Why don't RevenueCat and Superwall line up on their own?
They answer different questions about the same user. Superwall knows which paywall and variant someone saw; RevenueCat knows what that person paid, renewed, refunded, or cancelled. Until you connect RevenueCat and Superwall data on a shared user ID, neither tool can tell you which paywall produced revenue that stuck.
That gap matters more than it looks. Superwall's experiment view judges variants on what happens near the paywall, while RevenueCat's charts show revenue with no idea which variant earned it. If you're picking paywall winners today, you're probably picking them on the half of the funnel that's easiest to measure. If you're still deciding whether you need both tools, start with RevenueCat vs Superwall: which do you need.
Here's the playbook, ordered by impact rather than setup order:
- Give both SDKs the same user identity.
- Write Superwall experiment and variant context into RevenueCat as subscriber attributes.
- Rank variants on trial-to-paid and revenue per paywall view, not trial starts.
- Control for price differences and sample size before calling a winner.
- Cohort renewals and cancellations by variant.
- Strip involuntary churn out of variant comparisons.
A method note before any numbers: benchmark figures below come from RevenueCat's State of Subscription Apps 2026, which reports medians across 115,000+ apps, $16B+ tracked revenue, and 1B+ transactions, primarily 2025 data. They describe the market, not what your app will do.
Step 1: How do you give both SDKs the same user identity?
Set one stable app user ID and pass it to both RevenueCat and Superwall at the same moment, before the first paywall can present. Every later step depends on this join key, which is why it ranks first.
The failure mode is quiet. If you never identify users, each SDK may track them under its own anonymous identifier, and nothing guarantees those match. Each dashboard looks complete on its own, but a join produces orphans on both sides.
Check RevenueCat's and Superwall's current docs for the exact identify calls and for how each handles anonymous-to-identified aliasing. That behavior is vendor-specific and can change between SDK versions, so don't rely on a blog post's code snippet.
What to change: generate your own ID (a UUID stored in Keychain works if you don't have accounts) and call both identify methods from one function so they can't drift. Avoid email addresses as IDs.
How to verify: make a sandbox purchase, then search for that same ID in both dashboards. If either shows an anonymous ID, your identify call is firing after the paywall.
Step 2: What Superwall context should you log into RevenueCat?
Log the experiment, variant, paywall, and placement as RevenueCat subscriber attributes when the paywall presents. That puts the which-paywall answer on the same record as the what-did-they-pay answer.
Superwall surfaces paywall and experiment information through its events and delegate callbacks; confirm in its docs which fields your SDK version exposes. RevenueCat supports custom subscriber attributes. The names below are a suggested convention, not something either vendor requires:
| Attribute (suggested name) | What goes in it | Why you need it |
|---|---|---|
| sw_experiment_id | Superwall experiment identifier | Groups users into one test |
| sw_variant_id | Assigned variant | The comparison you actually care about |
| sw_paywall_id | Paywall shown | Separates paywall design from experiment |
| sw_placement | Placement or trigger name | Onboarding and feature-gate paywalls behave differently |
| sw_first_seen_at | Timestamp of first presentation | Lets you cohort by exposure date |
Why it works: attributes travel with the subscriber, so later renewals, refunds, and cancellations inherit the variant label without you rebuilding the join every week.
How to verify: open one test subscriber in RevenueCat and confirm all five attributes are populated. Then find a user who saw two paywalls. Decide now whether you keep first-touch or last-touch, and stick with it. For the experiment side of this, see how to set up Superwall paywall experiments.
Step 3: Why rank variants on trial-to-paid instead of trial starts?
Trial starts measure willingness to tap, not willingness to pay. The standard advice, pick the variant with the higher start rate, is wrong whenever variants change who starts the trial, and most variants do.
Here's the mechanism, worked through with made-up numbers for illustration only. Say variant A turns 8% of paywall viewers into trial starts and variant B turns 6%. A looks like the winner on Superwall's view.
Now say, again purely as an illustration, that 30% of A's trials convert to paid while 50% of B's do. Per hundred paywall views, A yields 2.4 paying users and B yields 3.0. The variant that won the visible metric lost the one that pays rent.
That reversal won't happen in every test, but you can't rule it out without joined data. Lower-friction variants, such as a shorter trial or a skippable screen, can plausibly draw less committed users, and only trial-to-paid will show it. So rank variants on paying users or revenue per paywall view; treat starts as a diagnostic, not a verdict.
Trial length is the most common variant, so read its benchmark carefully. Per RevenueCat's State of Subscription Apps 2026, median trial-to-paid is 42.5% for trials of 17–32 days vs 25.5% for trials under 4 days. That's a comparison across different apps, not a controlled test, so treat it as a reason to test trial length rather than proof it'll work for you.
