# Onboarding and Paywall Best Practice Master Doc

Everything we know about the App Store listing, onboarding and the paywall of a consumer
subscription app, in one file. It is built from teardowns of 89 onboarding flows and 52 paywalls
across health, wellness, fitness, sports, finance, kids and education, the 2026 industry reports
from RevenueCat and Adapty, the published experiment results we found worth trusting, and the
behavioural research that explains why the patterns work.

**Who it is for.** A growth or product lead who wants a rigorous second read on their own store
listing, onboarding flow or paywall, and does not want to wait for one. It is written to be
handed to an LLM alongside screenshots.

**How to use it.** Screenshot the flow in order, from the store listing through the paywall and
one screen past it. Fill in the profile lines in the prompt below, give an LLM this file and the
screenshots, and run the prompt. Naming the specific pattern behind every call is what turns a
vague answer into a usable one.

**One warning before you start.** Most of what follows is a pattern that worked somewhere, and
each entry says how strongly it is evidenced. A pattern is not a prescription. The apps that
convert best are not the ones that implemented the most items here; they are the ones whose
every screen has a reason to exist.

---

## What is in here

- [The flow, and what each step is for](#the-flow-and-what-each-step-is-for)
- [The App Store listing](#the-app-store-listing)
- [Onboarding patterns](#onboarding-patterns)
- [What changes by vertical](#what-changes-by-vertical)
- [Paywall and monetisation configurations](#paywall-and-monetisation-configurations)
- [Onboarding and first-session patterns with measured results](#onboarding-and-first-session-patterns-with-measured-results)
- [The psychology behind the patterns](#the-psychology-behind-the-patterns)
- [Pricing and packaging, field notes](#pricing-and-packaging-field-notes)
- [Paywall redesign case studies](#paywall-redesign-case-studies)
- [What the teardowns found](#what-the-teardowns-found)
- [Named examples, principle by principle](#named-examples-principle-by-principle)
- [Benchmarks](#benchmarks)
- [Evidence from outside the industry, and the pattern it supports](#evidence-from-outside-the-industry-and-the-pattern-it-supports)
- [Numbers to stop quoting](#numbers-to-stop-quoting)
- [The judgment calls an audit has to make](#the-judgment-calls-an-audit-has-to-make)
- [Three things that are easy to get wrong](#three-things-that-are-easy-to-get-wrong)
- [How to read the evidence in this document](#how-to-read-the-evidence-in-this-document)

---

## The prompt

> You are auditing a consumer subscription app's store listing, onboarding and paywall. The
> attached file is your reference. The images are the flow, in order.
>
> **The app.** Category: ____. Access model: hard paywall (no free path) or freemium (a free path
> exists): ____. Paywall placement: in onboarding or in-app: ____. Trial: none / length ____.
> AI-powered: yes/no. Platform: iOS / Android / web. Primary plan shown: weekly / monthly /
> annual. Any funnel numbers you have (install to trial, trial to paid, install to paid): ____.
>
> Work in three passes and do not skip the first.
>
> **Pass 1, describe.** For each screen say what it asks, what it promises, what it shows, and
> where it sits in the sequence. Quote the actual copy. Count the questions and the asks
> (fields, permissions, account, payment). Do not evaluate anything yet.
>
> **Pass 2, assess.** Place the app on the profile axes in the Benchmarks section first, so
> every number you use is the matching row and never the cross-vertical median. Benchmarks are
> cut by store category, so a wellness, sleep or meditation app reads the Health and Fitness
> rows, and a journaling or habit app usually reads Lifestyle or Productivity. Then go through
> the patterns in this reference that apply to this app. For each one say whether the flow does
> it, name the screen that decided the answer (or say it is a property of the whole flow, such
> as reason-giving across every question), and give the one change that would fix it.
> If a pattern does not apply to this app, say so and say why instead of forcing a verdict.
>
> **Pass 3, return.** In this order:
> 1. Setup. Confirm the flow is evaluable from what was supplied, confirm or challenge the stated
>    category and access model, and name the metric the audit should be measured against: by
>    default install-to-trial and trial-to-paid when there is a trial, install-to-paid when there
>    is not, always read against the app's own profile rows.
> 2. Scorecard. Score the listing, the onboarding and the paywall out of 5 against this app's
>    profile, one line of justification each.
> 3. Gaps. A table ranked by impact against effort, highest leverage first, each row naming the
>    pattern from this reference that supports it and how strong that evidence is.
> 4. Recommended flow. The ordered steps for this app, each marked required or optional, with
>    current steps flagged to cut, move or merge.
> 5. Changes as testable hypotheses. One row per change: "if we do X, metric Y moves because Z",
>    the metric to watch, the guardrail metric that would show it backfiring, and how to test it.
> 6. Ship first. The single highest impact-to-effort change in one sentence, with the metric
>    that confirms it.
> 7. Next loop. What to re-measure after shipping, so the audit becomes a loop.
>
> Rules. Be specific to these screens, never generic. Never name a step, a field or a prompt
> the screenshots do not show. If something is missing entirely, say it is missing rather than
> saying it could be improved. Every number you quote comes from this reference with its
> publisher, or from the app's own data; do not invent one, and check what each number is a
> number of before you use it. If the evidence behind a recommendation is weak, say so in the
> row. Where two figures in this reference differ for the same metric, the Benchmarks section
> is the one to use, and check "Numbers to stop quoting" before you cite anything. When you
> are challenged on a claim, recount from the screens before defending it.

---

## The flow, and what each step is for

Every screen answers a question the user is already asking. A screen that answers a question
nobody asked is the screen they leave on.

| Step | The question in their head | What the step is for |
|---|---|---|
| The ad | Worth a tap? | Getting the click. Not explaining the app. |
| The store listing | Worth downloading? | About seven seconds on the page, and about nine in ten people never scroll past the third screenshot. The first three screenshots are the listing. |
| Onboarding | What is this, and is it worth it for me? | They already know a trial is coming. Every screen has to earn the next one. |
| The paywall | Do I understand it, and is it worth it to me? | Both have to be true before anyone starts a trial. |
| The screen after | Was that a good decision? | Confirming the choice is part of the purchase, not a separate problem. |

Commitment escalates at every rung and so does drop-off. The strongest flows are not the
shortest; they are the ones where every ask carries a reason. Onboarding is not there to teach
how the product works. It is there to make clear what the product is and why this user should
care, and to let the product teach itself once they are in it.

---

## The App Store listing

The listing is where the funnel starts and the part of the flow most teams never audit. It has two jobs, being found and converting the visit, and a single test for the second: every screenshot has to land one message in half a second, and the caption has to sell the value rather than describe the feature. What follows is the platform mechanics, the copy, the visuals, the keyword and rating levers, and named examples, each with its evidence.

### What changed in 2025-2026

Two platform-level changes redefined ASO strategy for subscription apps in 2025, and both remain in early adoption. **Apple's CPP Keyword Linking** (launched July 30, 2025) lets developers assign keywords from their iOS keyword field to Custom Product Pages, so that the right CPP surfaces in organic search, not just paid. The catch: keywords from the title and subtitle are ineligible, forcing a strategic rebalance of where you place your most valuable terms. The feature is US-only at launch (with limited UK availability), assignments depend on the latest approved app version, and Apple Search Ads takes precedence when both paid and organic CPPs target the same keyword. Apple increased the CPP cap from 35 to **70 per app** in October 2025. No published case study yet connects organic CPP impressions to subscription conversion or LTV, the feature is too new and Apple's privacy framework limits post-install organic attribution.

On the Android side, **Google Play Guided Search** went live September 23, 2025, powered by Gemini. Users now type goals ("learn Spanish," "get fit at home") instead of app names, and the store organizes results into intent-based refinement paths. This shifts optimization from keyword density to **semantic relevance and outcome alignment**. Subscription apps must rewrite short descriptions and screenshot captions around jobs-to-be-done, not feature lists. Alongside Guided Search, Google introduced the **"You" tab**, a personalized hub that surfaces subscription management front-and-center, creating elevated churn risk. RevenueCat warns that win-back flows, pause options, and downgrade paths are now non-negotiable. Google reported a **2% decrease in voluntary churn** from surfacing subscription benefits more prominently, and extended default account hold to 60 days (December 2025), reducing involuntary churn by **10%**.

The net effect: Apple now rewards intent-matched creative at the keyword level, while Google rewards intent-matched copy at the semantic level. Both platforms are moving toward a world where **relevance to specific user motivation**, not keyword volume, determines visibility and conversion. Practitioners at AppTweak, Phiture, and RevenueCat converge on one directive: map 6-8 user jobs your app solves, build distinct creative/copy treatments for each, and distribute them across CPPs (iOS) and Custom Store Listings (Android).

**Known gap:** No public source currently connects keyword-level organic installs to subscription LTV. CPP Keyword Linking is too new for published downstream subscription impact data. MobileAction and Airbridge discuss the theoretical connection, but no quantitative evidence exists as of April 2026.

---
### Platform-specific: Apple App Store

#### Apple CPP Keyword Linking

*iOS*

**What it is.** Developers assign keywords from their iOS keyword field (100-character backend metadata) to specific Custom Product Pages in App Store Connect. When the app ranks organically for an assigned keyword, the matched CPP displays instead of the default product page. Keywords from title and subtitle are **not eligible**, only keywords Apple auto-suggests from the keyword field can be assigned. No manual keyword entry is supported. Publishing requires no additional review if the CPP is already approved.

**What the data shows.** No published organic conversion data exists. Paid CPP benchmarks provide directional guidance: AppTweak reports **5.9% average CVR lift** for apps and **3.5% for games** using CPPs in Apple Ads. SoundCloud achieved a **58% conversion rate increase** and **39% lower CPI** with competitor-keyword CPPs. Apple claims a **156% conversion increase** (2.5 percentage point lift) when referring users to CPPs versus default pages. Phiture is running structured experiments but has not published results.

**What to do.** ARPU Brothers (Peter Plachta) calls this "NOT a game changer" but a meaningful addition. His guidance: treat the keyword field as a separate activation channel for CPPs, create themed pages (e.g., "Homework Helper" vs. "Image Generator"), and avoid repeating title/subtitle keywords. AppTweak recommends a CPP prioritization framework evaluating keywords by **Reach** (search volume × organic rank) and **Impact** (potential CVR improvement × incremental installs). Phiture urges caution, open questions remain about long-tail keyword combination behavior, traffic redistribution between CPPs and default pages, and cross-language indexing.

**By vertical.** Babbel was observed using "learn Spanish" CPPs organically in the Canadian App Store (August 2025). Education and fitness apps with multiple distinct use cases (e.g., language-by-language, workout type) are ideal CPP candidates.

**Evidence strength.** Mechanics: strong. Limitations: strong. Conversion data: mixed (paid context only; organic data unavailable). Assignment reset behavior on version update: weak (no explicit confirmation; inferred from documentation).

**Source.** Apple Developer documentation, AppTweak WWDC 2025 recap, Phiture ASO Stack, ARPU Brothers

#### Apple Product Page Optimization (PPO) for A/B testing

*iOS*

**What it is.** Apple's native A/B testing framework that lets developers test up to three treatment variants against the default product page for icons, screenshots, and app preview videos. Tests run on live traffic and require a minimum volume of impressions to reach significance. Unlike CPPs, PPO tests compete on the same keyword surface, they change creative, not the keyword routing.

**What the data shows.** Headspace used PPO to test one creative element at a time. Key finding: users responded positively to screenshots showing social proof and expanded content on key features. "Bright colors in background design and concise captions performed well." Critically, the same features performed **differently across Google Play vs. iOS** for different English-speaking markets.

**What to do.** Thomas Petit emphasizes CPPs are NOT A/B tests, the only valid comparison is before/after, not CPP vs. default page metrics. Use PPO for genuine creative testing and CPPs for intent-matching. SplitMetrics recommends pre-launch testing in their isolated environment before committing to a PPO test with live traffic.

**By vertical.** Wellness apps (Headspace, Calm) have published the most PPO testing data. Finance and productivity apps have limited published results.

**Source.** SplitMetrics, AppTweak, M+C Saatchi Performance (Headspace case study)

### Platform-specific: Google Play

#### Google Play Guided Search and AI discovery

*Android*

**What it is.** An AI-powered discovery system replacing traditional keyword matching with intent-based semantic search. Users type goals (e.g., "deck-building games," "learn to cook") and Gemini interprets intent, organizing results into refinement paths. The system evaluates keywords, creatives, reviews, and store pages as **connected signals** rather than isolated optimization levers. The semantic shift is rolling out unevenly across languages, with stronger signals in English.

**What the data shows.** Google's Curated Spaces (related feature) drove a **30% boost in user search activity** in pilot markets (India, Japan). Audio sample testing among meditation apps showed users who engaged were **3x more likely to install**. Phiture's Avakin Life case study showed **57% conversion rate increase** using intent-tailored Custom Store Listings. No published before/after case study specifically isolates Guided Search impact on an individual app.

**What to do.** RevenueCat prescribes three immediate actions: (1) rewrite store listing for outcomes, lead with the result your app delivers, (2) audit subscription cancel → win-back flows, (3) ship a content update within 30 days. AppTweak recommends grouping keywords into **semantic clusters informed by intent**, not topic similarity. Build intent-specific store pages using Custom Store Listings. Phiture recommends investing in Engage SDK integration and structured experimentation with localized experiences.

**By vertical.** Education apps should lead with learning outcomes ("Speak Spanish in 30 days") not features ("500+ lessons"). Fitness apps should describe transformation ("Run your first 5K") not tools ("GPS tracker").

**Source.** Google official blog (Sept 23, 2025), RevenueCat, AppTweak (Micah Motta, Jan 2026), Phiture PlayTime 2025 recap

#### Google Play "You" tab and churn dynamics

*Android*

**What it is.** A personalized hub combining profile, subscriptions, Play Points, recommendations, and engagement stats. Rolled out starting September 23, 2025 in Play Points markets (US, UK, Japan, Korea), expanding October 1, 2025. Users can now review and manage subscriptions directly within the You tab, making cancellation frictionless. Apps integrating the Engage SDK can surface in-app events, promotions, and embedded videos within this space.

**What the data shows.** Google reported a **2% decrease in voluntary churn** from surfacing subscription benefits more prominently. Default account hold duration increased to **60 days** (from 30) on December 1, 2025, yielding a **10% decrease in involuntary churn**. Play Points campaigns showed organic installs rising by **more than 150%**, retention improving by ~33%, and day-two conversion rates nearly doubling.

**What to do.** RevenueCat explicitly warns: "Users can review and churn more easily, so **win-backs, pauses, and downgrade paths matter more than ever.**" Audit subscription flows immediately. AppTweak's Simon Thillay notes the You tab is becoming central to Google Play's engagement strategy. Phiture recommends leveraging Engage SDK for re-engagement surfaces.

**By vertical.** Fitness and wellness apps with seasonal usage patterns face the highest churn risk from easier cancellation access. Finance apps with strong habitual engagement (daily portfolio checks) may benefit from the You tab's visibility.

**Source.** RevenueCat, Phiture PlayTime 2025 recap, AppTweak PlayTime 2025, Google official

#### Android Vitals and Play Store ranking visibility

*Android*

**What it is.** Android Vitals ties app quality directly to search and recommendation visibility, checked daily against a 28-day rolling average. An app that exceeds a 1.09% user-perceived crash rate or a 0.47% user-perceived ANR rate overall (both cap at 8% on a per-device basis) gets reduced visibility in search and recommendations, plus warning labels on the listing and a quality alert to users before install.

**What to do.** For a Google Play client with an unexplained drop in search visibility, check Vitals before touching the listing. A metadata rewrite cannot fix a ranking penalty caused by crash or ANR rate. The fix is a stability release; visibility recovers on its own once the 28-day rolling average clears the threshold.

