | Purchase happens where you can see it, which restores signal quality on iOS and removes store commission from the unit economics | You can support a web funnel and the offer survives a browser checkout |
The Admiral Media Paywall-to-Bid Loop
The Admiral Media Paywall-to-Bid Loop is the framework the Admiral Media team uses to convert paywall changes into bid changes without guessing. It exists because the failure mode in most subscription accounts is not a bad paywall test, it is a good paywall test that never reaches the bidding algorithm.
The six steps of the Admiral Media Paywall-to-Bid Loop
- Measure revenue per install, not conversion rate. Conversion rate is a vanity ratio if price and plan mix move at the same time. Fix a cohort window that matches your reporting reality, usually day 7, day 30 and day 60, and track revenue per install inside it. This is the number that becomes a bid.
- Set the bid ceiling from the payback target, not from habit. Divide target revenue per install by the payback multiple the business will tolerate. If day 60 revenue per install is 3 euros and the business wants money back by day 60, the ceiling is 3 euros of blended cost per install, minus store commission and refunds. Everything above that is a financing decision, not a marketing one. Admiral Media covers the full calculation in the CAC payback period guide.
- Choose the proxy event that arrives inside the attribution window. On iOS the measurement window is short and coarse. Pick the earliest event that reliably predicts paid conversion, typically trial start or a specific onboarding completion, and map it to a conversion value. A perfect signal that arrives too late is worth less than a rough signal that arrives on time.
- Encode value, not just occurrence. Send revenue or modelled revenue to the platform rather than a flat conversion. Value rules and predicted lifetime value inputs let target ROAS bidding separate a 9.99 monthly signup from a 79.99 annual one. Without that, the algorithm optimises toward whichever plan is easiest to sell, which is rarely the profitable one. See predictive LTV bidding for the mechanics.
- Rerun the ladder after every paywall change. A paywall test that lifts revenue per install by 20 percent has raised your bid ceiling by 20 percent. Raise target CPA or lower target ROAS deliberately and in one step, then let the campaign re-learn. Teams that leave bids untouched after a winning paywall test capture the margin and forfeit the volume.
- Feed auction learning back into the paywall. The channels tell you which promise wins attention. If the winning creative angle is a specific outcome, that promise should be the first line of the paywall. Mismatch between the ad hook and the paywall headline is the most common and most fixable leak the Admiral Media team finds in subscription funnels.
Step six is the one teams skip. It costs nothing, it requires no engineering, and in most accounts it is worth more than another round of button colour tests.
How long should your free trial actually be?
Longer trials convert better, and the industry is moving in the opposite direction. RevenueCat’s 2026 data shows trials of 17 to 32 days converting at a median of 42.5 percent trial to paid, against 25.5 percent for trials shorter than four days, roughly 70 percent better. Yet 46.5 percent of apps now use trials of under four days, up from 42.1 percent in 2025.
That is not irrational. Short trials return cash faster and generate experiment data faster, which matters when you are funding media spend from revenue. It is a cashflow decision dressed up as a conversion decision, and it should be named as such rather than defended on conversion grounds it cannot win.
Business of Apps reports the same tension from the cancellation side: a three day trial averages 26 percent cancellations, while a 30 day trial averages 51 percent. More time to evaluate means more time to change your mind. The right trial length is therefore a function of how quickly your product can prove itself, not a number you copy from a benchmark table.
| Trial length | Median trial to paid | What it does to cashflow | What it demands from the product |
|---|---|---|---|
| Under 4 days | 25.5% | Revenue lands inside a week, so paid media can be refunded quickly and tests iterate fast | The aha moment must land in the first session, because most cancellations never see day two |
| 17 to 32 days | 42.5% | Cash arrives weeks later, which requires working capital and patience from finance | Habit formation, so onboarding, notifications and lifecycle messaging must carry the user across weeks |
| No trial, pay up front | Not applicable | Fastest possible cash return and the cleanest signal for value based bidding | A proposition strong enough to be bought on one screen, and a price low enough to be an impulse |
Why do most trial cancellations happen in the first hour?
Because users treat trials as trials of the product, not trials of the billing arrangement, and they make the call almost immediately. RevenueCat’s 2026 data is stark on this: 55.4 percent of all three day trial cancellations occur on day zero, and 84 percent occur between day zero and day one.
