Forty percent of people who install one astrology app from an ad end up paying, and its founder ties that number to more than a million dollars of A/B tests. That figure is worth taking seriously and not copying. It is one founder's account of one app, and the tests that lost are as instructive as the ones that won. Start with a plain paywall, then test design before price The first paywall and the order of tests The designer behind a set of 4,700 paywalls begins with a deflating admission: paywall optimization is "as a starting point, honestly not that important." The first paywall needs very little, namely a headline that describes the product, a couple of bullet points, and a continue button. The highest-converting paywall is not automatically the one to keep, so early effort belongs on getting users into the app. Once a paywall exists, the order of work matters. Design tests come first. Teams should try radically different layouts, including video, bullet lists, trial timelines, and long-form pages, until one of them clearly beats the rest. Personalization comes second: calling out the user's name and goals, or showing a strong man on a fitness paywall for a male user so the screen reflects the buyer's identity. Price packaging comes after that, covering whether yearly and weekly products sit in vertical or horizontal groups and how prices are displayed. Device-based pricing, with higher prices for newer phones, is last because the work is heavy and the returns shrink. Judge a win by the business it funds ARPU can flatter a weekly product ARPU alone misleads. The same speaker once used it as the North Star metric and then stopped. A three or four dollar ARPU can come from weekly products whose monthly churn runs 90 to 95 percent, so the number looks healthy while hiding how quickly subscribers leave. The working set is paywall percentage, trial-to-paid conversion, LTV, D1, D7, D30, and churn. Paywall percentage is read as a benchmark: over 80 percent of installers seeing the paywall is good, and 90 percent or more is strong. The metric choice depends on the business. Teams chasing quick revenue should weight ARPU, while teams building for the long term should weight LTV and retention and keep churn in view. Patience, touchpoints, and the after-onboarding slot Wins are not guaranteed. The agency's method uses the existing paywall as a control against four or five new designs. It beat the control in week one for every client except two, and for those two the statistically significant result took until the second or third experiment. That is a useful limit: a single failed test does not show that a paywall approach is wrong. Touchpoints matter as much as the screen. The speaker lists onboarding paywalls, gated feature paywalls, settings-page paywalls, cancel-flow offers, and a delete-time offer. The after-onboarding slot is the most used, and the speaker cites 78 percent of iOS apps on Mobbin showing a paywall there. Whether the account should be created before or after the paywall is less settled. The speaker prefers account-first so a user ID ties to the subscriber, but the video does not test that claim against conversion. Where pricing tests and aggressive offers disagree Pricing is where the founder's experiments were largest. Moonly ran 40 pricing experiments in a year, and the speaker calls pricing tests "the most important experiment you can run in a product." A high-priced lifetime plan anchors users toward the plan you want them to buy. On Mobin, only 4 percent of iOS apps offer a lifetime plan, usually priced at about twice the annual plan. The founder also reports keeping the same number but changing the unit, from $8 per month to $8 per week, which brought money in weekly and shortened the feedback loop on ad spend. Free trials were the sharpest reversal. The founder says trial users who received the birth chart free felt they had already taken about 90 percent of the value, and that 59 percent of iOS paywall screens on Mobin offer no trial at all. Moonly dropped trials and used a one-time offer around day three, which the founder credits with about a 15 percent revenue lift between day three and day 30. Billing mattered too. On iOS, native payments led to 50 percent of cancellations within three minutes of purchase. Adding Stripe as an alternative in the US doubled LTV, and about 80 percent of US customers now pay through Stripe. The founder advises against Android alternative billing through Google's SDK, because the developer does not control the interface and Google Pay is not supported. The limits are clear. These numbers come from the founder's own app and are not independently audited, and the material does not show sample sizes, test durations, or significance thresholds. On aggressive tactics, the speaker disagrees with the predetermined spin-the-wheel paywall, which can work for weekly-price revenue but suits a long-term business poorly. The speaker also warns that an 80 percent-off offer shown right after onboarding can cannibalize revenue before the user has seen the paywall a few more times. Steep discounts are presented as fine for win-backs, where a subscriber is already considering cancelling. Ship a simple first paywall and test radically different designs before optimizing personalization or price packaging. Read a paywall result alongside ARPU, churn, LTV, and D1, D7, and D30 retention, because a high ARPU on weekly products can hide 90 to 95 percent monthly churn. Treat a single clean win as a hypothesis: some apps needed a second or third experiment before the result was statistically significant. Test a lifetime or annual anchor and a unit change, such as $8 per week instead of $8 per month, against your own conversion data. Use discounts for win-back offers to subscribers considering cancelling, and be cautious about immediate deep discounts after onboarding. The practical promise of this chapter is that you can now plan a paywall test sequence, interpret a win against the metrics that match your business, and decide whether a result is trustworthy enough to ship. Start with the design tests, hold each winner to churn and retention, and treat the founder's largest claims as a reason to run your own experiments.