**Source.** Google Play Console Help, Android Developers documentation

#### Google Play Store Listing Experiments (native A/B testing)

*Android*

**What it is.** Google Play's native store-listing test runs up to 3 variants against a control, testable on icon, feature graphic, screenshots, video, and short or full description. Only one default-graphics experiment can run at a time. The audience is signed-in Google Play users only, measured on first-time installers plus first-time installers retained after 1 day. Google states a minimum run of 7 days, to clear weekday/weekend variance, and localized tests can run across up to 5 languages simultaneously.

**What to do.** For a client's first ASO test on Android, this native tool is the default before recommending a paid platform like SplitMetrics: no separate vendor, real production traffic, and a stated minimum duration that keeps the test from being called early.

**Source.** Google Play Console Help

### Listing copy optimization

#### Title and subtitle keyword optimization

*Both*

**What it is.** Strategic placement of keywords across the title (30 chars iOS / 50 chars Google Play) and subtitle (30 chars iOS; Google Play uses short description instead). The algorithm assigns priority by position, keywords earlier in the title rank higher. Apple combines keywords from title + subtitle but gives **more weight to title keywords**. Title > Subtitle > Keyword Field in ranking weight.

**What the data shows.** More than half of top US apps include keywords in their name (Appfigures). **70% of App Store visitors use search** to discover apps (Apple Search Ads data). **65% of App Store downloads** follow a keyword search (AppTweak). Apple's algorithm now rewards **focus over stuffing**, concentrating on your most important keyword theme outperforms cramming 30 characters with disparate terms.

**What to do.** Well-known brands should place brand name first and append keywords (e.g., "Duolingo - Language Lessons"). Lesser-known brands should lead with the primary keyword and trail with brand. **Never duplicate keywords** across title, subtitle, and keyword field, Apple ignores repetition. Appfigures recommends updating metadata every **30-45 days** to feed Apple's algorithm more information.

**By vertical.** Education: "Duolingo - Language Lessons" (brand + primary keyword). Rosetta Stone uses all 30 characters: "Rosetta Stone: Learn Languages" with subtitle "Speak Spanish, French & More." Fitness: Calm uses "Calm, Meditation and Sleep" despite strong brand recognition, even well-known brands include category keywords. Productivity: Todoist packs name + subtitle with keywords while remaining human-readable, "balance you always want to strike" (Appfigures).

#### iOS keyword field strategy

*iOS*

**What it is.** The 100-character hidden metadata field in App Store Connect where developers place keywords that the algorithm uses for search indexing. This field has gained strategic importance since CPP Keyword Linking draws exclusively from it. Use commas as separators with **no spaces after commas** (saving 10-15 characters). Do not repeat title or subtitle keywords, Apple does not increase weight for repetition.

**What the data shows.** Use singular forms, Apple automatically indexes plurals in English (but not all languages; e.g., French "animal" ≠ "animaux" per AppTweak testing). Stop words ("the," "and," "with") appear in less than **3.2% of top-ranked titles**; the algorithm ignores them. The US App Store indexes keywords from both English (US) and Spanish (Mexico) locales, effectively providing **200 characters** of keyword space for the US store.

**What to do.** AppTweak: "Do not repeat keywords that you have already added to your title or subtitle. Apple does not take into account keyword density." Include synonyms, related terms, and modifiers that combine with title/subtitle words. Avoid competitor brand names, Apple "does not approve of this practice and can take measures" (AppTweak). Thomas Petit recommends using Apple Search Ads data to validate keywords before committing them to organic metadata: "The insights from Apple Search Ads data, knowing which keywords actually convert, are more valuable than the immediate downloads they bring."

**By vertical.** Duolingo maintains a localized Spanish keyword list for additional US App Store coverage (cross-locale indexing). Education apps should include learning-outcome terms ("fluency," "practice," "conversation") not already in title. Fitness apps should add condition-adjacent terms ("anxiety," "insomnia," "stress") in the keyword field to capture intent-based searches.

**Evidence strength.** Cross-locale indexing: good (well-documented by AppTweak and practitioners but not officially acknowledged by Apple). Competitor brand name risk: strong for Apple's stance: mixed for enforcement consistency.

**Source.** AppTweak, App Radar, Appfigures, ASO Pulse

#### Description optimization for semantic search

*Both (critical for Android; secondary for iOS)*

**What it is.** On Google Play, the long description is fully indexed for keyword search and now evaluated semantically by Gemini-powered Guided Search. On iOS, the description is not directly indexed for keyword search but is read by Apple's editorial team for featuring decisions and may influence relevance signals. With Guided Search, Google Play descriptions should be rewritten around outcomes and jobs-to-be-done rather than feature lists.

**What the data shows.** AppTweak's January 2026 analysis found that natural language processing now enables app stores to interpret a single query as representing **multiple possible user intents**, shaping which apps surface across different intent paths. No quantitative case study directly isolates description optimization impact under Guided Search.

**What to do.** RevenueCat recommends leading the short description with the result your app delivers. AppTweak advises thinking in "intent clusters to own" rather than flat keyword lists. ASO World recommends ensuring descriptions address user needs in natural language and combining with strong engagement metrics (retention, ratings, update frequency).

**By vertical.** Education: describe transformation ("Speak conversational Spanish in 3 months") not feature set ("500 audio lessons, flashcards, grammar drills"). Finance: emphasize outcomes ("Start investing with $5") not mechanics ("Fractional share purchasing available").

**Source.** RevenueCat, AppTweak (semantic clusters article, Jan 2026), ASO World

### Screenshot and visual strategy

#### First three screenshots determine conversion

*Both*

**What it is.** Approximately **90% of users do not scroll past the third screenshot** on an app store product page, making the first three screenshots the primary conversion surface. The average user spends only **7 seconds** on a product page. Less than **2% of App Store visitors** tap "read more" (SplitMetrics), meaning visuals dominate the install decision.

**What the data shows.** Well-designed screenshots can lift page conversion by **20-35%** (ASOMobile). The recommended three-screenshot framework follows a Value → Usage → Trust sequence: Screenshot 1 answers "What do I get?" with the strongest value promise; Screenshot 2 shows the most popular use case in action; Screenshot 3 provides social proof (ratings, awards, user count). Todoist's final screenshot uses a direct CTA, "Get Todoist now", called "the easiest ASO hack" by Appfigures.

**What to do.** AppTweak recommends updating screenshots **2-4 times per year**. ASOMobile advises refreshing creatives every **6-8 weeks**. SplitMetrics found that **57% of top Google Play games** A/B tested screenshots at least twice in 2024, compared to only 34% of apps, subscription apps are undertesting. Use 3-5 word benefit-driven captions. Formula: Verb + Benefit + Result.

**By vertical.** Calm uses a problem-solution-benefit sequence: "Improve Sleep Quality" → "Listen to Sleep Stories" → app UI → "Join 100 Million People." Duolingo uses bright multi-color backgrounds that diverge from category norms, most education competitors use solid/clean backgrounds. Appfigures noted Duolingo's screenshots say "learn a new language" but fail to highlight simplicity, their core differentiator.

**Source.** ASOMobile 2025 guide, SplitMetrics scroll depth data, AppTweak

#### Portrait vs. landscape orientation

*Both*

**What it is.** In iOS search results, portrait orientation displays **3 screenshots** versus landscape displaying only **1**, giving portrait 3× the visual real estate at the critical first-impression moment. Landscape has slightly better scroll depth (15% of users reach the 5th screenshot vs. 11% for portrait), but portrait dominates initial visibility.

**What the data shows.** Rovio's Angry Birds 2 case study found portrait screenshots outperformed landscape by **13% conversion uplift**, generating **2.5 million additional installs** in the first post-release week. The insight: Angry Birds 2 users were casual gamers who held their phones in portrait mode.

**What to do.** For B2C subscription apps, portrait is strongly recommended. Games and media apps with landscape-native experiences may be exceptions. Always A/B test before committing, orientation effects vary by audience behavior.

**By vertical.** All major subscription app verticals (education, fitness, finance, productivity) should default to portrait. Gaming subscription bundles (e.g., Apple Arcade competitors) may benefit from landscape to show gameplay.

**Source.** SplitMetrics (Rovio case study), AppTweak

#### App preview video best practices

*Both*

**What it is.** Short promotional videos (15-30 seconds on iOS, up to 2 minutes on Android) that demonstrate app functionality. On iOS, videos **autoplay silently** in search results and on the product page, making them extremely high-impact. On Android, videos **require a user tap** to play, and only **6% of Google Play users** actually play the video.

**What the data shows.** High-quality preview videos can boost conversion by **20-40%** (SplitMetrics). However, poorly designed videos can **hurt conversion by up to 20%** (ASOWorld). Moburst's Lior Eldan confirms: "After examining dozens of ASO projects, we've noticed that videos sometimes have a negative effect on conversion rates." ASO Stack found that additional videos beyond the first didn't appreciably increase conversion, **one good video is sufficient**. Users who engaged with preview videos were **2× more likely to install** (Leanplum).

**What to do.** Design for muted viewing on iOS, use text overlays and visual cues. The first 3-5 seconds are critical; most users decide immediately whether to keep watching. For subscription apps, show the premium experience, what users get after subscribing. If you lack resources for an exceptional video, **skip it entirely** rather than publish a mediocre one. Use PPO on iOS to A/B test video vs. no video before committing. On Google Play, invest in the Feature Graphic thumbnail rather than the video itself, the thumbnail gets far more impressions than the video.

**By vertical.** Wellness apps (Calm, Headspace) benefit from demonstrating the in-app experience, ambient UI, audio quality, guided sessions. Duolingo notably has **no app preview video** despite being a strong candidate for one (Appfigures flagged this as a missed opportunity). Fitness apps with dynamic movement content (workout demonstrations) see strong video engagement.

**Source.** SplitMetrics, Moburst, ASO Stack, ASOWorld

### Keyword strategy

#### Semantic keyword clustering for the AI era

*Both*

**What it is.** With both Apple and Google moving toward semantic understanding of app listings, keyword strategy must evolve from flat keyword lists to intent-driven clusters. Group keywords by **user motivation**, not topic similarity. Each cluster maps to a distinct job-to-be-done, and each cluster should have corresponding creative assets (CPPs on iOS, Custom Store Listings on Android) that communicate that specific use case.

**What the data shows.** AppTweak's analysis found that AI changes are rolling out as incremental system improvements, not dramatic feature launches. Natural language processing enables stores to interpret a single query as representing multiple user intents. The semantic shift is stronger in English and unevenly distributed across markets.

**What to do.** Build an "intent map" of 6-8 user jobs your app solves (RevenueCat). Align headline, screenshots, and promo video to each intent. AppTweak advises deciding which user motivations your app is best positioned to satisfy, then claiming those intent clusters with matching metadata and creative. Keywords, creatives, reviews, and store pages should work together as **connected signals**.

**By vertical.** A meditation app might cluster around: (1) sleep improvement, (2) anxiety management, (3) focus/productivity, (4) daily mindfulness habit, each with distinct CPP/CSL, screenshot captions, and keyword field terms.

**Source.** AppTweak (Micah Motta, Jan 2026), RevenueCat

#### Ratings and reviews as a conversion lever

*Both*

**What it is.** Star rating directly impacts both conversion rate and search ranking. Both Apple and Google factor overall rating, review volume, rating velocity, sentiment, and recency into search algorithms. The **4.0-star threshold** is a critical inflection point: approximately **90% of featured apps** on the App Store maintain 4.5+ ratings (AppTweak), and Google Play offers literal filter buttons for 4.0 and 4.5 stars.

**What the data shows.** Going from **3 to 4 stars** correlates with approximately **89% increase** in install conversion (AppReply, citing behavioral data). Apps below 4.0 stars lose up to **50% of potential installs** (ASOMobile, multiple sources). Responding to reviews leads to an average rating improvement of **0.7 stars on Google Play** (AppReply), and **70% of users who receive a response** will revise their original review. Hotellook's StoreKit in-app rating implementation generated **5× more ratings in 2 months** than the entire previous year, with impression-to-install conversion rising **15%** (AppFollow). The Robinhood case study (AppTweak) demonstrated that massive download velocity can temporarily override rating drops, during the GameStop affair, Robinhood surged to #1 despite plummeting ratings.

**What to do.** Implement in-app rating prompts at moments of peak satisfaction (completion of a workout, finishing a lesson, positive portfolio performance). Aim for a **4-hour SLA** on responding to 1-star reviews. Every negative review requires roughly **10 positive ones** to offset. Google Play caps developer replies at 350 characters; Apple allows 5,970.

**By vertical.** Finance apps face extreme rating volatility from market events (Robinhood, TD Ameritrade, Webull all experienced dramatic drops during GameStop). Education apps benefit from prompting ratings after lesson completion streaks. Fitness apps should prompt after personal records or milestone achievements.

**Evidence strength.** The 89% conversion lift figure comes from a single source (AppReply) and should be treated as directional rather than definitive.

**Source.** AppTweak (Robinhood case study), AppFollow, ASOMobile, AppReply

#### Localization strategy and ROI

*Both*

**What it is.** Localization encompasses three tiers: (1) metadata translation, keyword fields, title, subtitle, description (lowest effort, immediate search visibility), (2) creative localization, translated screenshot text, localized imagery, local currency (medium effort, significant conversion lift), and (3) culturalization, completely different store pages reflecting local values and usage patterns (highest effort, best ROI in competitive markets). Apple expanded to **50 localizations** (from 39) in March 2026, including 9 Indian languages.

**What the data shows.** Localized store pages deliver **5.1% to 24.6% CVR lift** (StoreMaven, pre-2023, based on 500M+ user sample). SoundCloud achieved approximately **400% download uplift** from translating App Store keywords into Russian (Phiture). Duolingo's downloads in Brazil increased **200%** and Japan **120%** after visual and naming localization (ASOMobile). Flipster (crypto app) saw **25.72% increase in app visibility** in Vietnam within 4 days of localizing (AppTweak). Including the target keyword in screenshot captions produced **2.4× better average rank** (Phiture). Tinder localizes not just text but positioning: US says "Know what they're looking for" (pragmatic), Germany says "Forever? Situationship? You decide" (bold), Japan says "Are you looking for a lover or a friend?" (formal).

**What to do.** Start with metadata localization for visibility gains, then advance to screenshot localization for conversion, then full culturalization for top-priority markets. Phiture notes: go beyond translation, use locally correct references (local units, punctuation, city names, mobile carriers in screenshots). AppTweak advises focusing localization on tier-one markets with proven revenue potential rather than localizing broadly. RevenueCat emphasizes pricing localization: higher subscription prices correlate with **higher trial conversion rates**, 9.8% for high-priced apps vs. 4.3% for low-priced (RevenueCat 2025 report).

**By vertical.** Education: Duolingo adjusts app name by market ("Duolingo: Aprende Inglés" in Spanish, "Duolingo: Englisch lernen" in German) and uses country flags in Europe, cartoon elements in Japan. Finance: Revolut highlights Bizum integration in Spain, "Reimagine your salary" in France, "Great exchange rates" in Japan, practical and service-oriented by culture. Education apps in South Korea show **10.7% download-to-paid conversion** vs. 1.7% all-categories average (RevenueCat 2024).

**Source.** Phiture ASO Stack, AppTweak localization guide, StoreMaven (pre-2023 data), ASOMobile

### Vertical examples

#### Duolingo, the education ASO benchmark

*Both*

**What it is.** Duolingo earns an A-grade ASO score from Appfigures with **616K estimated monthly US App Store downloads**, 2× Babbel and 4× Rosetta Stone. The app name "Duolingo - Language Lessons" with subtitle "Learn Spanish, French and more" covers primary keywords, and 97% of new ratings are positive. The brand keyword "duolingo" has the highest popularity of all its ranked keywords, indicating massive outside-the-store demand.

