The subscription app metrics that matter most are the funnel rates (store conversion, install-to-trial, trial-to-paid), revenue per user (ARPU, ARPPU and LTV) and retention or churn by cohort.
Track all of them, but pick one revenue-based number, usually ARPU over a fixed window or LTV per install, as the goal for A/B tests.
Rates higher up the funnel are easy to move and easy to fool yourself with.
Below, each metric comes with its formula, where to find it, and real tests from our app test database that show how it behaves when someone changes a paywall or an onboarding flow.
The funnel metrics
A subscription app funnel has four steps: store impression, install, trial start (or direct purchase), and first paid renewal.
Each step has its own rate, and each one is worth tracking on its own.
App Store conversion rate
This is the main app store optimization metric.
Apple defines it in App Store Connect as total downloads and pre-orders divided by unique device impressions.
Product page views and first-time downloads sit next to it in the same report.
Our App Store conversion rate benchmarks and ASO guide cover this step in depth.
Install-to-trial rate
Install-to-trial rate = trial starts / installs in the same cohort.
It mostly reflects paywall placement and design. Greg moved its paywall into onboarding for everyone and saw sign-up-to-trial go from 3% to 15%, about +400% trial sign-ups (Greg's test, a bundled change).
Trial-to-paid rate
Trial-to-paid rate = trials that convert to a first paid period / trials started.
Measure it only on cohorts whose trials have fully ended, or the number will look low.
In RevenueCat's State of Subscription Apps 2026, the median trial-to-paid rate was 42.5% for 17-32 day trials and 25.5% for trials under four days. See our trial conversion benchmarks for more.
Download-to-paid rate
Download-to-paid = paying users / installs, usually read at a fixed day such as day 35. It combines the two steps above and is harder to game.
Zumba tested a 7-day trial against a 14-day one and got +8% install-to-paid (Zumba's trial test).
The steps often move in opposite directions. Slopes replaced the paywall with a one-tap trial card and got +25% trial starts while trial-to-paid stayed about the same (Slopes' test).
A wellness app that tried a 14-day trial instead of 7 days got slightly more trial starts but fewer paid conversions (the 14-day trial test, reported by Daphne Tideman).
Revenue metrics (ARPU, ARPPU, LTV)
Revenue metrics are where pricing, plan mix and conversion all come together.
Use proceeds (after store commission and tax) if you can, so web checkout and in-app purchases compare fairly.
Apple's proceeds definition is customer price minus taxes and Apple's commission.
ARPU
ARPU = revenue from a cohort over a window / all users in that cohort, paying or not.
Always state the window (D30, D90). Superwall calls the same idea "proceeds per user."
ARPU is the best single number for most paywall tests because it counts both how many people paid and how much they paid. Our ARPU guide goes deeper.
Price tests show why. BuyBye raised its yearly price by 50% and saw -19% conversions but +25% proceeds per user (BuyBye's price test, vendor-reported by Superwall).
Judged on conversion, it lost. Judged on ARPU, it won.
ARPPU
ARPPU = revenue / paying users only. It tells you how much a payer is worth, not how many people pay, so it rises whenever you push out low-value buyers.
Ryn VPN added a "Continue with Ads" option to its paywall: revenue rose 158%, conversion went from 0.08% to 1.2%, and ARPPU rose 98% (Ryn VPN's test, a before-after comparison reported by Superwall). Read ARPPU next to conversion, never alone.
LTV
Simple LTV = ARPPU per period x expected number of paid periods, or, for a monthly plan, monthly ARPPU / monthly churn rate.
A more honest version is realized LTV: actual cumulative revenue per install at day 90, 180 or 365. Our subscription LTV calculator runs the simple formula.
Worked example: a monthly plan nets $8 after commission and 10% of subscribers cancel each month.
LTV per payer = $8 / 0.10 = $80. If 5% of installs become payers, LTV per install = $4.
LTV catches what first-week numbers miss. Zumba sent paywall taps to a web checkout and saw immediate conversions drop 25% while LTV rose 17% (Zumba's web checkout test).
Teaching.com swapped its trial for a money-back guarantee: conversion dipped slightly, but revenue and LTV went up (Teaching.com's test).
