How an English learning app reached 3M+ downloads and scaled subscriptions faster
RedKiwi teaches English listening and speaking through real-world video clips and has passed three million downloads, with notable traction in Japan and Korea.
Its monetization problem is the familiar one for education apps: converting free learners into paying subscribers and keeping them engaged, which requires continuous testing of subscription lengths, trial offers and paywall designs. This guide walks through the three areas where the team changed its subscription operations — cross-platform entitlement logic, dynamic paywalls, and analytics event forwarding — and gives you a checklist for deciding which of them matters for your own app. All figures below are the case-study numbers as reported, not an independent measurement.
The starting problem: frequent experiments against slow release cycles
An app sold in many countries has to serve different storefronts, currencies and learner behaviors at once. RedKiwi's testing agenda was broad — trial offers, paywall layouts, upsell triggers, pricing — and each test carried the same two costs: a possible wait on app store review, and new eligibility logic written for that campaign alone. The real loss was throughput: fewer tests per month, and slower learning about what converts.
Use this as a diagnostic: if your team can only ship a handful of paywall or offer tests per quarter because each one needs a release, the bottleneck is release coupling, not ideas.
Reported outcomes at a glance
The case study reports three headline results from consolidating subscription work: experiment setup dropped from two weeks to roughly two to three days; engineering effort for paywall and offer changes fell by 30%; and conversion for targeted segments improved by 20%. Treat these as the publisher's reported figures for this customer at the time of writing, not as benchmarks you can expect to reproduce. The mechanism behind them is described in the following sections.
One configuration for both stores
Rather than maintaining parallel subscription code for iOS and Android, the team manages all subscription products from a single dashboard. Three properties do the work:
- A single configuration serves both platforms, so entitlement logic is not written twice.
- Products or tiers already present in the app build can be swapped in a paywall without touching app code.
- Purchases are validated with Apple and Google and access is granted through one entitlement system, so a learner sees the same subscription state on any device.
That last point matters more than it looks: mismatched entitlements and store-specific quirks are the usual source of "I paid but I'm still locked out" support tickets. Centralizing validation removes the class of bug rather than fixing instances of it.
The constraint that trips people up: configuration vs. new build
A common misreading of remote paywall configuration is that anything can change at any time. The actual boundary in this case study is narrower and worth stating plainly:
- Changeable without a release: which already-shipped products or tiers a paywall displays, its layout and copy, localization, and which user segments see it.
- Still requires a new build: adding a product that does not yet exist in the app build.
So the practical rule is to ship a slightly wider catalog of products than you currently promote, then decide remotely which ones to surface. That keeps the fast path open.
Targeted promotions and the paywall experiment set
Paywalls are assembled and edited in a dashboard rather than in code, then published through remote configuration, so an idea can go live without a store resubmission. Currency, language and price adapt to each market on their own, and a paywall can be aimed at a segment defined by behavior, location or predictive score.
That targeting gained a second layer from Monetai, RedKiwi's own machine learning engine, which estimates which learners are unlikely to buy. Those low-likelihood groups can be shown a tailored, localized paywall right away, with no engineering queue in between. The experiments the team ran through the platform included localized discounts in Japan and Korea, machine-learning-driven offers for predicted non-payers, time-sensitive offers to re-engage dormant users, and different pricing tiers for new versus returning users.
With deployment delays and per-campaign eligibility code out of the way, a new promotion went from weeks to a few days — which is what allowed more ideas to be tested and conversion for targeted segments to improve.
Forwarding subscription events into Mixpanel
To see how learners move through the subscription lifecycle, the team connected the subscription platform to Mixpanel through the built-in integration — no custom tracking code. A few dashboard settings enabled automatic forwarding for three events:
- Trial started — a learner begins a free trial.
- Subscription renewed — an active subscription renews successfully.
- Cancellation detected — a learner cancels or turns off auto-renew.
Each event arrives enriched with product identifiers, revenue data and user attributes. That enrichment is what makes the downstream work possible: retention and churn dashboards that show where users drop off, trial-to-paid conversion funnels, and segmentation by plan type, region or purchase channel. Because the pipelines are maintained inside the subscription platform, there is no tracking code to write or keep accurate across both platforms. The case study reports analytics setup time falling from weeks to hours.
Decision checklist: which pieces do you need?
Work through these in order; each one points to a specific section above.
- Do paywall or offer changes currently wait on a store release? If yes, remote configuration (section 3) is your first win.
- Do you maintain separate entitlement code per platform? If yes, a single entitlement system (section 3) removes a recurring bug class.
- Is your product catalog in the build wider than what you promote? If not, ship extra products first, or remote swapping will not help (section 4).
- Are you segmenting offers by predicted likelihood to purchase? If not, that is the highest-leverage experiment type described here (section 5).
- Can you see trial starts, renewals and cancellations in your analytics tool without custom code? If not, event forwarding (section 6) is the gap.
If most answers are "no", the sequence above is also a reasonable implementation order.
| Area | Before | After | Reported change |
|---|---|---|---|
| Experiment setup | 2 weeks | 2–3 days | Faster setup |
| Engineering effort for paywall and offer changes | Baseline | — | Reduced by 30% |
| Conversion for targeted segments | Baseline | — | Improved by 20% |
| Subscription analytics setup | Weeks | Hours | Faster setup |
| Subscription management engineering time | Baseline | — | Half or less of prior time |
These are the figures reported in the original case study at the time it was published. They are not independently verified and should not be treated as expected results or current benchmarks.
| Event | Trigger | Enrichment |
|---|---|---|
| Trial started | A user begins a free trial | Product identifiers, revenue data, user attributes |
| Subscription renewed | An active subscription renews successfully | Product identifiers, revenue data, user attributes |
| Cancellation detected | A user cancels or turns off auto-renew | Product identifiers, revenue data, user attributes |
Event names and enrichment fields reflect the integration as described in the case study; confirm current event schema in the integration documentation before implementing.
Frequently asked questions
Can I change which products appear on a paywall without shipping a new app version?
Yes, provided the products already exist in the app build. The paywall can then be reconfigured remotely to show a different product or tier. Adding a product that has never shipped in the build still requires a release.
What does a single entitlement system actually prevent?
It removes the situation where a purchase is valid on one platform but not recognized on another. Purchases are validated with Apple and Google and access is granted through one entitlement record, so subscription state stays consistent across devices and platforms.
Do I need custom code to send subscription events to Mixpanel?
In this case study, no. The integration was configured in the dashboard and forwarded trial starts, renewals and cancellations automatically, enriched with product, revenue and user data.
What kinds of paywall experiments did the team run?
Localized discounts for Japan and Korea, tailored offers for users predicted to be unlikely to purchase, time-sensitive offers aimed at dormant users, and different pricing tiers for new versus returning users.