**What the data shows.** Screenshots use bright multi-color backgrounds, chunky fonts, and actionable 4-5 word captions. Weaknesses identified by Appfigures: screenshots emphasize "learn a new language" without highlighting ease/simplicity, their primary competitive advantage. "language lessons" has a popularity score of only **7**, while competitors Babbel and Rosetta Stone optimize for "learn languages" (popularity **45**). Duolingo has no app preview video despite strong candidacy for one. Cross-locale strategy: maintains a localized Spanish keyword list that Apple uses alongside English in the US store. Localization drove Brazil downloads up **200%** and Japan downloads up **120%**.

**What to do.** Duolingo's brand strength creates organic demand, but keyword selection could improve. Appfigures specifically notes the missed opportunity around "learn languages" vs. "language lessons."

**Source.** Appfigures ASO Teardown, ASOMobile, Geeklab

#### Rosetta Stone and Babbel, education challengers

*Both*

**What it is.** Rosetta Stone uses all 30 title characters ("Rosetta Stone: Learn Languages") with a keyword-rich subtitle ("Speak Spanish, French & More"). It ranks #4 for "speak Spanish" and "speak French", excellent specific-intent targeting. Apple Ads achieved a **90% conversion rate lift** and now accounts for ~20% of new-user acquisitions. Babbel uses competitive counter-positioning: targeting countries where Duolingo is less visible and capturing downloads from misspellings of "Duolingo", an ASO arbitrage tactic. Babbel's bold orange icon provides strong differentiation against Duolingo's green and Rosetta Stone's blue.

**What the data shows.** Rosetta Stone implements thousands of keywords across multiple Apple Ads campaigns using Search Match for keyword discovery. Babbel runs more aggressive Apple Search Ads campaigns than Rosetta Stone, bidding on competitor brand names.

**Source.** Gummicube ASO Spotlight, Apple Ads success story, MobileAction

#### Calm and Headspace, wellness category leaders

*Both*

**What it is.** Calm organically ranks for **10,000+ keywords** (nearly 40% in top 30), compared to Headspace's **3,900+ keywords** (30% in top 30). Calm uses 7 Custom Product Pages and runs 280+ live ad creatives monthly. Headspace uses 5 CPPs and 120+ ad creatives. Both invest heavily in creative testing. Calm holds **100% impression share** for the "mindfulness" keyword in Apple Search Ads. Calm has **4 million paid subscribers** versus Headspace's 2 million.

**What the data shows.** Headspace A/B testing via M+C Saatchi found that social proof and expanded feature content improved conversion. Bright backgrounds and concise captions performed well. Critically, **identical features performed differently across iOS vs. Google Play** in different English-speaking markets, reinforcing the need for platform-specific and locale-specific testing. Calm's screenshot strategy follows a problem-solution-benefit model: "Improve Sleep Quality" (pain point) → "Listen to Sleep Stories" (unique solution) → app UI → "Join 100 Million People" (social proof).

**What to do.** Calm's 10K+ keyword footprint and 7 CPPs represent the gold standard for subscription wellness app ASO. Headspace's systematic one-variable-at-a-time PPO testing methodology is the model for creative optimization.

**Source.** MobileAction, Vermillion agency, M+C Saatchi Performance

#### Strava, fitness community with ASO gaps

*Both*

**What it is.** Strava's title "Strava Running and Cycling GPS" targets specific keywords but misses higher-volume terms, "running and cycling gps" has only **30 searches**, while "fitness tracker" and "run tracker" have significantly higher volume. yellowHEAD identified Strava's screenshots as a major weakness: plain app screens with no backgrounds, no captions, and no emotional design.

**What the data shows.** Strava ranks top 5 for "running and cycling," "cycling," and "cycling gps", but these are low-volume terms. RevenueCat names Strava as an app that experimented with **reverse trials** (full premium access without upfront payment), leveraging strong product-market fit to convert at trial end. Strava's global community of **750,000+ clubs** is its competitive moat but is underrepresented in store listing creative.

**What to do.** yellowHEAD recommends adding runner/cyclist imagery to the icon, introducing caption overlays on screenshots, and shifting keyword targeting to higher-volume terms. The community angle ("train with 100M athletes") could serve as a powerful social proof screenshot.

**Source.** yellowHEAD ASO analysis, RevenueCat

#### Todoist, productivity ASO done right

*Both*

**What it is.** Todoist achieves **75,100 estimated monthly US App Store downloads**, ahead of closest competitors by ~50%. It is regularly featured by Apple. Its name + subtitle combination is keyword-rich yet human-readable. Keywords target specific intent types, "grocery" and "project" as types of lists people make. The app frequently updates its keyword list, taking advantage of the fact that Apple's algorithm remembers previous keywords.

**What the data shows.** Todoist's final screenshot simply reads **"Get Todoist now"**, a direct CTA that Appfigures calls "the easiest ASO hack." Todoist nearly doubled downloads year-over-year, benefiting from Wunderlist's shutdown and a Verizon partnership. Appfigures concludes: "Todoist is simply doing it right", beautiful presentation and strong keyword coverage with top-10 rankings across many terms.

**What to do.** Todoist's strategy of targeting usage-context keywords (grocery lists, project management) rather than generic productivity terms is a model for subscription apps with multiple use cases. The direct CTA screenshot is a low-effort, high-impact tactic worth testing across verticals.

**Source.** Appfigures ASO Teardown

#### Robinhood, finance and the rating volatility lesson

*Both*

**What it is.** During the GameStop crisis (January 2021), Robinhood's rating cratered due to review-bombing, yet the app surged to **#1 in Finance and #1 in All Apps** on Google Play, demonstrating that massive download velocity can temporarily override rating penalties. TD Ameritrade's rating dropped from ~3.04 to **1.13 in a single day**; Webull went from 4.49 to 2.77.

**What the data shows.** The case study is the most cited example of the tension between ratings and download velocity in ranking algorithms. Both Apple and Google factor overall rating, review volume, rating velocity, sentiment, and recency into search ranking, but extreme viral events can overwhelm the rating signal through sheer install volume.

**What to do.** Finance apps must build ratings resilience through proactive review management and rapid response protocols. Maintain a review response team with authority to escalate during market volatility events. Consider implementing in-app rating prompts that are triggered by positive portfolio events and suppressed during market downturns.

**Source.** AppTweak (priority source)

---

## Onboarding patterns

These are the patterns that recur across app categories, grouped by where they sit in the
flow. They are observed prevalence and mechanism rather than measured win rates: treat each as
a hypothesis with a good prior, not a proven lift.

### A. Welcome screens

| # | Pattern | What it looks like | Why it works |
|---|---|---|---|
| A1 | Lead with a concrete, near-term outcome | First-screen headline like "Sleep better tonight," "Save up to 80% on prescriptions," "Learn Spanish in 5 minutes a day" | Users instantly understand relevance, reducing early bounce |
| A2 | Establish credibility early | "Backed by science," expert endorsements, certifications, user counts, institutional logos | Reduces skepticism before users invest time or trust |
| A3 | Differentiate from the status quo immediately | Comparison cards like "YouTube vs. Guided Plan" or "Bank vs. Smart Account" | Helps users justify switching from existing solutions |
| A4 | Make value legible without reading | One-line headline paired with a simple outcome-focused visual | Lowers cognitive load and speeds comprehension |
| A5 | Story-style first screen | Auto-advancing, tap-to-skip panels instead of a swipe carousel | A familiar pattern that moves on its own lowers first-open drop-off |

### B. Personalised questions and setup

| # | Pattern | What it looks like | Why it works |
|---|---|---|---|
| B1 | Frame personalization as value creation | Copy like "We'll build this for you" instead of "Answer a few questions" | Users perceive onboarding as helpful, not a chore |
| B2 | Keep personalization clearly finite | Progress bars, step counts, "Takes 2 minutes" labels | Increases completion by reducing uncertainty |
| B3 | Prefer structured inputs over open text | Single select, multi select, rating scales, agree/disagree statements | Lowers cognitive and emotional effort |
| B4 | Use personalization to surface the problem | Rating sleep quality, stress, skill level, or financial confidence | Makes the need for the product feel concrete |
| B5 | Interleave questions with reinforcement interstitials | Social proof, validation ("Most users start here"), science facts | Prevents fatigue and builds confidence mid-flow |
| B6 | Deliver micro value during onboarding | "Did you know?" facts, insights based on answers | Keeps users engaged and reinforces expertise |
| B7 | Normalize beginner or early-stage status | "Most people start here," "You're not behind" | Reduces shame and fear of starting |
| B8 | Micro-commitments before the big ask | Draw a checkmark instead of tapping Next (Ahead), slide to confirm, pick a goal, placed just before account creation or the trial ask | Active participation raises investment and makes leaving costlier |
| B9 | Offer "suggest for me" at decision points | A default or auto-pick where users stall on a choice (Tiimo's routine picker) | Removes decision paralysis, which drives drop-off on choice screens |
| B10 | Match social proof to the stated goal | Muscle-gain user sees a muscle-gain testimonial, weight-loss user a weight-loss one | Proof reads as "people like me", not generic praise |
| B11 | Reuse the stated goal everywhere | The chosen goal phrase repeats on every later screen and in the paywall headline (Zing); one question changes the paywall title (Tiimo) | Continuity of relevance with no modeling; the user sees their answer was used |
| B12 | Loading screens that sell | The pre-paywall loader shows value, social proof or the plan being built; a persona guide can carry the voice from onboarding into the paywall (Zing's coaches) | Builds anticipation before the ask; never use it to hide extra questions, which makes every loader read as suspect |
| B13 | The recommendation says why it was picked | "Why this tutor fits you" with reviews from learners like you (Preply), a behavioral profile label the user maps to (Noom), a report on blind spots from the answers (Purpose) (operator interviews, 2026) | The user sees the questions were used; the best matching is wasted if its reasoning is invisible |

### C. Account creation

| # | Pattern | What it looks like | Why it works |
|---|---|---|---|
| C1 | Delay account creation until after initial value | Let users explore content or complete personalization before sign-up | Increases sign-up rates through prior investment |
| C2 | Keep initial account creation lightweight | Email only or SSO first, deeper details later | Reduces drop-off at the commitment boundary |
| C3 | Justify heavy data collection contextually | Asking for details only when enabling a feature | Users tolerate friction when it feels necessary |
| C4 | Use progress framing for long flows | Checklists or step indicators during setup | Converts overwhelm into reassurance |
| C5 | Split account creation from the paywall | Tiimo moved signup to right after the welcome screen: signups rose and paywall conversion held. Offer SSO wherever there is an account | Each ask lands lighter and at its own peak of intent; C1's delay is the other valid placement, so test which fits |

### D. Engagement, habit and commitment setup

| # | Pattern | What it looks like | Why it works |
|---|---|---|---|
| D1 | Ask for explicit but appropriate commitment | Learning schedule, wellness intention, training goal | Shifts users from passive intent to action |
| D2 | Match commitment tone to the domain | Soft commitments in wellness, clearer cadence in education/sports | Prevents guilt while still encouraging consistency |
| D3 | Tie notifications directly to user goals | "We'll remind you at your chosen time to practice" | Increases opt-in and long-term engagement |
| D4 | Frame notifications as support tools | Language like "We'll help you stay on track" | Builds trust and reduces notification fatigue |
| D5 | Reinforce progress psychologically early | Messaging like "You've already started" | Encourages continuation before results appear |
| D6 | Preview the real notification before the system prompt | Swipeable previews of actual notifications tied to benefits ("know when you get paid", "avoid overdraft fees"), then the OS dialog (Copilot) | A concrete benefit beats an abstract ask; protects the one-shot OS prompt |
| D7 | Continue onboarding past the paywall | Subscribed or not, the user goes straight into a guided first task personalized from onboarding answers, not a home screen (Imprint, before and after) | Activation and early retention are won in the first task, which a generic dashboard skips |
| D8 | Ask "how did you hear about us" after key actions | Placed late so it does not cost conversion; wording decides the answers (does "App Store" mean search, featuring or Search Ads?); joined to analytics, paired with age | Self-reported source catches channels last-click attribution undercounts |

### E. Paywall and monetisation, at the level of the screen

| # | Pattern | What it looks like | Why it works |
|---|---|---|---|
| E1 | Show a personalized outcome before monetization | "Your plan is ready," "Here's what we recommend" | Users feel their effort was rewarded |
| E2 | Reduce perceived risk explicitly | Free trials, reminder promises, cancel-anytime language | Lowers anxiety and increases trial starts |
| E3 | Make subscription value concrete | Clear "What's included" lists and benefit breakdowns | Helps users justify the decision rationally |
| E4 | Use choice architecture to simplify decisions | Highlighted "Most popular" plan, recommended tier | Reduces decision fatigue |
| E5 | Unlock immediate utility after commitment | Sample lessons, virtual cards, free credits | Reinforces that the decision was correct |
| E6 | Sequence the asks after a decline | Decline survey routes price objections to a discount or lifetime, doubt to more explanation; lifetime only as a second offer, never on the first paywall for annual-intent users | Users differ in what stopped them; a fitted second ask recovers what one generic paywall loses |
| E7 | Recover abandoned and failed purchases | In-app message if still in session, push or email right after exit; failed payments get a retry prompt and an acknowledging paywall next session | These users already chose to buy; friction or hesitation failed, not interest |
| E8 | Match the paywall to the entry point | Content-specific paywall for the locked item tapped (Calm); higher price from high-value features, trial from low-interest ones like Settings (Instasize); ad keyword to matching store page to matching paywall | The tap or the ad is declared interest; selling that exact thing converts it |
| E9 | Toggle trial with two prices | Trial on costs slightly more, pay now costs less, terms stated under each state | Serves trial-averse users and prices the flexibility; same price either way gives no reason to commit |
| E10 | Earned and revealed offers | "You qualified for a free trial" (Noom), a trial card the user taps to claim, a skippable "reveal your discount" story (Flo) | Endowment effect: people value what they feel they earned; prefer a guaranteed reveal over a chance draw |

---

## What changes by vertical

The structural playbook is close to universal. What differs is how each category earns trust
at the moment of purchase, and that difference is the most actionable cut in this document.
Copying a paywall you admire from another category is the most common expensive mistake.

