The best paywall A/B tests in our archive changed what users pay for and how the price is framed, not the button color.
Stardust's 7-day free trial lifted paywall conversion 131%, Preu AI's added annual plan raised proceeds per user 80%, and a weekly-equivalent price with a discount badge raised Stardust's proceeds per user 62.5%.
Below are 20 app paywall tests ranked, with the exact numbers and who ran them.
How we ranked them: we pulled every app paywall test from the LiftArchive database and kept split tests (A/B or multivariate) with a reported number.
Results measured in revenue (proceeds per user, ARPU, LTV) rank above conversion-only results, and tests with thin details rank lower.
Most of these were published by paywall vendors (Superwall, RevenueCat, Purchasely), so they may lean positive. We note that on each entry.
For the full set, browse all paywall tests, or start with our guide to app paywalls.
More ranked lists live in research, and the app growth hub covers onboarding, pricing and retention.
Ranked table: the best paywall tests at a glance
| Rank | Test | Change | Result | Who ran it |
|---|---|---|---|---|
| 1 | Stardust | 7-day free trial on the $24.99 yearly plan | +131% paywall conversion | Stardust |
| 2 | Preu AI | Annual plan added next to monthly | +80% proceeds per user; +22% trial conversion | Preu AI |
| 3 | Stardust | Discount badge and weekly-equivalent price | +50% conversion; +62.5% proceeds per user; yearly share +30% | Stardust |
| 4 | Unchaind | Paywall that previews premium features | Over +50% conversion; +64% estimated ARPU | Unchaind |
| 5 | Pepper | Two-step paywall with a day-by-day trial timeline | +50% LTV | Pepper |
| 6 | Mojo | Annual plan shown as a monthly price | +45% Brazil, +26% Mexico, +8% USA | Mojo |
| 7 | PropGPT | Weekly price from $6.99 to $9.99 | +47.5% proceeds per user | PropGPT |
| 8 | Speak4Me | Multi-step paywall showing features first | +27% conversion; +36% proceeds | Speak4Me |
| 9 | $1.99 paid 30-day trial instead of monthly | $1.99 paid 30-day trial instead of monthly | D90 ARPU $2.23 → $2.89 (+30%) | Vahe Baghdasaryan |
| 10 | BuyBye | Yearly price 50% higher | -19% conversions; +25% proceeds per user | BuyBye |
| 11 | Rash ID | Three plans with yearly as best offer | About +20% ARPU; over +50% conversion to paid | Rash ID |
| 12 | Fits | Toggle paywall vs timeline paywall | +19% ARPU for the toggle paywall | Fits |
| 13 | Zumba | Paywall button to a web checkout | Immediate conversions -25%; LTV +17% | Zumba |
| 14 | Stardust | Video background paywall | +67% paywall conversion; +12.5% proceeds per user | Stardust |
| 15 | Three-tier cascading paywall offers | Three-tier cascading paywall offers | +8.3% ARPU; about +$14.7K/month proceeds (about $177K/year) | Superwall |
| 16 | TipStop | News-style paywall vs classic paywall | +190% direct conversion | TipStop |
| 17 | Photoroom | Localized upsell paywalls | +50% upsell screen conversion | Photoroom |
| 18 | Blinkist | Trial timeline instead of a feature list on the paywall | +23% trial starts | Blinkist |
| 19 | Pixelcut | Weekly plan instead of monthly on the paywall | +16% paying customers | Pixelcut |
| 20 | Zumba | 7-day free trial instead of 14 days | +8% install-to-paid | Zumba |
The 20 best app paywall A/B tests
1. Stardust: a 7-day free trial on the $24.99 yearly plan
The change: Stardust kept its $24.99 yearly plan and added a 7-day free trial.
It was one of five variants that also tried higher prices with and without trials.
The result: +131% paywall conversion for the $24.99 yearly plan with a 7-day trial vs no trial.
After the trials ended, proceeds per user rose 78% and paying subscribers 45%.
Run by Stardust (multivariate test, published by Superwall). See the test.
Why it likely worked: A trial removes the risk of paying for an app the user hasn't tried yet.
Every trial variant raised conversion over the control, while higher prices without a trial converted worse.
Try it if: your paywall has no trial at all, or you've only tested price and never the trial itself.
