I Made 4,000 App Paywalls and Learned This (video)

Joseph Choy (Consumer Club founder, now at Superwall) interviews Jonathan Parra, the lead designer at Superwall who has hand-built 4,500+ paywalls across the platform’s customer base — more than anyone on the planet. The episode walks through a deck Parra normally presents at Superwall’s office hours: a sequence of A/B-tested paywall before/afters with the design reasoning behind each lift. It is the most data-dense paywall-craft session in the wiki.

The headline finding, which Parra frames as the one that changed his philosophy:

I really thought we would be able to create a set number of golden paywalls that you can use everywhere and they’ll perform about the same. … One paywall in an app in the same app category, you would expect the performance to be around the same and it could just flounder completely in one app versus another. — Jonathan Parra

There is no perfect paywall. There is, however, a baseline that ports well — a single-page bullet-list design — and a small set of consistent additive moves (“no commitment, cancel anytime”; chevron arrows; large 65pt CTAs) that lift conversion almost everywhere.

The thesis

Three claims do the load-bearing work:

1. No universal winner. Same paywall, same app category, opposite outcomes. The space is local, not global.
2. Design and packaging are bigger levers than pricing. Founders obsess over price testing first; the wins are downstream of layout, copy, hierarchy, and trial framing. Price testing also creates cohort-management overhead that compounds the further you push it.
3. Less beats more. The recurring failure mode is paywalls that try to inform the user. Users don’t read. Comparison tables, descriptive CTAs, and multi-product selectors all underperform stripped variants in repeated tests.

The bullet-list baseline

Bullet-list paywalls actually are one of the few paywalls that perform well across the board. Just single-page bullet-list paywall. That’s a benchmark. — Jonathan Parra

The canonical baseline (what Parra builds first for new customers):

• A heading with a clear USP and short subtitle
• A short bulleted list of what the user gets (no table, no comparison)
• A social-proof row (stars / accolades / reviews) if the app has the install base to justify it
• A subtitle: “no commitment, cancel anytime”
• A single large CTA — 65pt button height (56pt minimum), generic Continue copy, optional right-chevron arrow
• Annual selected by default

This is the design Parra uses as the null hypothesis for every customer experiment. Subsequent tests add one thing at a time (image, video, multi-page, social proof page, free-trial timeline) and compete against this baseline.

A/B case 1 — Simplify the picker (+10%)

Two paywalls for an “AI assistant” app:

Control	Variant
”AI Assistant Weekly / AI Assistant Annual / AI Assistant Monthly” naming	”Weekly / Annual / Monthly”
Limited-time 70% off in small text	70% off, centered, no “limited time”
Descriptive CTA: Put AI into action	Generic CTA: Continue
(no subtitle)	“No commitment, cancel anytime”

Result: +10% conversion. Every change is a subtraction. The single counter-intuitive move is removing limited time — Parra’s read: scarcity copy at small font sizes doesn’t fire; if you want urgency, it has to be in a heading on a one-time-offer paywall, not as a 12-point badge subtitle. See Scarcity principle.

A/B case 2 — Hide the picker, surface the trial

Two paywalls, same app:

Control	Variant
Three products visible (weekly / annual / monthly)	Two products (annual / weekly), annual still selected
CTA: Put AI into action	CTA: Start my 3-day free trial

The variant moves the trial framing onto the button itself rather than burying it in the product row. Parra’s read:

• A three-product picker creates decision fatigue — see Cognitive load. The cleaner move is to default to the LTV-maximizing plan (annual) and hide the rest behind a view all plans drawer.
• Trial copy on the button transfers the safety signal from the product row to the user’s eye-line at the moment they’re deciding.
• Plan order matters: if you do show multiple products, order them by length cadence (longest-to-shortest or shortest-to-longest). A jumbled weekly/annual/monthly row is harder to parse and converts worse.

A/B case 3 — The “we want you to try” reframe (+111%)

Two paywalls, the most dramatic delta in the deck:

Control	Variant
Before/after comparison table (Free vs Premium)	Single big phone screenshot
Three-product selector (3 / 6 / 12 months)	(no selector on this screen)
CTA: Continue	Headline: “We want you to try [app] for 50% off” + Continue

Result: +111% in favor of the stripped variant. The continue button on the variant leads to the actual product/price screen — the simpler paywall is doing the pre-suasion work; the real choice is one step later.

Parra’s read traces back to the cognitive-load thesis: comparison tables look like they help the user but in practice they overwhelm. The variant’s win is partly the simpler layout, partly the gift-framed copy (“we want you to try [app]”) — see Gift vs receipt for the broader principle that ceremonial reveal beats flat transaction.

Inspiration credit: Cal AI (“Cali”), which Parra had observed running similar paywall structures.