Retention and churn
Mobile app retention metrics come in two kinds, and teams often mix them up.
Usage retention counts who opens the app. Subscriber retention counts who keeps paying.
A subscription app needs both, because usage drops usually show up as cancellations a few weeks later.
Usage retention (D1, D7, D30)
Day-N retention = users active on day N / users who installed on day 0.
Apple's analytics count a session as at least two seconds of use, so check how your own tool defines "active."
See app retention rate benchmarks and our retention rate calculator.
Subscriber churn and renewal rate
Monthly churn = subscribers lost in the month / subscribers at the start of the month.
Renewal rate is its mirror: the share of subscribers due to renew who did.
Split churn by plan length, since a yearly plan only gets a renewal decision once a year.
The churn rate calculator and subscription churn benchmarks help here.
Two things to watch in particular:
- Trial cancellations on day 0. RevenueCat reports that 55.4% of 3-day trial cancellations happen on day 0. Many people cancel right away and keep using the trial, so a trial start is a weak sign of intent.
- Involuntary churn. In the same report, billing failures caused 31% of subscription cancellations on Google Play and 14% on the App Store. That churn is fixed with grace periods and billing retries, not with product changes.
Refunds belong in this section too. BoldVoice tested a hard paywall with no close button and got more trial starts, but refunds rose by more than the gain (BoldVoice's test).
For ways to bring lapsed users back, see our retention guide and the win-back tests.
Which metric to use as the A/B test goal
Use the metric closest to money that you can still measure in a reasonable time.
For most paywall, pricing and onboarding tests, that means ARPU (or proceeds per user) per install over a fixed window, with trial-to-paid and refunds as guardrails.
| Goal metric | Good for | Main risk |
|---|---|---|
| Install-to-trial | Fast reads on paywall layout and placement | Ignores whether trials pay; easy to inflate with aggressive designs |
| Trial-to-paid | Trial length, reminders, trial timeline screens | Shrinks the denominator; fewer but better trials can look like a win |
| ARPU / proceeds per user | Price, plan mix, discounts, most paywall tests | Needs to wait for trials to end; noisier than a rate |
| Realized LTV (D90+) | Web checkout, trial vs no trial, freemium vs hard paywall | Slow; you may wait months for a result |
| Retention / renewal rate | Onboarding, engagement features, plan-length changes | Takes a full billing cycle; hard to power with small apps |
Real tests show why the goal decides the answer:
- Mojo hid credit packs and kept only subscriptions on sale. The goal was revenue, and Mojo saw +14% new subscription revenue (Mojo's test, vendor-reported by Superwall).
- Vahe Baghdasaryan tested a $1.99 paid 30-day trial against a monthly plan with no trial for health and fitness apps. D90 ARPU went from $2.23 to $2.89, +30% (the paid trial test). A trial-start goal could not have shown that.
- Stardust's trial timeline design lifted conversion 13.6% with trial-to-paid unchanged (Stardust's timeline test). With trial-to-paid flat, the extra trials should turn into real revenue.
- Opal replaced its hard paywall with a free tier, and pay penetration fell from 20% to 9% (Opal's change, before-after). RevenueCat's 2026 data points the same way: a median day-35 trial-to-paid rate of 10.7% for hard paywall apps against 2.1% for freemium (source).
Set the goal metric and the read-out window before the test starts.
Picking it afterwards, from whichever number moved, is how teams ship changes that lose money.
Our guides to A/B testing subscription apps and choosing A/B test metrics cover setup and guardrails.
Use the sample size calculator to see whether a revenue goal is realistic at your traffic.
Calculators
These free calculators run the formulas on this page.
Each one asks only for numbers you can pull from App Store Connect, Google Play Console or your subscription platform.
- Subscription LTV calculator: ARPPU, churn and payer rate to LTV per payer and per install.
- Churn rate calculator: subscribers at start and subscribers lost to monthly churn and implied lifetime.
- Retention rate calculator: cohort size and active users to Day-N retention.
- A/B test significance calculator: check whether a conversion difference is real before you trust it.
When you change a metric definition, write it down and keep it fixed across tests.
A 5% ARPU lift means little if the window moved from D30 to D14 between experiments.