2. Preu AI: an annual plan next to the monthly one
The change: The paywall had one monthly plan with a "Try Free" button.
The variant kept it and added an annual plan described as priced at 60% of the monthly price.
The result: +80% proceeds per user and +22% trial conversion rate.
Run by Preu AI (A/B test, published by Superwall).
Exact prices and plan mix were not shared. See the test.
Why it likely worked: The monthly price becomes an anchor that makes the annual plan look cheap, and annual buyers pay more up front.
Try it if: you sell only one plan, especially only monthly.
3. Stardust: a discount badge and a weekly-equivalent price
The change: Same prices ($24.99 yearly, $2.99 weekly).
The variant added a discount percentage on the yearly plan and showed its price as a weekly equivalent.
The result: +50% paywall conversion and +62.5% proceeds per user, driven by more users choosing yearly.
New yearly subscriptions rose 30% as a share of sales.
Run by Stardust (A/B test, published by Superwall). See the test.
Why it likely worked: Users compare plans in the unit they see most.
Putting yearly in weekly terms made the saving obvious without changing the price.
Try it if: you show a weekly or monthly plan next to yearly and only print the full yearly total.
4. Unchaind: a paywall that previews premium features
The change: A text paywall with a headline and bulleted benefits was replaced by a visual one that used rich media to preview the premium features.
The result: Over +50% conversion and +64% estimated ARPU.
Run by Unchaind (A/B test, published by Superwall).
The source calls the ARPU figure an estimate. See the test.
Why it likely worked: Showing the product beats describing it.
A preview answers "what do I actually get?" faster than a bullet list.
Try it if: your paywall is mostly text and your premium features are visual.
5. Pepper: a two-step paywall with a day-by-day trial timeline
The change: Pepper replaced its single-screen paywalls with a two-step flow that included a timeline of what happens on each day of the free trial.
The result: +50% LTV, and it beat every other paywall design Pepper had tested.
The gain held as the rollout grew from 5% to over 50% of users.
Run by Pepper (A/B test, published by Superwall). See the test.
Why it likely worked: A timeline tells the user exactly when they'll be charged, which lowers the fear of a surprise bill.
The two steps also give the pitch room before the price.
Try it if: you've already tuned headlines and highlights on a single-screen paywall and the gains have flattened.
6. Mojo: the annual plan shown as a monthly price
The change: Instead of one yearly price, the annual plan read "$X/month, billed annually".
The result: New revenue per paywall impression rose 45% in Brazil, 26% in Mexico and 8% in the USA, from more paywall-to-trial conversions and a bigger share of annual plans.
Run by Mojo (A/B test, published by RevenueCat). See the test.
Why it likely worked: A smaller per-month number feels more affordable, and that matters most where incomes are lower.
The effect was much smaller in the US.
Try it if: a large share of your users are outside high-income markets.
Check the App Store's pricing display rules first.
7. PropGPT: weekly price raised from $6.99 to $9.99
The change: The paywall's weekly price went from $6.99 to $9.99, a 43% increase, with the same images and copy.
The result: +47.5% proceeds per user. The conversion drop was described as minimal, but no number was shared.
Run by PropGPT (A/B test, published by Superwall). See the test.
Why it likely worked: Many apps underprice. If conversion barely moves, the higher price flows straight to revenue.
Try it if: you set your price once at launch and have never tested above it.
8. Speak4Me: a multi-step paywall that shows features first
The change: A one-screen paywall with a trial checkbox was replaced by a flow that walked users through the app's features, then showed prices and trial options.
The result: +27% conversion and +36% proceeds. Run by Speak4Me (A/B test, published by Superwall).
The whole flow changed, so the lift can't be tied to one screen. See the test.
Why it likely worked: Users who reach the price screen already know what they're buying, so the price is easier to accept.
Try it if: users hit your price before they understand the app.
Note that Stompers saw the opposite result (see Watch out for).
9. A $1.99 paid 30-day trial instead of a monthly plan
The change: Next to an annual plan with a 7-day free trial, the variant replaced a no-trial monthly plan with a $1.99 paid trial for the first 30 days.
The result: Day-90 ARPU went from $2.23 to $2.89 (+30%), and conversion to paying from 10.2% to 12.7%.
Run by Vahe Baghdasaryan (A/B test, pooled across 10 apps with similar price points). See the test.