Read that again in media buying terms. You paid for the install, you paid for the trial start, and more than half of the people who will cancel had already cancelled before your dashboard refreshed. Your three day trial is functionally a one hour trial. The window in which you can influence the outcome is the first session.
The implication is that onboarding is not a product nicety sitting adjacent to the paywall, it is the paywall’s conversion engine. Every screen between install and the aha moment is a tax. The Admiral Media team’s standing recommendation for subscription clients is to instrument the onboarding funnel step by step, find the step with the largest drop, and fix that before touching paywall copy. Copy tests on a paywall nobody reaches with intact motivation are theatre.
This is also why creative and paywall must tell one story. If the ad promised a specific outcome and the first post install screen asks for notification permissions, the user has already been handed a reason to reconsider. Admiral Media’s work on creative testing frameworks exists partly to make that promise explicit, so it can be carried through to the paywall rather than lost at the store listing.
What happens to renewals after the first payment?
Annual subscriptions are not the twelve month guarantee most forecasts assume. RevenueCat reports that the first month accounts for 35 percent of all annual cancellations, after which cancellations settle to 3 to 10 percent per month before spiking again in month 12 ahead of renewal. In the 2026 dataset, roughly 72 percent of annual subscribers cancelled during year one, up from around 56 percent a year earlier.
Users are switching off auto renew early and treating the annual plan as a one off purchase for the current year. For anyone building an LTV model to bid against, that has a direct consequence: a model that assumes annual plans renew at historical rates will overstate lifetime value and licence bids the business cannot actually afford. Admiral Media’s guidance here is deliberately conservative. Bid against realised cohort revenue you have already banked, and treat renewal upside as headroom rather than as budget.
There is a platform specific leak worth naming too. RevenueCat’s data shows 31 percent of Google Play subscription cancellations are involuntary billing failures, against 14 percent on the App Store. That is not a marketing problem and no amount of creative fixes it. It is dunning logic, retry windows and grace periods, documented in the Google Play Billing subscription lifecycle documentation. On Android accounts, recovering involuntary churn is frequently a higher return activity than another round of audience testing, and it costs no media budget at all.
How do you get paywall revenue into the bidding algorithm?
By sending value rather than events, and by sending it inside the window the platform can actually use. This is the step where most subscription accounts quietly lose the advantage their paywall work created.
Three mechanisms do the heavy lifting. First, proxy events: choose the earliest reliable predictor of paid conversion, usually trial start or a defined onboarding completion, and optimise toward it while the true purchase event remains too sparse or too late to train on. Second, value assignment: attach revenue or modelled revenue so that target ROAS bidding can distinguish plan types. Third, conversion value mapping on iOS, where a limited set of values must encode both the event and its worth inside a short postback window described in Apple’s App Store subscriptions documentation and implemented through SKAdNetwork or AdAttributionKit.
Volume matters as much as accuracy. Value based bidding strategies need enough weekly conversion events to exit the learning phase, which is why consolidating campaign structure usually beats adding another test campaign. In Admiral Media’s work with ChatPDF, the account restructure came first: analyse the data, remove overlaps, consolidate ad sets, and only then test target CPA against target ROAS with value rules and an LTV signal layered on. Sequencing matters. Value bidding on a fragmented account learns nothing.
What did paywall aware bidding do for ChatPDF?
Admiral Media rebuilt ChatPDF’s paid acquisition around value based bidding and a cadenced creative testing programme, and the account delivered a 320 percent increase in ROAS, a 156 percent increase in subscriptions and a 42 percent reduction in CAC. The numbers are indexed year one against year two year to date, and the goal was to scale paid acquisition during the low season while improving both ROAS and CAC efficiency at the same time.
The channel split is the interesting part, because it shows two different routes to the same outcome.
Google delivered the larger ROAS improvement at 320 percent year over year growth with a 38 percent CAC reduction and 142 percent subscription growth. Meta delivered less ROAS growth at 280 percent, but more subscription volume at 171 percent and the deeper CAC cut at 45 percent. Same product, same paywall, two different efficiency profiles. That is exactly the kind of split that only becomes visible once revenue signal is flowing correctly into both platforms, and it is the basis for how budget should then be allocated between them.
The account team’s own summary of what moved the needle was blunt: a new campaign structure based on Admiral Media best practices increased efficiency drastically, value bidding plus audience analytics increased ROAS while spend scaled, and cadenced testing improved conversion rate and CAC over time. Structure, then signal, then creative cadence. In that order.