### The critical checks, per vertical

**Every vertical.** Check these alongside the vertical's own list.

- No two big asks back to back: account creation and the paywall sit apart, each where intent is highest.
- Every choice screen has a default or a "suggest for me" way through.
- The stated goal is reused on later screens and in the paywall headline, not asked and dropped.
- Social proof matches the goal the user just picked, not one generic testimonial.
- The pre-paywall loader carries value, proof or the plan being built, never hidden questions.
- The notification primer previews the real notification before the system prompt.
- After the paywall, subscribed or not, the user lands in a guided first task built from their answers, not a generic home screen.
- Any copied competitor mechanic is named for the mindset it plays on and fits this product's value; added steps answer measured drop-off.
- When the core output renders in seconds (a photo edit, a draft, a dictation), the tool comes before any question, login or paywall, and the asks follow the first output; a plan or habit product keeps the quiz as its onboarding.
- The first screens are written for a first-timer, never the team's own expert user: no jargon, no failure-marking of the user's own attempt (red X, "wrong", "blunder") before the first win, and a positive line after a miss.
- The first screen states the one result the product is known for, in the words of its store listing's first screenshot or subtitle (a photo turned into calories, a draft from a blank page); a first screen that only greets, lists features or asks for sign-in is Missed. What follows it is set by the tool-first check above; a plan or match product may open on its first question when the question names the result.
- The first paywall is not the only ask: a decline leads somewhere (second offer, "why not today?" question routed to the next offer, welcome offer in the first 2-3 days); any discount in that sequence runs against a no-offer arm.
- An abandoned store sheet and a failed payment each trigger their own fast follow-up, not the generic paywall.
- A locked-feature tap opens a paywall about that feature; ad- or keyword-acquired users see the promise that brought them.
- Risk is answered on the screen: refund or cancel line, and the trial terms stated plainly under any trial toggle.
- Decline copy states the consequence neutrally; no guilt-trip wording.
- Freemium: premium items carry locks and upgrade entry points across the app, not only on the paywall.
- Every new user meets the paywall inside the first-session flow (in onboarding or at its end), not only behind a feature tap or a settings link; where analytics exist, day-1 view rate is read against the 80% target and apart from view-to-purchase.
- Each tier is named and sold for what the user gets, never for the technology behind it ("conversation practice", not "the AI tier"), and a user can tell the tiers apart from the paywall alone.
- Where a free tier exists, the actions that bring the next user in (inviting a friend, sharing or publishing what was made, creating what others find) are reachable on the free tier: a lock or paywall on the invite or share control is Missed; a free allowance of creation (a daily count, starter credits, a set number of free uses) is Met, because a use with real cost is metered, not gated. What is gated is depth, power-user capacity or cosmetics, and a share surface carries a free state as well as the paid clean one.

**Health.** Mark the transition from assessment to action with an explicit commitment moment. Use direct, clinical assessment tone to create trust and perceived value. Establish institutional trust early through social proof and professional framing. Move quickly from branding to outcome-driven value to establish seriousness. Justify data collection through structured educational sequences, not scattered proof. Deliver a clear, personalized outcome immediately after assessment. Allow commitment earlier in the flow, including account creation before personalization. Position personalization as the core product. Audit the question flow against one frame: it exists to triangulate the user's goals, their level of experience, and where they are today, then measure the gap, because closing that gap is the product. Every question should feed one of the three; a question that feeds none is a data grab. Check which job the app is doing before assuming a motivation question belongs. Medical care and access apps route by treatment or service selection instead, and asking them for a goal is asking the wrong question. Present the paywall directly after personalization with outcome-based justification. Keep trial initiation simple, relying on trust rather than heavy reassurance.

**Wellness.** Lead with an immediate problem/solution promise on the first screen. Introduce social proof and scientific credibility early to reduce uncertainty. Use intent-based entry questions to anchor relevance ("What brings you here?"). Frame personalization as a professional or clinical intake. Use ratings, satisfaction scales, and agreement statements to surface and quantify pain. Insert reflective feedback or insights mid-flow to validate the user. Reinforce value and education during onboarding with "Did you know?" moments. Keep signup and trial conversion low-friction and anxiety-reducing.

**Sports and fitness.** Frame the product as a long-term performance companion tied to athletic identity. Position personalization as performance optimization, not setup. Anchor onboarding around goals with clear time expectations and visible progress. Move quickly from splash to product value to preserve momentum. Deliver a clear, personalized performance outcome at the end of setup. Reinforce trust mid-flow with guidance, social proof, and coaching reassurance. Request notifications contextually and frame them as coaching support, not alerts. Establish which of two populations you are. Fitness and training apps triangulate the user's goals, experience and current state so the product can measure and close the gap. Sports fan apps do one job instead: find out who the user follows, then personalise everything downstream off that. For a fan-facing app, treat three core questions as the ceiling, not the starting point. Justify pricing through concrete performance benefits and trial- or credit-based exposure.

**Education.** Communicate a sharp, time-bound learning value proposition in the first 1-2 screens. Anchor learning to personal motivation, goals, and intent early on. Calibrate difficulty through level or skill assessment before delivering content. Frame personalization as a short, finishable flow with visible progress. Use a friendly, humanized guide or voice to lower learner anxiety and build trust. Let users select topics or interests to create autonomy and fast relevance. Normalize beginner status with social proof to reduce insecurity. Reinforce credibility and differentiation mid-flow through proof and comparison. Set expectations for habit and consistency before the first lesson. First, decide whether the app has a transactional job or a curriculum job. If the user came to get an answer, onboarding friction lowers completion and the right question count is zero. Ask prior knowledge only if something downstream sequences on it. Treat 15 questions as the observed ceiling and four as the observed floor, then justify every question against what it changes for the learner. Introduce the paywall only after personalization, with clear value and risk reversal.

**Finance.** Defer compliance and financial detail requests until they are contextually required. Establish user-first or fairness positioning early to counter baseline distrust. Use structured thoroughness and checklist-style progress to signal legitimacy. Frame security and identity verification as user protection, not friction. Require explicit account intent and role definition before deeper setup. Lead with a concrete, immediate financial upside on the first screens. Enable instant usability via virtual cards, funding, or early access. Lower trust barriers with a usable free tier before payment.

### How each category earns trust at the paywall

- **Health** leads with clinical proof: outcome statistics, press logos, before-and-after
  projections. Trust is data-driven rather than testimonial. Billing cadence is often
  non-standard in a way that obscures the annualised cost.
- **Wellness** uses the calmest visual language in the set, spacious and single-CTA, with
  emotional, identity-based testimonials. Post-decline discount offers appear here.
- **Sports** leans on feature-comparison matrices, sometimes several screens deep, and uses
  long free trials of one to two months instead of discounts. Hardware-bundled subscriptions
  and consumable credit packs appear only in this segment.
- **Kids** is the only segment where the buyer and the user are different people. It carries
  three signatures found nowhere else: parental gates as purchase friction, child-safety trust
  badges, and licensed characters. Copy addresses the parent while imagery sells the child.
- **Education** splits in two: book-summary apps using mascots and "read in 15 minutes"
  framing, and language apps using streaks and multi-seat family plans. It makes the heaviest
  use of the transparency timeline and carries heavy post-decline discount overlays.
- **AI companions** use escalating tier ladders and sell chat-time limits, memory and
  intelligence as the gated value, paired with character-attachment hooks. They lead with
  feature-gap tables and defer price to the end.

### Conditions that change the approach

Category is a proxy. These are the underlying conditions that actually decide the design.

| If | Then |
|---|---|
| The product genuinely cannot work without the data | Ask early and say why. Drop every question that is not needed |
| People arrive curious rather than convinced | Show them the problem before you sell the fix |
| They doubt that it works | Lead with numbers, outcomes and credentials |
| They doubt it is for someone like them | Lead with identity and real testimonials |
| They came to get one answer, fast | Skip onboarding. Every question loses people |
| You are a health or fitness app | Sell annual. Almost every other category moved to weekly |
| People take days to decide rather than minutes | Do not put a hard paywall in the first session |
| You sell in more than one country | Retest the paywall in each. The winner changes |
| You are about to run a test | Change the plan or the trial, not the colours |

---

## Paywall and monetisation configurations

41 configurations, grouped. Each says what it is, what the data shows, why it works, when to use it, and how strong the evidence is.

### Paywall architecture and placement

#### Hard paywall with trial vs. freemium

**What it is.** Users encounter a paywall immediately or after onboarding, requiring a trial start or payment before accessing core features. No meaningful free tier exists. The paywall typically shows a free trial CTA prominently.

**What the data shows.** Hard paywall apps generate **8.1x higher D60 RPI** ($3.09 vs $0.38) and **5x higher D35 conversion** (10.7% vs 2.1%) compared to freemium. Hard paywalls produce **21% higher LTV per subscriber** (Adapty). However, Year 1 annual retention is roughly equal (~27-28%) between models, suggesting the advantage is front-loaded. Hard paywall conversion declined from 12.1% to 10.7% year-over-year.

**Why it works.** Forces a decision at peak intent (moment of download). Eliminates "free rider" dilution of the install base. Every user who proceeds has demonstrated willingness to engage with premium.

**When to use it.** Apps where core value requires full access (fitness programs, creative tools, content libraries). Less suitable for apps depending on network effects or viral growth where free users create value.

**Apps doing it.** Mojo (creative tools), BetterMe (fitness), Headspace (evolved from generous freemium to near-complete content lock, reporting "double-digit" conversion improvement)

**Evidence strength.** Strong. Confirmed by a large dataset or a controlled test. Independently confirmed by two large datasets.

**Source.** RevenueCat 2026 (revenuecat.com/state-of-subscription-apps), Adapty 2026 (adapty.io/state-of-in-app-subscriptions-report)

#### Onboarding paywall placement

**What it is.** Paywall presented at the end of the onboarding flow, before users access the main app. The user completes a brief onboarding (3-10 screens demonstrating value) and then sees a full paywall with trial offer.

**What the data shows.** At Mojo, the onboarding paywall accounts for **~50% of all trial starts**. Adapty data shows onboarding paywalls with trials convert at **1.35%** vs 0.89% for in-app placement (2x). Adapty confirms **89.4% of all trial starts happen on Day 0**. Rootd (anxiety app) saw a **5x revenue increase** from moving its paywall to the beginning of onboarding. FitnessAI saw **2x install→trial and 2x install→paid** by moving the paywall before onboarding and adding video. Greg (plant care app) went from 3% to 15% sign-up-to-trial (5x) by moving paywall to onboarding.

**Why it works.** Captures users at peak intent, immediately after download, motivation is highest. Aligns with industry data showing Day 0 is overwhelmingly the conversion moment. Apple's App Store team directly recommended this approach to Mojo.

**When to use it.** All categories except Education (where only 71.3% of trials start on Day 0, users browse and return). Especially effective for Health & Fitness, Entertainment (94.5% Day 0), and Photo & Video (91.2% Day 0).

**Apps doing it.** Mojo (50% of trials from onboarding), Rootd (5x revenue), FitnessAI (2x conversion), BetterMe, Headspace

**Evidence strength.** Strong. Confirmed by more than one large dataset. Confirmed by multiple primary sources and two benchmark reports.

**Source.** Parízek RevenueCat guide (revenuecat.com/blog/growth/guide-to-mobile-paywalls-subscription-apps), Francescu Santoni Sub Club Ep. (subclub.com), Rootd Sub Club episode, Adapty 2026

#### Reverse trial (full premium, no payment upfront)

**What it is.** New users receive full premium access immediately without entering payment information. After a set period (7-30 days), they revert to a free tier and must pay to re-access premium features. It runs in two placements: as the front door (every new user gets premium for the period, the Strava and Mumigo shape) and as the decline path (premium granted only after a dismissed paywall, no purchase UI, the shape Apple currently accepts on iOS). The mechanism is the same; the population it touches is not.

**What the data shows.** Elena Verna (Dropbox Growth) reports **15-40% higher conversion than pure freemium**, with ~25% conversion rate possible (vs 2-5% for standard freemium). Mumigo (transit app) moved from **0.5% to 8% conversion** with reverse trial as primary driver. Strava uses a variant with 30-day preview for new users. No specific published conversion data from Strava or Ladder.

**Why it works.** Loss aversion, after experiencing premium, the free tier feels like a downgrade rather than the status quo. Users form habits around premium features that become painful to lose. Lower friction than traditional trial (no payment info required) attracts a larger top of funnel.

**When to use it.** Apps with strong network effects or complex feature sets that require time to appreciate. Particularly effective when the free-to-premium gap is experiential, not just functional.

**Apps doing it.** Strava (social fitness), Ladder (workout planning), Mumigo (transit)

**Evidence strength.** Good. Consistent across several independent reports. Elena Verna's framework is widely cited; app-specific data varies. The 0.4% to 4.5% decline-path figure that circulates is one practitioner's freemium app (Steve P. Young, quoted inside RevenueCat 2026), not platform data.

**Source.** cxl.com, stormy.ai

#### Quiz → personalized plan → paywall sequence

**What it is.** Users complete an extended quiz (10-96+ questions) about goals, preferences, and context. The app generates a "personalized plan" with a dynamic timeline. A paywall follows, framing the subscription as the cost of the personalized plan.

**What the data shows.** Noom generates **~$750M, $1B/year** using a 77-96+ step quiz flow. No published A/B test data comparing quiz vs. no-quiz exists from Noom directly. BetterMe uses a 41-step quiz with hard paywall and reports **$2M/month** revenue. Web-first quiz funnels avoid Apple's 30% cut (keeping 94-97%).

**Why it works.** Sunk cost effect (users invest 10+ minutes and feel committed), personalization increases perceived value ("this plan is made for me"), countdown timers create urgency, and weekly price reframing reduces sticker shock. The quiz also captures segmentation data for downstream personalization.

**When to use it.** Health & fitness, wellness, language learning, categories where personalization is a credible value-add. Requires sufficient content depth to make "personalization" feel real. Works best with web-to-app funnels.

**Apps doing it.** Noom (weight loss), BetterMe (fitness), Fastic (fasting), Yazio (nutrition)

**Evidence strength.** Strong. Confirmed by a large dataset or a controlled test. For strategy description; mixed for revenue attribution (Noom's quiz is part of a full-stack growth engine, not an isolated variable).

**Source.** screensdesign.com, retention.blog, lazertechnologies.com

#### Campaign/app-open triggered paywalls

**What it is.** Paywalls triggered by specific app events (app open, feature tap, usage milestone, push notification) rather than only during onboarding or at feature gates.

**What the data shows.** At Mojo, an app-open triggered campaign paywall generated **15% of new iOS revenue** with "minimal complaints from free users." Superwall data shows exit-intent offers (paywall shown when user dismisses primary paywall) add **15-20% to total revenue**. Jake Mor reports transaction abandon discounts account for **25-40% of revenue** in his own apps.

**Why it works.** Multiple touchpoints increase the probability of catching users at a moment of high intent. Event-triggered paywalls match the offer to the context (e.g., showing the paywall after a user tries a locked feature). Frequency and context matter, aggressive triggering causes fatigue.

**When to use it.** Apps with clear value moments (completing a workout, reaching a content limit, receiving a notification). Requires careful frequency capping to avoid degrading the experience.

**Apps doing it.** Mojo (app-open campaign), Avast ("Get Pro" button generating +10-20% revenue)

**Evidence strength.** Strong. Confirmed by a large dataset or a controlled test. First-party data from practitioners.

**Source.** Parízek RevenueCat guide (revenuecat.com/blog/growth/guide-to-mobile-paywalls-subscription-apps)

#### Post-paywall welcome offers / exit offers

**What it is.** A secondary, discounted offer shown only to users who close the primary paywall without converting. Typically a 24-hour limited-time offer at 30-50% discount. Can also trigger at the next session or after trial cancellation.

**What the data shows.** Adapty identifies post-paywall welcome offers as producing **10-15% ARPU uplift**. Superwall reports exit offers contribute **15-20% of total revenue**. Jake Mor notes that only **~50% of users who start an In-App Purchase complete it**, exit/abandon offers capture this drop-off.

**Why it works.** Catches high-intent users whose objection was price, not interest. Time-limiting the offer creates urgency. Presenting a lower anchor after the primary price makes the discount feel significant.

**When to use it.** Hard paywall apps where the primary paywall sees significant dismiss rates. Works across all verticals. Test discounted vs. no-offer (Mojo found no offer sometimes won during Black Friday in the US; Runna removed its win-back discount after the no-offer cohort retained and paid more, and uses extended trials instead of discounts to protect a premium brand: its Head of Growth in a 2026 podcast interview, direction only).

**Apps doing it.** Mojo (tested during Black Friday), Runna (win-back discount pulled), apps using Superwall's transaction abandon flow

**Evidence strength.** Good. Consistent across several independent reports. Consistent across multiple practitioner reports.

**Source.** Adapty 2026 report, Superwall blog (superwall.com/blog), Jake Mor X/Twitter

#### Paywall placement timing (immediate vs. delayed)

**What it is.** The strategic decision of when users first encounter a paywall, immediately post-install, after onboarding, after demonstrating value, or triggered by engagement events.