Why it likely worked: A low paid entry point turns hesitant users into payers, and paying anything filters for people who meant to stay.
Try it if: your monthly plan has few takers and you don't want to give away a long free trial.
10. BuyBye: yearly price raised 50%
The change: Same paywall and trial. Only the yearly price changed, up 50%.
The result: Conversions fell 19%, but proceeds per user rose 25%.
Run by BuyBye (A/B test, published by Superwall).
Long-term renewals were not reported. See the test.
Why it likely worked: Losing a fifth of buyers is fine when each remaining buyer pays half as much again.
This is why conversion rate alone is the wrong metric for price tests.
Try it if: you judge paywall tests on conversion rate only.
Switch the primary metric to proceeds per user.
11. Rash ID: three plans with yearly marked as the best offer
The change: Two plans (monthly, yearly) with no label became three (monthly, yearly, weekly), with yearly marked as the best offer.
The result: About +20% ARPU and over +50% conversion to paid.
Run by Rash ID (A/B test, published by Superwall).
The extra plan and the badge changed together. See the test.
Why it likely worked: A third option gives users a comparison point, and the badge tells them which one to pick.
Try it if: you offer two plans and users seem unsure which to choose.
12. Fits: toggle paywall vs timeline paywall
The change: Fits tested a toggle-style paywall (trial switch, per-period price, "Most Popular" badge, larger annual option) against a Blinkist-style timeline paywall.
The result: +19% ARPU for the toggle paywall.
A follow-up that isolated the toggle found it wasn't the main source of the lift; the visual hierarchy of the plans was.
Run by Fits (A/B test, retold on the RevenueCat blog). See the test.
Why it likely worked: Making the preferred plan bigger and badged steers users toward it.
The trial switch itself mattered less than it looked.
Try it if: your plans look equal on screen.
Check the App Store's current rules before using a trial toggle.
13. Zumba: the paywall button opens a web checkout
The change: The paywall button opened a Stripe web checkout with Apple Pay and Google Pay preselected, instead of native in-app purchase.
The result: Immediate conversions fell 25%, but annual share and retention rose, and LTV rose 17% overall.
Run by Zumba (A/B test, published by RevenueCat). See the test.
Why it likely worked: Web checkout avoids store fees and made annual plans easier to sell, which outweighed fewer instant buyers.
Other apps saw it lose (see Watch out for).
Try it if: you have a strong brand and users who already trust you with a card.
14. Stardust: a video background paywall
The change: On top of the earlier winners (yearly trial and the new price display), the text-heavy paywall got a video background.
The result: +67% paywall conversion and +12.5% proceeds per user.
The same setup without video gave +55.7% conversion and +9% proceeds per user, so the video added the extra lift.
Run by Stardust (multivariate test, published by Superwall). See the test.
Why it likely worked: Motion catches attention and shows the app's feel.
Stardust also stacked it on proven winners, a good way to compound gains.
Try it if: your app has a visual product and your paywall is a wall of text.
15. Three-tier cascading paywall offers
SuperwallThe change: Instead of two paywalls by view count, users saw the standard paywall on view one, an extended-trial paywall on views two and three, and a discount paywall from view four on.
The result: +8.3% ARPU, about +$14.7K per month in proceeds (about $177K annualized).
Run by Superwall (A/B test on an anonymized app). See the test.
Why it likely worked: Users who keep coming back without buying are interested but not convinced.
A better offer each time meets them where they are, while first-time viewers still see full price.
Try it if: you show the same paywall every time, no matter how often a user has dismissed it.
16. TipStop: a news-style paywall vs the classic one
The change: TipStop, a sports betting tips app, tested a paywall in the style of the New York Times against its classic paywall. Neither design was described.
The result: +190% direct conversion. Run by TipStop (A/B test, published by Purchasely).
Sample size, duration and revenue were not shared, which is why it ranks this low despite the size. See the test.
Why it likely worked: We can't say for sure.
A big lift with no details can also come from a weak baseline, so treat it as a lead.
Try it if: you run a content app and want a bold layout test rather than a copy tweak.
17. Photoroom: localized upsell paywalls
The change: Upsell paywalls were localized for each user's market, then tested and refined over several rounds.