Timing differs too. The same report finds 55.4% of all 3-day-trial cancellations happen on day 0, so a 3-day variant's outcome settles within days, while a month-long trial variant needs weeks before you can read it.
How to verify: pull trial starts, conversions, and revenue from RevenueCat grouped by sw_variant_id, then divide by Superwall's paywall views per variant.
Step 4: How do you handle price tests and sample size?
When variants differ in price, compare revenue per paywall view over the same window, never conversion rate alone. And don't call a winner until every variant's trials have resolved.
A higher-priced variant can lose on conversion and still win on revenue, and an annual-first variant shifts plan mix in ways conversion rate hides. Compare realized revenue per view, and note how much of each variant's revenue is annual so you can revisit it at renewal.
There's no universal minimum sample, but here's a workable rule of thumb for indie apps. Don't read a paywall test until every variant has roughly 100 or more paid conversions, and the test has run at least one full trial length plus about a week for billing retries to settle. Below that, a handful of lucky subscribers can flip the result. Run in whole weeks so weekday and weekend traffic weigh equally. Change one thing per test: if variant B has a new price and a new layout, you won't know which moved revenue. If your traffic can't reach that volume in a reasonable window, test bigger swings, like trial vs no trial or monthly-first vs annual-first, instead of copy tweaks. Small changes need samples most indie apps simply don't have.
How to verify: write the stop rule down before launch, and only check results against it.
Step 5: How do you track renewals and cancellations by variant?
Cohort subscribers by sw_variant_id and first-seen date, then watch cancellations and renewals through the first billing cycles. A variant that converts more but churns faster can lose on a one-year view.
The first month is the early signal. Per RevenueCat's State of Subscription Apps 2026, the first month accounts for 35% of all annual-plan cancellations, so first-month cancellation rate by variant is a reasonable early read on annual plans.
Don't assume a stricter paywall buys better retention. The same report puts year-one retention of yearly subscribers at ≈27% for hard paywalls vs ≈28% for freemium, concluding that the paywall model changes who converts, not who stays. Measure retention per variant instead of inferring it.
How to verify: check whether your RevenueCat charts filter by custom attribute; if not, export subscribers with attributes and pivot in a spreadsheet. RevenueCat cohort analysis explained covers reading those curves.
Step 6: Why strip involuntary churn out of variant comparisons?
Billing failures aren't a verdict on your paywall design. Per RevenueCat's State of Subscription Apps 2026, billing errors cause 14% of subscription cancellations on the App Store across the report's 115,000+ app sample.
If one variant's cohort happens to hit more card failures, it looks worse for reasons unrelated to the paywall. Separate billing-issue cancellations from voluntary ones; check RevenueCat's event docs for how your setup labels them. Compare variants on voluntary churn only.
Route billing-failure users into recovery and a winback campaign, not into your paywall test results. How to verify: if voluntary churn per variant tells a different story than total churn, the cleanup mattered.
Can AppApex merge RevenueCat and Superwall data for you?
Yes. AppApex merges RevenueCat and Superwall subscription data into one view, something single-vendor tools can't do. It integrates with both and isn't affiliated with either.
The first sync pulls roughly 90 days of history in about 30–60 seconds. The first recommendation typically arrives within about 10 minutes.
The Conversion agent reads that merged data for funnel-leak diagnosis and proposes paywall, trial, and onboarding fixes. Each recommendation is tracked from status to applied to measured impact, with per-agent trust scores, so you can see which advice actually held up.
On the Growth and Studio tiers, Autopilot can execute changes through Superwall, RevenueCat, and App Store Connect within the controls you set: kill switches, quiet hours, per-agent budgets, and a full audit log. Starter is advisory only.
Every Autopilot action ships with rollback metadata and a 7–14 day watch window. If lift falls below threshold, the agent rolls it back automatically, and you can roll back manually anytime.
Write access to a provider is never requested until an Autopilot action needs it. Credentials are encrypted AES-256-GCM at rest, and provider reads happen server-side only. AppApex doesn't touch your app's source code, so the in-app identity work in Step 1 stays yours.
What should you do with 30 minutes today?
Run the Step 1 check. Make one sandbox purchase and search for that user ID in both RevenueCat and Superwall.
If the IDs match, spend the rest of the half hour adding sw_variant_id to your paywall presentation handler; that one attribute makes every future test readable. If they don't match, fix identity first, because nothing downstream is trustworthy without it.
Want a read on where your funnel leaks before you start testing? Run the free Growth Audit: enter your App Store ID and get a health score and analysis, no account required.
Last updated October 30, 2026