**What the data shows.** **89.4% of trial starts** happen on Day 0 (Adapty). Purchases are slower than trial starts: the Day 0 share of paid conversions runs from 15.6% (Education) to 71.4% (Photo & Video) in North America (Adapty 2026), and RevenueCat 2026 puts about a third of all paid conversions on Day 0. Education is the major outlier: only 71.3% of trials start on Day 0, with **23.5% starting at Day 31+**. Adapty's data on discounts for aged users: offering discounts to users who haven't converted after 30 or 90 days "works especially well" (Parízek). ~50% of new revenue comes in the first 7-14 days post-install (Mojo data).

**Why it works.** Download intent decays rapidly. Users who delay trialing rarely return. The counterargument (let users experience value first) holds only in categories with complex, non-obvious value (Education, Finance). For most categories, earlier = better.

**When to use it.** Immediate paywall in H&F, Photo & Video, Utilities, Entertainment. Delayed/contextual paywall in Education and apps requiring relationship-building.

**Apps doing it.** Rootd (5x revenue from earlier placement), Headspace (double-digit conversion improvement from locking content), Duolingo (freemium with progressive feature locking, different model)

**Evidence strength.** Strong. Confirmed by more than one large dataset. Corroborated across both major benchmark reports.

**Source.** Adapty 2026, RevenueCat 2026, Parízek guide

### Pricing and plan structure

#### Weekly subscription + trial as highest-LTV configuration

**What it is.** Offering a weekly plan with a free trial (typically 3-day) as the primary subscription option.

**What the data shows.** Adapty 2026 reports weekly+trial produces **$54.50 in 12-month LTV**, the highest of any configuration, with **636% growth from Day 0 LTV of $7.40**. Annual+trial produces $49.92. Weekly plans generate **55.5% of all app revenue** (up from 43.3% in 2023). Weekly plans convert **1.7-7.4x better** than annual across all price tiers. Adapty's best-by-LTV setup: **weekly $5.99 with 3-day free trial** (1.5x avg LTV of all setups). However, RevenueCat shows annual-dominant apps have **2x higher RPI** ($0.46 vs $0.32), and weekly subscribers retain at only **5.5% at Day 380** vs 19.9% for annual.

**Why it works.** Weekly billing produces higher cumulative payments from retained subscribers despite lower per-period revenue. The low initial commitment reduces friction. Trial + weekly creates a gentle ramp: $0 → small weekly charge → habit formation. The tension: weekly optimizes individual subscriber value but annual optimizes revenue per install.

**When to use it.** Gaming (82% weekly), Productivity (66% weekly), Utilities (73.6% weekly). Health & Fitness is the exception where annual dominates (60.6% of revenue). Monthly is described as "essentially a dead zone" at top of funnel by Adapty.

**Apps doing it.** Most gaming and productivity apps. Health & Fitness apps should consider annual-first strategies.

**Evidence strength.** Strong. Confirmed by a large dataset or a controlled test. But note the RPI vs. Subscriber LTV tension between the two reports. The "right" answer depends on CAC and growth model.

**Source.** Adapty 2026 (adapty.io/state-of-in-app-subscriptions-report), RevenueCat 2026

#### Plan count optimization (1 vs. 2 vs. 3 products)

**What it is.** The number of subscription options presented on the paywall. Options may vary by duration (weekly/monthly/annual), tier (basic/premium), or both.

**What the data shows.** Superwall data (top 15 apps, 10K+ paywall views each): **2 products vs 1 product = +61% conversion increase**. **3 products vs 2 products = +44% additional increase**. Baseline average conversion: 3.77%. Adapty data shows plan count experiments have a **46.1% CR uplift win rate** (second highest after localization). Purchasely case study: EarthSnap saw **+120% new paid subscribers** by offering only a yearly option (with trial) vs. multi-option, directly contradicting the "more options = better" rule.

**Why it works.** Multiple options provide anchoring and let users self-select by budget and commitment level. However, too many options can create decision paralysis. The EarthSnap counterexample suggests that for simple apps with clear value, removing choice friction can outperform. Jake Mor cautions: "Make sure your LTV for all products is the same. Otherwise, you could sacrifice long-term growth for short-term gains."

**When to use it.** 2-3 products works for most apps. Single-product may work for education/content apps with straightforward value propositions. Always validate with A/B testing.

**Apps doing it.** Superwall network (aggregate), EarthSnap (yearly-only winner), Tinder (3-tier: Plus → Gold → Platinum built incrementally)

**Evidence strength.** Mixed. Directionally supported, not proven. Superwall data is observational/cross-sectional (not randomized A/B test). Significant confounds acknowledged by Jake Mor.

**Source.** Jake Mor / Superwall blog (superwall.com/blog/how-many-products-should-you-offer-on-your-paywall), Purchasely case studies

#### Default plan selection

**What it is.** Which plan is pre-selected (visually highlighted or toggled on) when the paywall loads. Often used to steer users toward annual plans.

**What the data shows.** Superwall data: switching default from monthly to annual **"totally inverts" plan mix**, e.g., from 37% yearly to **63% yearly**. Among high-intent users (who didn't close the paywall), **82% selected the default yearly option**. Among "shrewd shoppers" (who closed then returned), **54% selected monthly**. Mojo hid the monthly plan behind "View All Plans" and showed only yearly by default, yielding a **+15-20 percentage point increase** in yearly subscription share with only minor conversion impact. Sunflower app (Superwall customer): pre-selecting annual plan drove **70% increase in annual revenue**.

**Why it works.** Default bias, users follow the path of least resistance. The pre-selected option functions as the app's "recommendation." Removing lower-priced options entirely forces the choice between annual and nothing, which can increase annual share but risks conversion volume.

**When to use it.** Any app where annual plans have higher LTV. Particularly effective when combined with monthly-equivalent price display.

**Apps doing it.** Mojo (hide monthly, show yearly only), Sunflower (pre-select annual)

**Evidence strength.** Good. Consistent across several independent reports. Primary source data from Superwall's network.

**Source.** Superwall newsletter (superwall.dev/blog), Parízek RevenueCat guide, Stormy AI blog

#### Annual price anchoring with monthly equivalent shown

**What it is.** Displaying the annual subscription price re-framed as a monthly equivalent (e.g., "$4.99/month, billed annually" instead of "$59.99/year").

**What the data shows.** Mojo regional results: Brazil **+45%** new revenue per paywall impression, Mexico **+26%**, USA **+8%**, with zero price change. Latin America specifically showed **+30% trial start rate and +10% yearly take rate** from re-anchoring annual to monthly equivalent. The effect works through both increased paywall-to-trial conversion AND greater share of yearly plans.

**Why it works.** Monthly equivalent reduces perceived cost ("$4.99 feels manageable"). Particularly effective in lower-income regions where monthly rates are more relatable and annual lump sums feel prohibitive. The price doesn't change, only the framing.

**When to use it.** All apps with annual plans. Highest impact in emerging markets (Brazil, Mexico, Latin America broadly). Lower but still positive impact in mature markets (US +8%).

**Apps doing it.** Mojo (primary test case across 3 regions)

**Evidence strength.** Strong. Confirmed by a large dataset or a controlled test. First-party data from Mojo's Growth PM with region-specific results.

**Source.** Parízek LinkedIn + RevenueCat guide, NeoAds Substack analysis

#### Decoy pricing

**What it is.** Including a plan option designed not to be chosen but to make the target plan appear more attractive by comparison. Classic example: showing a 3-year option at a high price to make the 2-year plan look like a deal.

**What the data shows.** Avast tested 1-year, 2-year, and 3-year licenses with the 3-year option priced much higher. **Successfully increased 2-year purchases** (no specific percentage given). Adapty recommends adding a higher-priced plan as "the most common way to increase ARPU without losing conversion." Superwall data shows adding an annual plan with 60% discount alongside monthly drove **proceeds per user up 80%** and **trial conversion up 22%**.

**Why it works.** Asymmetric dominance effect, when one option clearly dominates another, it becomes easier to choose. The decoy eliminates the "is this worth it?" question and replaces it with "which is the better deal?"

**When to use it.** Any multi-plan paywall. Weekly and lifetime plans can also serve as decoys (Parízek).

**Apps doing it.** Avast (3-year decoy for 2-year), OctaZone (added 3-month plan, raised annual price to redistribute users to shorter plans with faster ROAS payback), Nebula (weekly with a 3-day trial for the unsure, a no-trial 3-month priced just under annual as the decoy, annual as the value play; put the trial on the plan you want chosen)

**Evidence strength.** Good. Consistent across several independent reports. Behavioral economics principle well-established; specific app data from first-party sources.

**Source.** Parízek RevenueCat guide, Superwall conversion stats page

#### Free trial length optimization

**What it is.** The duration of the free trial period before billing begins. Common options: 3-day, 7-day, 14-day, 30-day.

**What the data shows.** RevenueCat 2026: long trials (17-32 days) convert at **42.5%** vs **25.5%** for short trials (<4 days), a 70% improvement. Despite this, **46.5% of apps now use trials of 4 days or less** (up from 42.1%). Short trials: **55.4% of 3-day trial cancellations happen on Day 0**, and 84% by Day 1. Adapty's best-by-LTV setup: **3-day free trial** with weekly $5.99 plan. Trial subscribers retain **1.4-1.7x better** than direct buyers (Adapty). Health & Fitness leads trial-to-paid at **35%**, Entertainment trails at **19.1%**.

**Why it works.** Longer trials produce higher conversion but delay revenue. Growth teams increasingly choose short trials for faster cashflow and ROAS payback. The optimal trial length depends on how quickly users experience core value, fitness apps may need 7+ days to form habits, while photo editing apps demonstrate value in minutes.

**When to use it.** 3-day for utilities, photo/video, creative tools (value demonstrated fast). 7-day for fitness, wellness (habit formation). 14-30 day for education (learning requires time). Always test, Adapty reports trial structure experiments have a **59.6% LTV win rate**.

**Apps doing it.** Industry-wide pattern

**Evidence strength.** Strong. Confirmed by more than one large dataset. Corroborated across both benchmark reports.

**Source.** RevenueCat 2026, Adapty 2026

#### Introductory offers and discounted first period

**What it is.** A reduced price for the first billing period (e.g., $0.99 for the first week, then full price), distinct from a free trial.

**What the data shows.** RevenueCat blog case study: **85% intro discount on weekly plan** drove a **433% conversion lift** (3% to 14% peak) and CPA dropped from **$105 to $13** (-86%). However, LTV-30 dropped from **$55 to $10** (-81%) and first-renewal retention fell from 50% to 40%. Only 13.5% of App Store transactions use intro offers (5.9% on Google Play). Superwall reports testing introductory offers increased **conversions by 20%**. Only **6.9% of all subscriptions** use discounts (Adapty). Purchasely's Stretch app: quarterly plan with trial drove **+18% conversion to paid** and **-22% post-trial churn**.

**Why it works.** Lowers the perceived risk of commitment. The downside: intro-offer subscribers have lower willingness to pay and higher churn at full price. The RevenueCat case study demonstrates this tension, massive conversion lift but equally massive LTV decline. Net positive in that case only because MRR exceeded $50K for the first time.

**When to use it.** Apps with strong retention mechanics that can recoup intro-period losses. Not recommended as a permanent strategy, better as a targeted tool for specific segments (aged non-converters, win-back campaigns).

**Apps doing it.** RevenueCat case study app (unnamed), Stretch by Breakthrough Apps

**Evidence strength.** Strong. Confirmed by a large dataset or a controlled test. Primary source case studies with detailed metrics.

**Source.** RevenueCat blog (revenuecat.com/blog/growth/introductory-offers-apps), Purchasely case studies

#### Pricing localization / purchasing power parity

**What it is.** Adjusting subscription prices by country or region based on local purchasing power, competitive landscape, and willingness to pay.

**What the data shows.** Adapty reports localization experiments have the **highest LTV win rate of any test type at 62.3%** (vs 45.5% for price changes, 34.6% for visual/text). Mojo saw different paywall designs outperform by market (Japan: long-scroll with social proof yielded **+20% conversions vs US**; Latin America: monthly anchoring). European apps now charge **29-39% more** than North American apps across all plan types, with European prices jumping **18% YoY**. Adapty Pricing Index: India = 0.6x US baseline, Indonesia/Turkey = 0.7x, UK/France/Germany = 1.2x.

**Why it works.** Local willingness to pay varies enormously. German users accept $17.19/month while Turkish users expect 4.4x less for the same H&F app. Localization is not just currency conversion, it includes layout, copy, social proof strategy, and payment method preferences.

**When to use it.** Any app with international users (>20% non-domestic). Adapty recommends testing localization third (after pricing, then visuals), delivering **up to 15% uplift**.

**Apps doing it.** Mojo (Japan, Brazil, Mexico variants), Gamma (localized pricing in 8-10 countries, the rest on a dollar rate; international retention was only worse at USD prices, and a small team localized content first, product chrome in year two, market-specific features in year three: its head of product in a 2026 podcast interview), industry-wide

**Evidence strength.** Strong. Confirmed by more than one large dataset. Highest-win-rate test type per Adapty's analysis of thousands of experiments.

**Source.** Adapty 2026, Parízek guide, Mojo experiments

### Paywall design and UX treatments

#### Paywall copy: outcome framing vs. feature framing

**What it is.** Headline and benefit copy that frames the subscription as a personal outcome ("Get your personalized 7-day plan," "Look your best") rather than a feature list ("Unlock 500+ workouts," "Unlimited reminders").

**What the data shows.** Clear (meditation app): replacing features carousel with preview of the first week's program drove paywall conversion from **20% to 30% (50% lift)**. Parízek's testing order recommendation: copy and imagery tests "rarely yield huge lifts" compared to pricing and layout, but can help when messaging is unclear. Calm's paywall "explains the why, not the what." Superwall's framework: "Outcomes > Features" consistently validated.

**Why it works.** Outcomes connect to the user's motivation for downloading the app. Features answer "what do I get?" but outcomes answer "what will happen to me?" Identity-framing ("Become a meditator") can be even more powerful than outcome-framing for habit-forming apps.

**When to use it.** All verticals, but particularly high-impact in wellness, fitness, and education where the gap between features and aspirations is widest.

**Apps doing it.** Clear (50% lift), Calm, Ladder ("Get results, without planning workouts"), Noom

**Evidence strength.** Mixed. Directionally supported, not proven. Clear app result is from a secondary source (dev.to). Principle is well-established in behavioral science.

**Source.** dev.to/paywallpro analysis, Superwall experimentation framework, Parízek guide

#### Social proof treatments at paywall

**What it is.** Including testimonials, star ratings, review counts, user numbers, or media logos on the paywall to build trust and reduce perceived risk.

**What the data shows.** Mojo's Japan-specific paywall with long-scroll social proof and Free vs. Pro comparison yielded **+20% conversions vs US layout**. "No Payment Due Now" label: Superwall reports this "increases conversions every single time." Adding video backgrounds produces **8-15% conversion increases**. RevenueCat case study: trial toggle + reviews + simplified plan display → **+17.02% ARPU boost**. Social proof is culturally variable, Japan strongly responds to testimonials and reviews, while US responds more to sale framing.

**Why it works.** Social proof reduces perceived risk ("others found this valuable"). Star ratings and review counts leverage authority bias. "Cancel anytime" and "No payment due now" address the #1 fear (being trapped). The key is matching social proof to cultural context.

**When to use it.** All verticals. Most impactful when the app has genuinely strong reviews (4.5+ stars). Video backgrounds and user testimonials are additive.