The result: +50% upsell screen conversion. Run by Photoroom (A/B test, published by RevenueCat).
The figure may come from several rounds, not one split test. See the test.
Why it likely worked: A paywall in the user's own language, with local framing, reads as made for them.
Try it if: a big share of installs come from non-English markets and your paywall is English-only.
18. Blinkist: a trial timeline instead of a feature list
The change: Blinkist dropped features and benefits from the paywall entirely.
It walked through the trial as a timeline and promised a reminder before the charge.
The result: +23% trial starts. Push opt-in rose from 6% to 74% and complaints fell 55%, though those may be from after launch.
Run by Blinkist (A/B test, case study by Growth.Design). See the test.
Why it likely worked: Fear of forgetting to cancel stops people from starting trials. Promising a reminder removes that fear.
Try it if: you have a free trial and users complain about surprise charges.
19. Pixelcut: a weekly plan instead of monthly
The change: The paywall offered a weekly subscription in place of the monthly one.
The result: +16% paying customers. Pixelcut also judged weekly better on LTV but didn't share those figures.
Run by Pixelcut (A/B test, published by RevenueCat). See the test.
Why it likely worked: A weekly price is a small number and a low commitment, which suits apps people use for a short job.
Try it if: your app solves a one-off or short-term task, like editing a few photos.
20. Zumba: a 7-day trial instead of 14 days
The change: New users got a 7-day free trial instead of a 14-day one. Nothing else changed.
The result: +8% install-to-paid conversion. Run by Zumba (A/B test, published by RevenueCat).
Retention and LTV effects were not reported. See the test.
Why it likely worked: A shorter trial reaches the payment moment while interest is still high, and fewer users forget about the app.
Try it if: your trial is long and many users go quiet before it ends.
Daphne Tideman saw a 14-day trial lose to 7 days too.
What the winners have in common
They change the offer, not the decoration
Most of the top ten changed the trial, the plans or the price.
Stardust's trial, Preu AI's annual plan and PropGPT's price rise all beat design tweaks. See free trial tests and subscription pricing.
They frame the yearly plan in small units
Stardust (weekly equivalent), Mojo ("per month, billed annually") and Fits (bigger, badged annual plan) all made yearly look like the easy choice. More in plan tests.
They measure revenue, not just conversion
BuyBye lost 19% of conversions and still won on proceeds.
Zumba's web checkout cut instant conversions 25% and still raised LTV 17%.
A conversion-only scorecard would have killed both winners.
They remove the fear of being charged
Pepper's and Blinkist's timelines tell users exactly when billing starts.
That honesty raised trial starts and LTV. Compare layouts in our app paywall examples and paywall design tests.
What to test next
- Trial vs no trial. If you've never tested it, this is the biggest lever on this list.
- Add an anchor plan. Put annual next to monthly, or weekly next to yearly, and mark one as the best offer.
- Reframe the yearly price. Show it per week or per month with the billing stated.
- Raise the price. Test 30-50% above your current price and judge it on proceeds per user.
- Change offers by view count. Show a better offer to users who have dismissed the paywall several times.
- Test timing. Where the paywall appears in onboarding matters as much as what it says. See paywall timing tests and hard vs soft paywalls.
Use our paywall checklist before you launch, and compare paywall tools if you need a way to run tests without an app release.
Watch out for
Trapping users. BoldVoice tested a hard paywall with no close button and got more trial starts, but refunds rose by more than the gain.
Its unclosable discount offer did the same. Count refunds and cancellations in your result. See the BoldVoice test.
Copying a winner blindly. Speak4Me's multi-step paywall won, but Stompers' multi-page flow lost to a single page by over 16% on conversion.
Stardust's yearly-only paywall lost, while Mojo's yearly-only default won.
Results depend on the app and audience, so retest on yours.
Web checkout everywhere. Zumba won with it, but Dipsea's web-only paywall dropped initial conversion from 27.0% to 18.1%. Read our web-to-app guide before you switch.
Vendor bias and missing sample sizes. Almost none of these posts share sample size or test length.
Treat the numbers as direction, not a forecast.
Our app A/B test win rate page shows how often app tests actually win.
Paywall tests that lost or backfired:
RevenueCatWant help planning and reading paywall tests for your own app?
Oliver Kenyon runs experimentation programs and can help you pick the right tests and metrics.