AI Cloud Storage · Subscriptions & paywalls
How AI Cloud Storage improved subscriptions through continuous experimentation
Instead of redesigning the paywall on instinct, the team used in-house tooling to test when, where and how the subscription offer appeared — and let user behaviour decide.
The challenge
Cloud storage is a product people value most after they have trusted it with something important. A paywall shown too early asks for commitment before that trust exists; shown too late, the moment of value has passed.
The team needed a way to change paywall timing, layout and offers without shipping a new app version for every idea, and a way to see the full funnel — from install to trial to renewal — in one place.
The in-house stack
AppsKit SDK
Remote-configured paywall placements and variants, so offers, copy and trigger points could be changed and split-tested without app releases.
Pentabit Insights
One view of acquisition source, in-app behaviour and subscription events, used to compare cohorts by variant rather than by averages.
What we tested
Experiment 1
Onboarding offer vs. value-moment offer
Hypothesis: Users who see the offer after their first successful backup will convert with more intent than users who see it during onboarding.
Learning: Timing the offer to a real moment of value changed who converted, not just how many — later-offer cohorts behaved differently through renewal.
Experiment 2
Storage-threshold prompts
Hypothesis: A prompt tied to approaching the free storage limit explains the upgrade better than a generic upsell.
Learning: Contextual prompts read as help rather than interruption, and were easier to judge because the trigger was tied to a clear user need.
Experiment 3
Plan presentation
Hypothesis: Simplifying the plan layout reduces hesitation on the paywall screen.
Learning: Fewer, clearer choices made results easier to interpret and removed noise from later tests.
The outcome
Paywall decisions moved from opinion to evidence: every change shipped as a measurable variant, and winning variants were rolled out remotely.
Because subscription events were tied back to acquisition source in Insights, the team could see which campaigns brought subscribers who stayed — not just users who installed.
Takeaways
- Test the moment, not only the design.
- Judge variants on cohorts through renewal, not on day-one conversion alone.
- Remote configuration turns every release into an experiment platform.