**Apps doing it.** Mojo (Japan variant), Blinkist (honest paywall timeline)

**Evidence strength.** Good. Consistent across several independent reports. Multiple practitioner reports, culturally variable effects documented.

**Source.** Parízek guide, Superwall blog, NeoAds analysis

#### Multi-screen / storytelling paywalls

**What it is.** Breaking the paywall into multiple sequential screens, showing benefits, social proof, feature comparison, and pricing as a guided narrative rather than cramming everything onto one screen.

**What the data shows.** Mimo (coding education app) saw a **60% lift in trial opt-ins** using multi-screen paywalls that broke benefits, pricing, and feature comparisons step by step. PaywallPro identifies 10 paywall archetypes, with the "Storytelling/Landing Page" style (scrollable, mini-website feel with trust badges, testimonials, comparison charts) among the highest-performing patterns. Purchasely's "Flows" product was built specifically to support multi-screen paywall experiences. OMENA replaced a timeline paywall with one long scrollable screen (value, testimonials, real faces, founder story, FAQ) and **doubled trial starts** in their test; trial starts alone can mislead, so read the paid conversion behind them.

**Why it works.** Reduces cognitive overload by presenting information sequentially. Each screen addresses a different objection. The narrative structure builds momentum toward the conversion decision. Mimics web landing page best practices adapted for mobile.

**When to use it.** Apps with complex value propositions (education, productivity, professional tools) where a single screen cannot adequately communicate value. Less suited for simple utility apps where speed-to-CTA matters more.

**Apps doing it.** Mimo (60% lift), Noom (multi-step quiz flow), OMENA (scrollable long-form, trial starts doubled)

**Evidence strength.** Strong. Confirmed by a large dataset or a controlled test. Primary source from Gauchet's direct curation.

**Source.** Sylvain Gauchet, Sub Club podcast (subclub.com/episode/how-to-build-more-successful-paywalls-sylvain-gauchet)

#### Animated paywall vs. static

**What it is.** Using animated elements (Lottie animations, video backgrounds, motion graphics) on the paywall screen rather than static images and text.

**What the data shows.** CoinStats (crypto portfolio tracker) achieved **2.9x higher conversion rate** on average with animated paywall vs. standard, across iOS and Android. The animations were meme-themed Lottie animations (Pepe the Frog for Black Friday, Giga Chad Santa for New Year) played before showing the paywall. FitnessAI: adding video to the paywall drove **2x install→trial** (Jake Mor / Superwall).

**Why it works.** Animations capture attention in a medium where users scroll quickly. Video and motion demonstrate the product better than screenshots. The CoinStats case also involved culturally targeted memes (Pepe in crypto community) and seasonal discounts, these confounds limit generalizability.

**When to use it.** Apps with visual outputs (photo/video editing, design tools) where the product can be demonstrated. Culturally targeted animations may work in niche communities. Generic gradient animations likely have minimal effect.

**Apps doing it.** CoinStats (2.9x), FitnessAI (2x with video)

**Evidence strength.** Weak to mixed. Directionally supported at best. For generalizability, CoinStats explicitly acknowledged no control groups, confounding with discounts and seasonality. This is NOT a Nami ML finding as commonly misattributed.

**Source.** Tigran Mkrtchyan / Business of Apps (businessofapps.com/insights/paywall-optimization-reimagined), GrowthGems #65, InsightOut Substack (tigranmk.substack.com)

#### Price presentation: anchoring, badges, per-day pricing

**What it is.** Visual techniques for presenting prices including crossed-out original prices, "Most Popular" badges, "Save X%" callouts, and per-day pricing breakdowns.

**What the data shows.** Parízek's testing priority ranks "plan/price display" as more impactful than value proposition format. Per-day pricing display lifts annual adoption **20-40%** (industry reports, not a single controlled study). "Save X%" badges prominently on annual plans are near-universal in top-grossing apps. "Most Popular" badge nudges users toward the highest-LTV option. Superwall data: adding annual plan with 60% discount display alongside monthly drove **proceeds per user up 80%**. Adapty: median annual plan costs 3x monthly price, annual subscribers save ~75%, and this savings display is standard practice.

**Why it works.** Anchoring bias, the crossed-out price establishes a reference point that makes the actual price feel like a deal. Per-day pricing leverages denominator neglect ("$0.54/day" feels trivial vs "$199/year"). "Most Popular" badges use social proof to guide selection.

**When to use it.** Universal. Every paywall should include at least one of these techniques. Per-day pricing is most effective for annual plans. "Save X%" is most effective when the savings percentage is >30%.

**Apps doing it.** Near-universal across top-grossing apps (BetterMe, Headspace, Calm, Duolingo all use some combination)

**Evidence strength.** Good. Consistent across several independent reports. Widely documented but controlled experiment data on individual treatments is sparse.

**Source.** Parízek guide, Superwall conversion stats, Adapty design patterns

#### "Honest paywall" / trial timeline transparency

**What it is.** A timeline visual showing the trial journey explicitly: Day 0 (start trial), Day 5 (reminder notification), Day 7 (charge begins). Makes cancellation process and timing transparent.

**What the data shows.** Blinkist reported **23% increase in conversion rate** and **55% drop in complaints** after implementing the honest paywall timeline. Strava adopted a similar 30-day trial timeline. Five Minute Journal's onboarding redesign using a JTBD framework produced **+20% ARPU, +10% trial start rate, +14% trial conversion**.

**Why it works.** Addresses the #1 user fear: forgetting to cancel a trial and being charged. Transparency builds trust, reducing friction at the point of purchase. The timeline creates a goal-gradient effect (the journey has "already started"). Counter-intuitively, making cancellation easier makes starting easier.

**When to use it.** Apps targeting mainstream/non-tech-savvy audiences, wellness apps, education apps. Particularly effective when the trial-to-paid conversion bottleneck is "fear of being charged" rather than "lack of perceived value."

**Apps doing it.** Blinkist (23% conversion lift, 55% fewer complaints), Strava (30-day timeline), Five Minute Journal

**Evidence strength.** Strong. Confirmed by a large dataset or a controlled test. Blinkist data is from their Head of Product (Eveline Moczko) via multiple primary sources.

**Source.** growth.design/case-studies/trial-paywall-challenge, revenuecat.com, purchasely.com

#### "Ugly paywall" / simplified design

**What it is.** Stripping the paywall of decorative elements (social proof, awards, detailed feature lists, gradient backgrounds) in favor of a clean, minimal design: big headline, few bullet points, clear CTA, often white background.

**What the data shows.** Golf GameBook switched from dark green to white background and saw an "unexpectedly good jump in trial conversion rate" (no specific number). Strava and Fishbrain both use radically simplified paywalls that outperform more complex versions. RevenueCat's blog documents the "ugly paywall" trend as a counter-movement to over-designed paywalls. RevenueCat case study: trial toggle + short-form redesign → **+31% install-to-trial, +64% revenue uplift** for a party game app.

**Why it works.** Over-optimized paywalls accumulate elements over time, creating visual noise and decision fatigue. Simplification reduces cognitive load and focuses attention on the CTA. Particularly effective after extensive A/B testing has added many elements, sometimes the best test is to remove everything and start fresh.

**When to use it.** Apps that have been iterating on paywall design for a long time and experiencing diminishing returns. Good reset strategy. Not a permanent best practice, rather a useful intervention.

**Apps doing it.** Strava, Fishbrain, Golf GameBook

**Evidence strength.** Mixed. Directionally supported, not proven. Conceptual with specific examples but limited quantified data.

**Source.** revenuecat.com

### Personalization and dynamic paywalls

#### ML-powered dynamic paywalls (product selection)

**What it is.** Machine learning models predict which product (plan, tier, duration) each user is most likely to purchase, then surface only that product on the paywall. Addresses decision overload by showing fewer, more relevant options.

**What the data shows.** Tinder's ML-powered dynamic paywall (matching users to the optimal product from 3 tiers + multiple durations + à la carte options) produced **"multimillion-dollar annual revenue increase"** (Shawn Gong, Sub Club podcast). Superwall's AI Demand Score (trained on 100M+ monthly paywall views) reportedly delivers **20-30% revenue lifts** for apps at $5-10M MRR. Adapty data: dynamic/segmented paywalls deliver approximately **35% higher conversion rates** than static alternatives. Adapty reports apps running 50+ experiments earn **18.7x more revenue** than single-experiment apps.

**Why it works.** Different users have different price sensitivity, feature preferences, and commitment timelines. A one-size-fits-all paywall is suboptimal for most segments. The ML model uses signals (device type, location, WiFi vs cellular, time of day, prior interactions) to predict purchase likelihood and optimal offering. Shawn Gong: "The real unlock wasn't better pricing, it was better decision design."

**When to use it.** Apps at scale ($1M+ MRR) with multiple products/tiers and dedicated growth teams. Smaller apps should focus on segmented paywalls by country/platform before investing in ML.

**Apps doing it.** Tinder (Shawn Gong), apps using Superwall Demand Score

**Evidence strength.** Mixed. Directionally supported, not proven. Tinder confirms "multimillion" but doesn't give exact figures. Superwall's 20-30% claim is self-reported.

**Source.** Shawn Gong Sub Club episode (subclub.com/episode/dynamic-paywalls-that-drove-millions-in-new-revenue-shawn-gong-tinder), Superwall Demand Score documentation

#### Personalized paywall (user data insertion)

**What it is.** Inserting user-specific data into paywall copy (first name, goals from onboarding quiz, usage data) to create a personalized experience.

**What the data shows.** Nami ML claims personalizing paywall with user's first name can improve conversion **"up to 17%."** However, **no primary study, sample size, app, or methodology is cited**, this is an unattributed marketing claim. Mojo's geo-personalized layouts (different designs for Japan, Latin America, US) produced measurable lifts (+20% Japan, +45% Brazil for pricing framing). Adapty reports localization (a form of personalization) has the **highest experiment win rate at 62.3%**.

**Why it works.** Personalization increases perceived relevance ("this is for me"). Name insertion is the simplest form. Goal-based personalization (showing outcomes aligned with quiz responses) is more impactful but requires onboarding data collection.

**When to use it.** All apps that collect user data during onboarding. Start with geo-personalization (highest win rate), then add goal/preference personalization, then test name insertion.

**Apps doing it.** Noom (quiz-based), Mojo (geo-based), Calm (use-case-based: sleep vs. meditation users have different WTP)

**Evidence strength.** Weak. Treat as a hypothesis. For the +17% claim specifically (unverified). Strong for geo-personalization and goal-based personalization as categories.

**Source.** Nami ML blog (namiml.com/blog), Parízek guide, Adapty 2026

#### Paywall frequency and fatigue management

**What it is.** Controlling how often and in what contexts users see paywalls to maximize conversion while minimizing fatigue and negative sentiment.

**What the data shows.** Conversion degrades meaningfully beyond **~2 paywall exposures per user per session** (industry practitioner reports). Mojo's app-open campaign paywall generated 15% of iOS revenue with "minimal complaints", suggesting moderate frequency works. Parízek recommends: discounts for users at 30+ days who haven't converted "work especially well," while exit discounts and win-backs "didn't perform as strongly" at Mojo. Adapty data: **44.5% of all purchases happen on Day 0**, supporting front-loaded exposure.

**Why it works.** Early exposures capture high-intent users. Repeated exposure to the same paywall creates banner blindness. Creative rotation (different designs, copy, offers) and frequency capping maintain effectiveness. Segmenting by days-since-install allows escalating urgency without fatiguing fresh users.

**When to use it.** All apps. Implement frequency capping from day one. Rotate creative every 2-4 weeks.

**Apps doing it.** Mojo (app-open campaigns), Duolingo (progressive feature locking creates natural frequency)

**Evidence strength.** Mixed. Directionally supported, not proven. Based on practitioner reports rather than published studies.

**Source.** Parízek guide, Adapty 2026, industry practitioner consensus

### Vertical-specific patterns

#### Fitness: aggressive onboarding quiz with web-to-app funnel

**What it is.** Extended multi-step quiz (40-96+ steps) on web, generating a personalized plan, leading to web checkout (avoiding App Store fees), then directing to the app for fulfillment.

**What the data shows.** H&F leads all categories in install LTV (**$1.21** Adapty), D60 RPI (**$0.66** RevenueCat), and trial-to-paid conversion (**35%** Adapty). Annual plans dominate (60.6% of revenue, growing). Trial LTV premium in H&F: **+63.6%** (second highest after Utilities). BetterMe: ~$2M/month revenue, $500K downloads/month. Noom: $750M-$1B/year. Web checkout keeps 94-97% vs 70% through App Store.

**Why it works.** High emotional intent at download (health anxiety, New Year's resolution, body dissatisfaction). Quiz builds commitment and captures segmentation data. Web-first avoids the 30% platform tax.

**When to use it.** Health & Fitness apps with content-driven value propositions (programs, plans, coaching).

**Apps doing it.** Noom ($750M-$1B annual), BetterMe ($2M/month), Fastic, Yazio, Runna (web-to-app added Dec 2024 beside the app flow: a core web flow with alternates by traffic warmth, a CMS so the ambassador team personalizes the landing flow per ambassador, and higher trial start, trial conversion and retention than the app flow in the first five months, direction only; it also opened Google Search as a channel: its Head of Growth in a 2026 podcast interview)

**Evidence strength.** Strong. Confirmed by a large dataset or a controlled test.

**Source.** Adapty 2026, RevenueCat 2026, screensdesign.com, retention.blog

#### Education: delayed conversion with honest trials

**What it is.** Education apps see uniquely delayed conversion patterns, with nearly a quarter of users starting trials after Day 31. Successful education paywalls emphasize transparency, multi-screen benefit explanations, and longer trial periods.

**What the data shows.** Education shows only **71.3% Day 0 trial starts** (vs 89.4% average) and **23.5% at Day 31+**, the highest late-start rate of any category. Education is the most discount-heavy: **14.3% of transactions** involve discounts (up from 8%). Duolingo: 128M MAU, **11M paid subscribers** (~8.8% penetration), $210.7M subscription revenue in Q2 2025 (46% YoY growth). Blinkist: **23% conversion increase** from honest paywall timeline. Mimo: **60% lift in trial opt-ins** from multi-screen paywalls. Trial LTV premium in Education: **+50.4%**.

**Why it works.** Learning has a longer value-demonstration period, users need time to feel progress. The "aha moment" takes days/weeks, not minutes. This is why Education is the one category where delayed paywalls and longer trials consistently outperform immediate conversion pressure.

**When to use it.** Language learning, skill development, reading/summary apps.

**Apps doing it.** Duolingo (freemium with progressive locking, 750+ A/B tests/quarter), Blinkist (honest paywall, 23% lift), Mimo (multi-screen, 60% lift), Babbel

**Evidence strength.** Strong. Confirmed by more than one large dataset. Public company data + multiple primary sources.

**Source.** Adapty 2026, Duolingo Q2 2025 earnings, Growth.design case studies, Gauchet Sub Club

#### Wellness/meditation: content gating evolution

**What it is.** Wellness apps have evolved from generous free tiers (designed to build habit) toward aggressive content locking, finding that users expect paywalls and that locking more content drives higher conversion.

**What the data shows.** Headspace: **~2.8M paid subscribers**, moved from 20% free content to near-complete lock, reporting "double-digit conversion improvement." Monthly plan loyalists exist because monthly billing acts as engagement reminder. Calm: **~4M subscribers**, ~$150M+ revenue. Habitual users (60+ min meditation) willing to pay **$22-35/month** vs occasional users at $4.90-5.90, a **4-7x WTP spread** within the same app.

**Why it works.** Subscription app users are habituated to paywalls from Spotify, Netflix, etc. The fear that locking content will cause backlash is typically unfounded. Content updates (Calm's celebrity sleep stories) drive retention. The WTP spread within wellness apps suggests significant personalization opportunity.