The Clear-30 “gray hat” cascade

Parra now defaults to a multi-screen drawer cascade for apps that have already done a lot of testing. Inspired by Clear 30 (the addiction-recovery app the format is named after):

Screen 1 (main paywall)
  Heading + USP + steps + "no commitment cancel anytime"
  Annual selected at default
  CTA: Continue
  ↓ user closes ↓
Screen 2 (drawer rises)
  "Not ready to commit for a year?  We have plans for everyone."
  Annual STILL selected (LTV optimization)
  Monthly added below
  ↓ user closes ↓
Screen 3 (one-time offer)
  "We want you to try [app] for free" (or 25–33% off)
  Same yearly price reframed — $39/year → 76¢/week
  Discounted yearly: $39 → $29.99

Three principles stack here:

1. Default-and-defend — annual stays the selection through every cascade; only the framing around it changes. The system optimizes for LTV without ever asking the user to pick.
2. Per-week reframe — the same $39 annual price is shown as 76¢/week at the second close. See Framing effect; the reframing doesn’t change the price, it changes how the user encodes it.
3. Discount discipline — Parra deliberately keeps in-flow discounts at 25–33%. Reserving 50%+ discounts for Black Friday / Cyber Monday avoids cheapening the brand and prevents user habituation to “open paywall, close, get cheap deal.”

The “gray hat” framing is Parra’s own. The cascade isn’t deceptive; it’s structurally honest about giving the user three off-ramps. But the design is unapologetically optimized to extract the highest-LTV outcome at each step.

The Swift UI surprise

I was like, “there’s no way this is going to work.” There’s no way it’s going to do good. And the SwiftUI style actually outperformed everything by quite a long shot. — Jonathan Parra

Parra built native-Swift-UI-looking paywalls for Plant Identifier (Pyometer Plus+) because the rest of the app already used Apple’s default design system. He expected custom variants to crush. They lost.

Parra has now reproduced the result on a second app. His read: when the rest of the app reads as native, a custom paywall registers as foreign — a visible context shift that breaks the user’s trust the same way an unfamiliar checkout flow does on the web. This is the Familiarity principle at the paywall surface: engineered familiarity beats engineered novelty when the broader app expects it.

Implication: paywalls are not a generic surface. The right paywall is the one that doesn’t break the design language of the app. Custom-design wins exist; they live in apps whose broader UX is already custom.

The premium contrast play (Hinge-inspired)

For multi-tier apps (Pro and Plus subscriptions), Parra deliberately runs opposite backgrounds on the two tiers:

• Plus plan paywall: white background
• Pro plan paywall: black background (fade contrast)

When the user toggles between tiers, the visual register changes — the higher tier feels qualitatively different, not just numerically more expensive. He cites Hinge as inspiration, and notes the result is more Pro subscriptions even when the underlying offer differences are modest. The Robinhood Gold credit card (black + gold inside a gold-themed installer app) is held up as an extreme case of the same principle.

The mechanism crosses categories: framing-via-color rather than framing-via-copy. The same encoding work that “guest pass” does for Headspace, the white/black contrast does here.

The video-paywall pattern (Mojo-inspired)

A high-performing format for visual-product apps:

• Bullets at top
• Phone-in-hand video demo (Parra uses Rotato for the phone-mockup rendering)
• Social proof row below
• Annual/Monthly product selector
• CTA: Start my 7-day free trial with chevron arrow

The video gets users a clear value prop without making them read. The format reads as a hybrid of paywall and TikTok-style product demo.

The free-trial timeline paywall (Blinkist-style)

Parra calls out the Blinkist day-by-day timeline paywall as a years-old format that still tests well:

• Today → unlock the app
• Day 5 → email/notification reminder
• Day 7 → billing begins

The format works because it visually previews the commitment as a guided journey rather than as an open-ended subscription. See Progressive disclosure — the same principle that drove Blinkist’s original +23% trial signup lift per Copy These SaaS Growth Tricks (video). The timeline is most effective for apps with a multi-day or multi-week transformation arc (recovery, language learning, fitness, mental health).

The pricing landscape (Feb 2026)

Several market-state observations from a designer with a wide vantage:

• AI wrapper apps have settled at $29.99–40/year. That’s the floor the market has consolidated on for text-based AI products; image/video apps charge more to cover inference. Parra describes it as the price “where you can still be profitable and have good margins and it’s still fairly affordable.”
• Weekly subscriptions are surging for AI rapper apps. High churn but users forget to cancel, so the weekly cohort often pays more than they would have on an annual plan. Useful for price-testing without locking users into a year. Social/multiplayer apps still skew weekly+monthly because the social value compounds at higher frequency.
• Two-tier billing (Pro + Plus, à la ChatGPT) is rising, especially in image/video AI apps where compute costs justify it. Parra notes the cohort-management cost: each tier is its own audience, each audience is its own experiment surface.
• Friends-and-family pricing is underused. Annual shareable plans (iCloud Family integration) bump LTV/ARPU; almost no consumer apps deploy them outside Apple’s own ecosystem and Spotify.