Does this hold for apps that are not AI tools?
It does, and the pattern shows up in older Admiral Media engagements too. Admiral Media managed NeuroNation’s user acquisition across creative testing and channel exploration using a systematic test and learn approach, and delivered a 117 percent increase in ROAS, a 66 percent increase in installs, a 32 percent increase in purchases, a 42 percent increase in net cohort revenue and a 39 percent reduction in CPI, on data from January to August 2019.
The detail worth pulling out is that net cohort revenue rose 42 percent while purchases rose 32 percent. Revenue grew faster than purchase count, which means the mix shifted toward higher value plans and cohorts, not just more transactions. That is the fingerprint of monetisation work and acquisition work moving together rather than in sequence.
Fastic is the volume end of the same idea. Admiral Media helped take Fastic to the position of the world’s number one fasting app, with growth across the engagement including a 639 percent increase in installs, a 1,655 percent increase in purchases, a 439 percent increase in revenue and a 952 percent increase in monthly active users. In 2026 the Admiral Media AI Creative Factory cut Fastic’s cost per result by 70 percent, work that won Silver in the AI category at The Drum Awards for Marketing EMEA 2026. Purchases growing at more than twice the rate of installs is the signal that the funnel behind the ad was converting better, not just receiving more traffic.
PURE shows the third lever, which is where you buy. Admiral Media tested a demand side platform against an established self attributing network for PURE’s US Android campaigns and recorded a cost per install of 2.44 dollars against 9.43 dollars, described in the case study as four times lower and reported as a 74 percent CPI reduction, with day 7 ROAS goals exceeded, which led to additional market launches. Same paywall, same product: a different traffic source changed what the unit economics could support.
How should you test a paywall without wrecking your campaigns?
Carefully, and one variable at a time, because a paywall test is also a bidding test whether you intended it to be or not. The moment revenue per install moves, the algorithm’s understanding of your account changes, and a test that runs across a learning phase reset produces a result you cannot trust.
The Admiral Media team’s working rules for subscription clients are straightforward. Hold media conditions constant during a paywall test: no bid changes, no budget step changes, no new campaign launches. Run the test long enough to capture the real decision window, which for a three day trial means at least the day 0 to day 1 cancellation spike and ideally a full renewal cycle. Read the result on revenue per install rather than on trial starts, because a paywall that lifts trial starts and lowers paid conversion is a loss dressed as a win. Then, once a winner is confirmed, change bids deliberately in one move rather than drifting them upward over a fortnight.
One more rule, learned expensively across accounts: do not run a paywall test and a creative test in the same window. Both move revenue per install. When they run together you get a number you cannot attribute to either, and the temptation is to credit whichever change you were more emotionally invested in.
Where does web to app fit into paywall optimization?
A web funnel moves the paywall in front of the install, which changes both the economics and the measurement. The purchase happens on your own property, so you see it directly rather than inferring it through a privacy limited postback, and the store commission on that first transaction is removed from the unit economics.
The trade is friction. You are asking someone to complete a checkout in a browser before they have the product, then get them through an install and a login. It works when the proposition is strong enough to survive that sequence, and it fails when the value is only obvious after use. Admiral Media covers the mechanics in detail in the guide to web to app funnels for subscription apps, including the comparison against running install ads directly.
For measurement constrained iOS accounts, the signal benefit alone is often the deciding argument. Clean, deterministic purchase data on a web funnel feeds value based bidding better than any conversion value schema, and better signal compounds into better targeting over time.
What should you fix first?
Fix the thing with the largest gap between current state and available benchmark, and fix it before you spend another euro on media efficiency. In most subscription accounts Admiral Media reviews, the order below reflects where the recoverable value actually sits.
- Onboarding drop off before the paywall. Cheapest to fix, largest effect, and it improves every downstream number. If most cancellations happen on day zero, the first session is where the money is.
- Ad promise to paywall promise mismatch. Free to fix. Copy the winning creative hook into the paywall headline and measure revenue per install.
- Signal quality into the bidding platform. Value rules, proxy events and conversion value mapping. Without this, every product improvement stays invisible to the algorithm buying your traffic.
- Involuntary churn on Android. Dunning and grace periods. No media budget required, and the industry data suggests the leak is material.
- Trial length and paywall model. Highest impact, highest risk, longest read time. Do this last, when the measurement around it is trustworthy.
- Price and plan architecture. Only once everything above is stable, because price changes contaminate every other test running at the same time.