**When to use it.** Meditation, sleep, mindfulness, journaling apps.

**Apps doing it.** Headspace (content gating evolution), Calm (celebrity content moat), Rootd (5x from paywall placement)

**Evidence strength.** Strong. Confirmed by a large dataset or a controlled test. First-party interview data from Headspace product leads.

**Source.** Headspace Sub Club episode (subclub.com), businessofapps.com, sbigrowth.com

#### Dating: entry point multiplication and tier laddering

**What it is.** Adding multiple premium access points throughout the app and building subscription tiers incrementally over time, rather than relying on a single paywall moment.

**What the data shows.** Jeff Morris Jr's experiment at Tinder: adding a **toggle in the top navigation bar** for subscription access generated **$50M in revenue** from a single change that was rated internally as "small to medium" priority. Tinder evolved from a single tier (Plus) to three tiers (Plus → Gold → Platinum), with Gold and Boost helping increase Match Group market cap by **$5-8B+**. Tinder Select ($500/month ultra-premium) **failed**, users didn't feel the environment matched the premium price. **85-90% of Tinder users never paid**, free users ARE the product.

**Why it works.** Dating apps have inherent urgency (matching is time-sensitive). Multiple entry points catch users at different moments of frustration or desire. Tier laddering captures users at different WTP levels. The $50M experiment worked because it exposed an untapped piece of premium real estate to high-intent users.

**When to use it.** Social/dating apps with large free user bases.

**Apps doing it.** Tinder ($50M from nav toggle), Hily ($3M/month)

**Evidence strength.** Strong. Confirmed by more than one large dataset. Verified first-person account with specific mechanism and dollar figure.

**Source.** Jeff Morris Jr Sub Club Ep. 148 (subclub.com/episode/how-a-single-paywall-experiment-generated-50m), Shawn Gong Sub Club episode

#### AI apps: premium pricing with fast cashflow models

**What it is.** AI-powered apps command premium pricing and drive higher per-user revenue but face faster churn, leading to strategies optimized for rapid cashflow recovery.

**What the data shows.** RevenueCat: AI apps generate **41% more revenue per payer** (Y1 RLTV: $30.16 vs $21.37) but churn **36% faster** on monthly plans. AI apps match H&F at **$0.63 RPI after 60 days**. Adapty: AI app install LTV is **$1.44** vs $0.84 average (+70%), but install-to-trial is only **5.31%** (vs 10.92% average, roughly half). AI apps drive **14% more direct purchases** (4.20% vs 3.71%), suggesting users who want AI tools are more decisive. Lifestyle AI apps saw **691% YoY revenue growth**.

**Why it works.** AI features have clear, demonstrable utility that justifies premium pricing. But novelty wears off faster than habit-forming categories. The strategy: price high, convert fast, accept higher churn, optimize for short-term ROAS.

**When to use it.** AI-powered creative tools, writing assistants, image generators, coding assistants.

**Apps doing it.** ChatGPT (RevenueCat Enterprise customer), AI photo editors, AI writing tools

**Evidence strength.** Strong. Confirmed by a large dataset or a controlled test. Confirmed by both benchmark reports.

**Source.** RevenueCat 2026, Adapty 2026

### Second chances, entry points and offer framing

#### Second-chance offer sequence

**What it is.** Treating the first paywall as one touch in a sequence rather than the decision. After a decline, the next ask changes what is offered or how it is sold: a short "why not today?" survey routes price objections to a discount or a lifetime plan and value doubts to more explanation; a time-boxed welcome offer (2-3 days after install) is shown at each app open; a lifetime plan is offered later to free users and weekly or monthly payers only.

**What the data shows.** No controlled figures published. The handbooks' premise is that 90%+ of users do not convert on the first paywall and that purchase likelihood is highest in the first 24 hours to a few days. Exit and post-paywall offers already carry a large share of subscriptions at some apps (see post-paywall welcome and exit offers above).

**Why it works.** Users differ in willingness to pay and in what stopped them. A second ask fitted to the stated reason recovers users a single generic paywall loses. Lifetime resets the frame for users averse to recurring charges.

**When to use it.** Hard-paywall and onboarding-paywall apps with high dismiss rates. Keep lifetime away from the first paywall for annual-intent users: it cannibalizes annual.

**Apps doing it.** Exit-offer flows across Superwall users; Sven Jürgens frames it as "think about paywalls like ad impressions"

**Evidence strength.** Mixed. Directionally supported, not proven. Consistent practitioner advice, no published control data. Mojo found no-offer sometimes beats an exit discount, so test offer against no offer.

**Source.** Botsi, 48 Laws of Subscription App Success (2026), Baghdasaryan and Rushfinn, The Subscription Optimization Handbook

#### Transaction abandonment and failed-transaction recovery

**What it is.** Two distinct high-intent triggers. Abandonment: the user tapped buy or start trial and stopped at the store sheet. Failure: the user tried to pay and the payment did not go through. Abandoners get a fast follow-up (in-app message if still in session, push or email right after exit) with a reason to finish: a limited discount, a longer trial or the benefit they have not seen. Failures get a retry prompt, an email with the fix, and a paywall next session that acknowledges the failed attempt.

**What the data shows.** Roughly half of users who start an in-app purchase complete it (Jake Mor). No published lift for the recovery flows themselves.

**Why it works.** Both groups already chose to buy. Hesitation and friction, not interest, are what failed, so a generic paywall wastes the strongest signal in the funnel. Speed matters: the sooner the follow-up, the higher the recovery.

**When to use it.** Every subscription app with a CRM or push connection. Requires the abandon and failure events to be tracked as properties the messaging tool can act on.

**Apps doing it.** Superwall's transaction-abandon flow

**Evidence strength.** Mixed. Directionally supported, not proven. Mechanism is sound and widely practiced, no published controlled result.

**Source.** Baghdasaryan and Rushfinn, The Subscription Optimization Handbook

#### Entry-point paywalls: matched content and matched price

**What it is.** The paywall a user sees depends on where they hit it. Tapping a locked feature opens a paywall about that feature, not the generic one. The price can follow the entry point too: a paywall reached from a high-value feature carries the standard or a higher price, one reached from a low-interest surface such as Settings leads with a trial.

**What the data shows.** No published figures. Calm customizes the paywall to the locked content selected; Instasize varies price by entry point from feature-preference research.

**Why it works.** The tap is a declared interest. Selling the feature the user just reached for converts that interest; pricing to the value of the entry point charges more where willingness to pay is demonstrably higher.

**When to use it.** Freemium and soft-paywall apps with several distinct premium features. Use MaxDiff or feature-usage data to rank entry points before pricing them differently.

**Apps doing it.** Calm (content-specific paywalls), Instasize (price by entry point)

**Evidence strength.** Weak to mixed. Directionally supported at best. Named examples, no published results.

**Source.** Baghdasaryan and Rushfinn, The Subscription Optimization Handbook

#### Ad keyword to paywall continuity

**What it is.** Carrying the ad's promise past the store page. A user acquired on a keyword or ad set ("background remover") lands on the matching Custom Product Page and then sees a paywall headlining that same feature, benefit and plan. Needs attribution data (an MMP or Apple Search Ads attribution) available when the paywall is chosen.

**What the data shows.** No published figures. Related evidence: swapping a generic "Unlock Premium" headline for the winning ad's promise moved trials and subscriptions (Price Power Playbook, de Baillenx).

**Why it works.** Message match. The promise that earned the install is the one most likely to earn the purchase; most apps drop it at the store page.

**When to use it.** Apps running keyword-specific Apple Search Ads or several distinct ad angles with Custom Product Pages.

**Apps doing it.** Background-remover example in the Tangent handbook

**Evidence strength.** Mixed. Directionally supported, not proven. Strong mechanism with adjacent evidence, no direct published test.

**Source.** Baghdasaryan and Rushfinn, The Subscription Optimization Handbook, Price Power Playbook 2026

#### Toggle trial with two prices

**What it is.** A paywall toggle lets the user choose trial or no trial, with the price differing by state: trial on costs slightly more, paying now costs less. The common mistake is the same price either way, which gives trial-averse users no reason to commit today.

**What the data shows.** No published figures for the two-price version. A trial toggle was one element of RevenueCat's bundled party-game redesign, so no lift can be attributed to it alone.

**Why it works.** Respects trial-averse users who do not want to remember to cancel, and prices the flexibility of the trial. Immediate payers are higher-intent and churn less after the first charge.

**When to use it.** Apps where a meaningful share of users buy without a trial. Apple requires clear disclosure of what is charged and when, so state the terms under each toggle state plainly.

**Apps doing it.** None named with results

**Evidence strength.** Weak. Treat as a hypothesis. Single practitioner recommendation.

**Source.** Baghdasaryan and Rushfinn, The Subscription Optimization Handbook

#### Choose your price

**What it is.** The user picks from a small fixed set of prices (lower, standard, supporter), framed as fairness or mission ("choose what feels fair to you"). Never an open range.

**What the data shows.** No published figures.

**Why it works.** Agency and self-image: many users avoid the lowest option so as not to feel cheap, and the choice makes the brand read as value-aligned.

**When to use it.** Mission-led or community brands, wellness, education. A poor fit for utilities where price is the decision.

**Apps doing it.** None named with results

**Evidence strength.** Weak. Treat as a hypothesis. Practitioner suggestion, untested in the source.

**Source.** Botsi, 48 Laws of Subscription App Success (2026)

#### Risk reversal and price framing copy

**What it is.** Four copy devices around the price. A plain refund line ("not for you? get your money back"), since many users do not know Apple refunds. The monthly cost compared to an everyday purchase this audience makes. Specific discount numbers (37%, 41%) rather than round ones, checked so the final price ends in.99. Decline copy that states the consequence ("Continue with ads") rather than a bare close.

**What the data shows.** No published figures. Simple shows a refund guarantee; Calm uses specific-number discounts.

**Why it works.** Lower perceived risk, anchoring to a familiar cost, and precision read as a considered offer. Consequence copy makes the trade-off of declining visible.

**When to use it.** All paywalls; the refund line especially where trial anxiety is high. Keep decline copy neutral and factual: guilt-trip wording ("No, I don't want to get fit") is confirmshaming, a trust and review risk.

**Apps doing it.** Simple (refund guarantee), Calm (specific discounts)

**Evidence strength.** Weak to mixed. Directionally supported at best. Established pricing psychology, no app-specific results published.

**Source.** Baghdasaryan and Rushfinn, The Subscription Optimization Handbook

#### Earned offers: qualification, claimable trials, reveals

**What it is.** The offer is framed as something the user earned or picked up rather than something pushed at them. Qualification: "based on your answers, you qualify for a free trial", time-limited. Claimable: a trial card on the home screen or in a notification that must be tapped to claim. Reveal: a story-style sale offer ("reveal your discount") the user can skip through. Lucky draws (spin for an offer) use the same effect with chance.

**What the data shows.** No published figures. Noom uses qualification framing; Flo uses an interactive reveal for sale offers.

**Why it works.** The endowment effect: people value what they feel they own or earned. A guaranteed, revealed offer keeps the effect without the downside of chance; for referral rewards, guaranteed rewards beat sweepstakes (Price Power Playbook, Scarpaci).

**When to use it.** Freemium apps without a standard trial (qualification), sale campaigns that must not look like the usual offer (reveal).

**Apps doing it.** Noom, Flo

**Evidence strength.** Weak to mixed. Directionally supported at best. Named examples and a well-studied principle, no published app results.

**Source.** Baghdasaryan and Rushfinn, The Subscription Optimization Handbook, Botsi, 48 Laws of Subscription App Success (2026)

#### Freemium upgrade surfaces

**What it is.** Making the paid tier visible everywhere a free user goes, not just on the paywall. Lock icons on premium items in lists, menus and cards; upgrade CTAs in settings, profile, feature screens and navigation; "Free Edition" in the header. A second free trial for lapsed trialists who come back on their own.

**What the data shows.** Lock icons alone increased free-to-paid conversion at Mojo (no figure published). Tinder's top-nav entry point generated about $50M (Jeff Morris Jr., Sub Club Podcast).

**Why it works.** Visible scarcity reminds free users what they are missing, and each contextual CTA catches a user at the moment of highest intent. A returning trialist may have missed the aha moment the first time; a second trial restarts intent without a price cut.

**When to use it.** Freemium apps. Size the second-trial idea from the share of active free users who already used a trial.

**Apps doing it.** Mojo (locks), Tinder (top-nav entry point)

**Evidence strength.** Mixed. Directionally supported, not proven. Consistent with the Tinder entry-point result, one named lock result without a figure.

**Source.** Botsi, 48 Laws of Subscription App Success (2026)

#### Consumable credits beside the subscription

**What it is.** A one-time credit pack alongside the subscription for users unlikely to subscribe ($15 for 150 photo edits), and paid extras for power users (premium support, extra content).

**What the data shows.** No published figures. Headway and Simple upsell extras to engaged subscribers.

**Why it works.** Subscription aversion is real in some segments and cultures, especially in content and creative apps. Credits capture revenue a subscription-only model forgoes; extras raise LTV from users already getting the most value.

**When to use it.** Photo, video and content apps; AI apps with usage-based cost. Offer credits to segments that have declined the subscription, not on the first paywall.

**Apps doing it.** Headway, Simple (upsells)

**Evidence strength.** Weak to mixed. Directionally supported at best. Practitioner pattern, no published results.

**Source.** Baghdasaryan and Rushfinn, The Subscription Optimization Handbook

#### Monetize beside the loop, never on it

**What it is.** In a freemium app, the action that brings the next user in (inviting a friend, sharing or publishing content, creating what others find) stays reachable on the free tier: a lock on the invite or share control is the failure, while a free allowance of creation (a daily count, starter credits, a set number of free uses) meters a use with real cost rather than gating it. What is gated is depth, power-user capacity or cosmetics.

**What the data shows.** No controlled figures. Discord gates file-upload size, custom emojis and profile effects and never the friend invite; a Photomath competitor that limited free scans ended, in Photomath's former product head's estimate, at roughly an order of magnitude lower value. Quizlet never gated set creation, the action that fed its SEO loop. ClassDojo charges for Memories albums and "never blocks learning"; Margins keeps everything social free and plans streak freezes and profile cosmetics; Opal's referral and leaderboard sit outside Pro; Suno's song page needs no login to listen; Grammarly's extension was "a crime to paywall".

**Why it works.** The free user who cannot invite, share or create is a cost; the one who can is acquisition. Gating the loop converts a little now and starves the top of funnel the model depends on; metering it (five songs a day at Suno, five live games a year at GameChanger) keeps the loop open while the cost is paid.

**When to use it.** Any freemium app whose growth depends on free users bringing the next user in (a social product, a marketplace side, user-made content that others find). A hard-paywall utility has no free loop to protect and gates at the door instead.

**Apps doing it.** Discord, Quizlet, ClassDojo, Margins, Opal, Suno, Grammarly; Gamma's badge and Photoroom's watermark are the same rule from the other side, a sharing surface that keeps a free viral state beside the paid clean one.

**Evidence strength.** Mixed. Directionally supported, not proven. Seven consistent operator accounts, no controlled comparison.

### What is not known, and what is misattributed

The gaps in the public evidence on paywalls, so an audit does not claim more than anyone can, and the figures most often pinned on the wrong source.

#### Paywall design → long-term LTV by vertical

No public source connects specific paywall design elements (layout, copy, visual treatment) to long-term LTV by vertical. RevenueCat's data connects **architecture** (hard vs. soft, trial vs. no trial) and **plan type** (weekly vs. annual) to LTV, but not visual design. Adapty comes closest with their experiment win-rate analysis showing that visual/text changes have the **lowest win rate (34.6%)** while localization and trial structure have the highest. The implication: design matters less than structure, but we lack granular data on which specific design elements drive LTV (not just conversion).