The custom-prompt-as-product thesis

Parra has watched apps charge $20–22/month — ChatGPT-level pricing — without the brand pull of OpenAI, and the model that works is vertical apps with a tuned prompt and a tight end-to-end flow:

The big bottleneck is still in the prompt. If you have an app that solves a very specific issue, you can get away with charging a fraction of what you charge with ChatGPT. You’re essentially paying for a custom n8n flow with good prompts so that you don’t really have to think about doing all this. You just have it work. — Jonathan Parra

Categories where the play works at premium prices:

• Insecurity-targeted products — Looks-Max apps charging multiples of the AI-wrapper baseline.
• Dating — text-message-screenshot → tuned reply apps.
• Health — vertical health apps still earning consistent revenue years after launch despite ChatGPT existing.

The closed-loop transformation app is the consistent winning archetype: the user can describe who they were before they started using the app and who they are after. Conversion rates for paywalls on closed-loop apps run materially higher than for generic-utility AI wrappers.

What founders should test first (and last)

Parra’s order of operations for a new customer:

1. Baseline. Ship a single-page bullet-list paywall. Use it as the null hypothesis.
2. Design tests. Image vs video vs static, copy variants, plan order, CTA framing, multi-page splits.
3. Multi-screen architecture. Drawer cascades, premium-tier contrast, free-trial timelines.
4. Price testing — last. Cohort management is painful in Superwall (user seeds, audience management). Save it until design tests have plateaued.

The advice flatly contradicts the founder default of “let me A/B test $9.99 vs $14.99 first.” The design lever is bigger and cheaper to move.

Where the gains come from

App stage	Typical lift
Large app (~$100k MRR) that hasn’t A/B tested	Immediate compounding wins on the first round
Newer app ($10–50k MRR)	8% conversion → 15–20% after a couple of iterations
Mature app already deep in testing	Smaller marginal lifts, but the gray-hat cascade and Swift UI variants often still find +10–30% pockets

The biggest delta is the apps that have been investing in product but not in paywall craft. The crew that “haven’t done a lot of AB testing because they’ve been working so much on product” is where the largest immediate gains live.

Connections to other wiki pages

• Jonathan Parra — guest entity, lead designer at Superwall
• Superwall — paywall A/B testing platform, publisher of this podcast, source of the data
• Joseph Choy — host, founder of Consumer Club
• Cognitive load — the recurring engine under “less beats more”; decision fatigue from multi-product pickers; tables as friction
• Framing effect — per-week price reframe ($39/yr → 76¢/week); “we want you to try [app] for 50% off” copy; the gift-framed CTA
• Anchoring bias — annual-as-default selection; longest-to-shortest plan order; high-tier-first contrast play
• Familiarity principle — Swift UI native-looking paywalls beat custom variants (Pyometer Plus and second-app reproduction)
• Loss aversion — the close-drawer cascade’s escalating off-ramps; one-time-offer drawer
• Decoy effect — inverse application: hiding the third product behind “view all plans” wins
• Scarcity principle — small-font “limited time” copy fails; scarcity has to live in a heading on a one-time-offer page
• Social proof — accolades placement on multi-page paywalls
• Progressive disclosure — the Blinkist day-by-day timeline paywall as the canonical example
• Gift vs receipt — the “we want you to try [app]” framing as gift-coded copy
• Tim Gabe — adjacent author; this video reproduces several of his SaaS-conversion findings from independent data
• Copy These SaaS Growth Tricks (video) — Tim Gabe’s earlier video on the same surface; many of the same biases recur in different products

Notable products mentioned (not promoted to entity pages)

Single-source mentions, flagged here for cross-reference if they recur:

• Clear 30 — addiction-recovery app; inspiration for the gray-hat drawer cascade
• Mojo — video-creation app; the video+bullets+social-proof paywall template
• Cal AI (“Cali”) — calorie-tracking AI app; the “we want you to try for 50% off” reframe
• Pyometer Plus+ — plant-ID app; the Swift UI native paywall case
• Hinge — inspired the premium-tier color contrast
• Robinhood Gold — extreme premium-color-signaling case
• Blinkist — the day-by-day free-trial timeline paywall
• Rotato — phone-mockup rendering tool for video paywalls

Glossary

• ARPU / ARPO — average revenue per user / per opportunity
• LTV — lifetime value
• CTA — call to action (the button)
• USP — unique selling proposition
• Hard paywall — paywall the user cannot dismiss without subscribing; soft paywall — dismissible
• One-time offer (OTO) paywall — final-step discount paywall shown only on close intent
• Closed-loop app — an app whose users can describe a before/after state from using it (Clear 30, Peloton, Duolingo); the highest-converting archetype

Sources

• raw/paywall-design-superwall.txt — auto-captioned transcript (yt-dlp, 2026-05-29)
• Original: https://www.youtube.com/watch?v=lkwX_kc0NS8