Teams routinely start at item five or six because those feel like the strategic decisions. They are, but they are also the ones you cannot read accurately until the first four are in order.
Working with Admiral Media on subscription growth
Admiral Media is a performance marketing agency for apps, games and ecommerce brands, with more than 500 million euros in managed ad spend, 150+ brands grown, a 5.0 rating on Clutch and Silver at The Drum Awards for Marketing EMEA 2026 in the AI category. The Admiral Media team works on subscription app marketing where paywall economics, measurement and media buying are treated as one system rather than three departments.
If the paid account is efficient and growth still stalls, the constraint is usually sitting on the paywall. That is a solvable problem, and it is usually cheaper to solve than another percentage point of CPI.
Frequently Asked Questions
What is app paywall optimization?
App paywall optimization is the process of improving the screen where users decide to subscribe, covering the paywall model, trial length, pricing, plan mix and when the paywall appears, with the goal of raising revenue per install. It matters to paid media because revenue per install determines the maximum cost per install a business can profitably pay. A paywall change that lifts revenue per install by 20 percent raises the affordable bid by roughly the same proportion. Admiral Media treats it as part of media strategy rather than as a product task handled separately.
Do hard paywalls convert better than freemium?
Yes on trial to paid conversion, according to RevenueCat’s State of Subscription Apps 2026, which reports a median day 35 trial to paid rate of 10.7 percent for hard paywall apps against 2.1 percent for freemium apps. Hard paywall apps also generate 3.09 dollars of revenue per install at day 60 versus 0.38 dollars for freemium. However, one year retention of yearly subscribers is almost identical at 27 percent for hard paywall apps and 28 percent for freemium, so the advantage is front loaded rather than compounding. The right choice depends on how expensive your traffic is and how long your business can wait for payback.
How long should a free trial be for a subscription app?
Longer trials convert better on the current public data: RevenueCat reports trials of 17 to 32 days converting at a median 42.5 percent trial to paid, compared with 25.5 percent for trials under four days. Despite that, 46.5 percent of apps now use trials shorter than four days, mainly for faster cash return and faster experiment cycles. Business of Apps reports a three day trial averages 26 percent cancellations against 51 percent for a 30 day trial, so longer trials also give users more time to reconsider. Choose based on how quickly your product can prove its value and how much working capital the business has.
Why do so many users cancel a trial on the first day?
Because users evaluate the product immediately rather than across the whole trial window. RevenueCat’s 2026 data shows 55.4 percent of three day trial cancellations occur on day zero and 84 percent occur by day one. In practice this means a three day trial is decided in the first session, so onboarding is the real conversion mechanism. The highest return fix is usually reducing friction between install and the first moment of genuine value, not rewriting the paywall.
How do you connect paywall revenue to Google and Meta bidding?
Send value, not just events, and send it early enough for the platform to use. That means choosing a proxy event such as trial start that arrives inside the attribution window, attaching revenue or modelled revenue so target ROAS bidding can tell plan types apart, and mapping conversion values correctly on iOS through SKAdNetwork or AdAttributionKit. Value based bidding also needs sufficient weekly conversion volume to exit the learning phase, which usually means consolidating campaign structure before layering value rules on top. In Admiral Media’s work with ChatPDF, restructuring the account came first, then value bidding, then cadenced creative testing.
What results has Admiral Media achieved for subscription apps?
Admiral Media rebuilt ChatPDF’s paid acquisition around value based bidding and cadenced creative testing, delivering a 320 percent ROAS increase, 156 percent more subscriptions and a 42 percent CAC reduction, indexed year one against year two year to date. For NeuroNation, Admiral Media delivered a 117 percent ROAS increase, 66 percent more installs, 42 percent higher net cohort revenue and a 39 percent CPI reduction, on data from January to August 2019. For Fastic, the engagement produced a 639 percent increase in installs and a 1,655 percent increase in purchases, and in 2026 the Admiral Media AI Creative Factory cut cost per result by 70 percent.
Should I move my paywall to the web?
Consider it if your proposition is strong enough to be bought before the product is installed. A web paywall gives you deterministic purchase data, which improves value based bidding on iOS where postback measurement is coarse, and it removes store commission from the first transaction. The cost is friction, because users must complete a browser checkout and then install and log in. Admiral Media’s guides to web to app funnels for subscription apps and web to app versus app install ads cover when the trade is worth making.