**Closest sources:** Adapty 2026 experiment win-rate analysis; Parízek's testing priority framework (pricing > layout > copy > imagery).

#### AI-personalized paywalls beyond basic claims

Superwall's AI Demand Score is the most documented AI-personalization system, reporting 20-30% revenue lifts at scale. Tinder's ML model (Shawn Gong) is confirmed but lacks specific numbers. Nami ML's +17% claim is **unverified**, no study, sample size, or methodology cited. Adapty has "AI forecasting" for experiments but not real-time per-user paywall personalization. No peer-reviewed or rigorously documented study exists on AI-personalized paywalls with control groups. This is the most significant evidence gap in the paywall optimization space.

**Best available:** Superwall Demand Score documentation (superwall.com/features/ai-pricing-demand-score); Shawn Gong Sub Club episode.

#### The Jeff Morris Jr. $50M experiment (verified)

**The experiment was a paywall placement change, not a pricing experiment.** In ~2018, Tinder (30-40M MAU at the time) added a toggle in the top navigation bar that allowed users to enter a new subscription tier directly. This previously untapped piece of premium real estate generated $50M (likely annualized). The design team initially resisted the top nav addition. It was rated internally as a "small to medium" experiment. Jeff Morris Jr.'s lesson: "Revenue teams should be a bit more aggressive in terms of pricing and entry points." His caveat: "If you tried putting a top nav into a meditation app, I'm not sure it would convert as high", scale and product category matter enormously.

**Source:** Jeff Morris Jr, Sub Club Podcast Ep. 148 (Jan 7, 2026), subclub.com/episode/how-a-single-paywall-experiment-generated-50m.

#### Nami ML animated paywall 2.9x (misattributed)

The 2.9x animated paywall conversion claim is **not from Nami ML**. The primary source is Tigran Mkrtchyan, Head of Digital Marketing at CoinStats, published on Business of Apps and his InsightOut Substack. The experiment used meme-themed Lottie animations during seasonal promotions (Black Friday, New Year) with significant discounts. The author explicitly acknowledged: "We did not use control groups." Discounts, seasonality, and a crypto-specific audience (Pepe memes) all confound the result. The 2.9x number is real but not generalizable.

#### Paywall screenshot databases lack performance data

PaywallPro's claim of "46K+ paywalls with performance data" is **misleading**. The platform has 46K+ screenshots with estimated app-level revenue metrics (ARPU, RPD) derived from third-party App Store analytics, not actual conversion rates or A/B test results tied to specific paywall designs. PaywallScreens.com (run by Superwall) similarly provides screenshots with estimated monthly revenue but no paywall-specific performance data. No public database connects specific paywall designs to actual conversion data.

#### Trials hurt LTV in half of app categories

Adapty's finding that trials produce **negative LTV impact** in Productivity (-13.7%), Lifestyle (-21.2%), Entertainment, and Graphics & Design is potentially the most important underreported finding in subscription optimization. RevenueCat's data does not directly contradict this but presents trials more positively overall. This discrepancy may reflect sample composition differences. Apps in these four categories should A/B test trial vs. no-trial rather than defaulting to trial offers.

#### Experiment sequencing matters more than individual treatments

Adapty's experiment win-rate data and Parízek's testing priority framework both point to the same conclusion: most teams test in the wrong order. The optimal testing sequence is (1) **pricing/trial structure** (59.6% LTV win rate), (2) **localization** (62.3%), (3) **plan count/duration** (58.7%), then (4) visual/copy (34.6%). Most teams start with visual changes (lowest win rate) because they're easiest to execute. Adapty estimates pricing experiments deliver **up to 80% uplift** while visual optimization delivers up to 30%. This sequencing insight could be the single most actionable recommendation in the library.

---
---

## Onboarding and first-session patterns with measured results

The named-company cases behind onboarding design and the first session, each with its evidence type. These decide whether the trial that just started converts.

### Onboarding design patterns

#### Noom's extended psychological onboarding

*Noom · Health / Weight Loss*

**What it is.** **Approximately 60-step onboarding** that functions as both a data collection and psychological commitment mechanism. Uses battery of assessments to categorize users into "diet psychology types," creates identity labels ("Noom Novice" from lesson 1), and sets the "10 minutes a day" expectation. North Star Metric: Quality Adjusted Life Years (QALYs), not revenue.

**What the data shows.** Artem Petakov (co-founder) described the onboarding as designed to "visualize the outcome and take ownership of that outcome." The extended quiz itself creates sunk cost and perceived personalization (leveraging the Barnum effect). Revenue grew **400% to $200M+** in 2019. No specific activation metric (e.g., "log food 3 days in a row") has been publicly shared.

**Evidence strength.** A way of thinking, not a measured result. Psychologically grounded approach; growth numbers are company-reported, not experimentally attributed to onboarding specifically.

**Worth knowing.** Noom's onboarding is an extreme example of front-loading activation before monetization, the 60-step quiz happens before the paywall. This directly addresses the timing paradox by creating psychological investment before asking for money. Casey Winters' principle "you CAN add friction if it increases value delivery" applies here. The approach works only when the friction itself delivers value (perceived personalization, commitment building).

**Source.** Artem Petakov, Breakout Growth Podcast, February 2021.

#### Reading.com's web funnel superiority

*Teaching.com / Reading.com · Education (Children's Learning)*

**What it is.** **Identical onboarding replicated on web vs. in-app.** Web funnel achieved **50% better conversion to trial start** and **30% better conversion to paid** than the same onboarding in the app.

**What the data shows.** Tim Dikun (COO) stated: "I think it was 50% better conversion to trial and then a 30% better conversion to paid than the same exact onboarding in the app, which was just astounding to me." Methodology was a 1:1 replication of existing in-app onboarding to a web funnel, then traffic was sent to both. Additionally, replacing free trial with a **30-day money-back guarantee** on web funnels yielded lower initial conversion but higher-quality users and better ad network signals. Price testing found **$12.49/month** as the LTV-maximizing price point (tested from $4.99 to $19.99), still winning after 2+ years. Removing the monthly trial option to push annual plans caused a **17% drop in realized LTV**.

**Evidence strength.** Strong. A/B tested. Controlled A/B test of identical onboarding across web vs. App.

**Worth knowing.** The web funnel finding has direct timing paradox implications, web separates the purchase decision from the app experience, potentially reducing the compression of activation and monetization into a single session. The money-back guarantee further decouples: payment happens first, but risk is reduced. The 17% LTV drop from removing monthly trials is a cautionary finding against over-optimizing for annual conversion.

**Source.** Tim Dikun, Sub Club podcast, "What Reading.com Learned Testing Prices and Funnels," May 28, 2025.

#### Blinkist's honest paywall

*Blinkist · Education / Productivity*

**What it is.** Redesigned paywall with transparent trial-end reminders (Day 5 notification before Day 7 trial expiration) and clear communication about what happens after trial ends.

**What the data shows.** A/B tested. **23% increase in trial conversion rate. 55% fewer customer complaints. Push notification opt-in surged from 6% to 74%** (a 12x increase). **4% higher trial retention.** Separately, Braze-powered personalized "Daily Pick" content campaigns drove a **64% increase in subscription purchases** and **50% increase in sessions**.

**Evidence strength.** Strong. A/B tested.

**Worth knowing.** This finding challenges the assumption that aggressive paywalls maximize conversion. Building trust at the monetization moment improved both conversion and retention simultaneously. The push notification opt-in surge (6% → 74%) is particularly notable, by earning trust at the paywall, Blinkist unlocked a retention channel (push notifications) that compounds over the user's lifetime. Directly relevant to the timing paradox: honesty at the Day 0 monetization moment improved long-term outcomes.

**Source.** Jaycee Day (designer), UX Planet article, Braze case study.

### First-session design patterns

#### Tinder's sub-2-minute onboarding

*Tinder / Match Group · Dating*

**What it is.** **Facebook login + photo selection + short bio = swiping in under 2 minutes.** Pre-Tinder, Match/eHarmony required 20-30 minute profile creation with 20-100 questions. Only "serious" daters completed this, constraining the addressable market.

**What the data shows.** Ravi Mehta (former CPO, Tinder) described the impact: "Massively decreased friction... increased the total addressable market because all of these people, who had very limited willingness to pay with their time, could now actually get into a dating app." This brought in younger users who had never considered online dating. The aha moment is defined as "getting into a conversation with someone." Tinder is freemium, **85-90% of users are entirely free**; 10-15% pay across tiered plans plus consumables. Mehta also referenced **Darius Contractor's "Psych Framework"** which tracks positive/negative "psych points" through onboarding (email verification = -5, credit card = -10, value moment = +5).

**Evidence strength.** A way of thinking, not a measured result. No specific numerical activation metric shared; the <2-minute onboarding impact is anecdotal/observational.

**Worth knowing.** Tinder sidesteps the timing paradox through freemium, monetization serves users who want more from an already-valuable free experience. Mehta's counter-example is instructive: **Sesame Care (weight loss) uses a 25-step onboarding that reportedly boosted conversion 40%** (second-hand, no methodology; use the device, not the number) because it built confidence for a high-stakes decision. The lesson: friction reduction is not universally correct. The right amount of friction depends on whether it builds confidence toward the aha moment.

**Source.** Ravi Mehta, Sub Club podcast, "How Tinder Captures More Value with Tiered Pricing and Consumables," late 2025.

#### Duolingo's delayed sign-up

*Duolingo · Education*

**What it is.** **Pushing account creation until after the user completes their first test lesson** rather than requiring signup before any product experience.

**What the data shows.** a roughly **20% lift in next-day retention**, as reported (secondary source; no baseline published, so no arithmetic on it). The insight: letting users experience the core value (a language lesson) before asking for commitment (account creation) dramatically increases the probability of return.

**Evidence strength.** Strong. A/B tested.

**Worth knowing.** This is a specific, actionable pattern directly transferable to B2C subscription apps. Note: the baseline 12% and 20% lift figures appear in secondary sources; primary Duolingo verification is limited.

**Source.** Referenced in Duolingo growth analysis, SaaS Factor reporting.

---

## The psychology behind the patterns

38 principles that bear on the store listing, onboarding, the paywall and pricing. Each one says what it claims, how to use it in a flow, and how it backfires. The backfire line is the one people skip, and it is where most dark patterns come from.

### Mostly relevant at: store listing

**Attentional Bias.** Users' current emotional state filters what they pay attention to, they notice what's already on their mind.

*Use it:* Ad targeting precision matters less than message-moment alignment. Serving a weight-loss ad to someone who just searched "how to lose weight" works because the ad matches an active mental preoccupation. Lifecycle messaging can use attentional bias by triggering messages at moments when users are most likely to be thinking about the problem the product solves.

*Backfires when:* Attentional bias works with the grain of existing thought. It can't manufacture interest that isn't there. If users have no active concern about your product's problem space, the bias has nothing to attach to.

**Confirmation Bias.** People seek out and trust information that confirms what they already believe.

*Use it:* Ads targeted at users who have already expressed interest in the problem space perform better because they're confirming something the user already suspects is true. Onboarding that frames the product's value in terms of what the user already believes about their problem converts better than messaging that asks users to adopt a new frame. First impressions set a confirmation bias trajectory, a great first experience makes users more likely to attribute subsequent positive outcomes to the product.

*Backfires when:* Confirmation bias can lock users into negative expectations if their first experience is poor. The bias amplifies whatever impression came first.

**Curiosity Gap.** Users feel compelled to close the space between what they know and what they sense they're missing.

*Use it:* Email subject lines, push notification copy, and ad headlines that open a gap outperform those that state conclusions. "The one thing most people get wrong about [topic]" outperforms "Here's how to do [topic]" because the first creates a gap; the second closes it before the user opens. Paywall design can use curiosity gaps to show blurred or teased premium content.

*Backfires when:* Gaps that feel manipulative rather than genuine ("You won't believe what happened") erode trust. The payoff must match the setup. If the revealed information is underwhelming, users feel tricked.

**Fresh Start Effect.** People are significantly more motivated to pursue goals at temporal landmarks, the start of a new year, month, week, or personal milestone.

*Use it:* High-intent acquisition windows for B2C subscription apps are not random. They cluster around fresh start moments, especially for products in health, learning, finance, productivity, and any self-improvement category. Winback campaigns sent around these moments significantly outperform generic reactivation. Campaigns that explicitly acknowledge the fresh start moment ("It's a new month, your streak restarts at zero") meet users where their motivation is already elevated.

*Backfires when:* Fresh start motivation is a peak, not a plateau. Retention strategy must convert the initial motivation spike into actual habit before the motivation normalizes, typically within 2-3 weeks. The window is short.

### Mostly relevant at: onboarding

**Aha! Moment.** The first moment a new user viscerally understands why they need your product.

*Use it:* Identifying and shortening the path to aha is one of the highest-leverage moves in onboarding. Most products take too long to get there, or bury it under setup friction. Every step before aha is a conversion risk. Every step after it compounds retention. For B2C subscription apps, this moment typically involves the user experiencing the core product value, not just understanding it conceptually.

*Backfires when:* Confusing aha with completion of onboarding steps. A user can finish the entire onboarding flow and never feel it. Aha is an emotional event, not a checkbox.

**Reciprocity.** Users who receive genuine value before being asked for anything are significantly more likely to say yes.

*Use it:* "Value before ask" is the structural application of reciprocity to onboarding and trial design. Products that ask for email, credit card, or commitment before delivering any value consistently convert worse than products that demonstrate value first. Free tools, free trials that start immediately, and "try before you buy" flows all operate on reciprocity.

*Backfires when:* Reciprocity requires genuine value. "Sign up and we'll show you a demo" doesn't create the obligation. The value must be real and immediate.

**Commitment and Consistency.** Once users commit to a small action, they feel compelled to stay consistent with it in subsequent decisions.

*Use it:* Onboarding flows that start with tiny commitments (selecting preferences, naming a workspace, answering a single question) unlock willingness to complete longer sequences. Habit-based products can use small daily actions to build behavioral consistency that makes cancellation feel like a break from identity. Trial-to-paid messaging can reference users' prior engagement ("You've done X, don't lose your progress").

*Backfires when:* Asking for too much too early triggers resistance instead of commitment. The first ask must feel genuinely low-stakes.

**Cognitive Dissonance.** When what users do conflicts with what they believe about themselves, they will change one or the other to restore consistency.

*Use it:* Onboarding flows that get users to state a goal ("I want to lose 10 pounds") create a commitment they will feel uncomfortable contradicting when they use the product. If the product then helps them take a small action aligned with that goal, it reduces dissonance and increases perceived value. Conversely, flows that feel manipulative or misaligned with users' self-image will generate dissonance that motivates churn.

*Backfires when:* Dissonance-inducing messaging (guilt-based retention) can cause reactance if overused. Use it to reconnect users to their own goals, not to shame them.

**IKEA Effect.** Users place significantly more value on things they partially built or configured themselves.

*Use it:* Onboarding flows that require users to make meaningful choices, name their workspace, select their goals, configure their preferences, build something, produce higher retention than flows optimized purely for speed. The effort creates ownership. That ownership raises switching costs: the user doesn't just lose access to a product, they lose something they built. Drift doubled their onboarding flow length to 12 steps and saw retention improve because users were building their own workspace.

*Backfires when:* Effort that feels pointless or that doesn't visibly affect the product creates frustration, not ownership. The effort must feel purposeful and the result must be visible.

**Labor Illusion.** Users value results more when they can see the work that produced them.

*Use it:* Onboarding flows that instantly produce a result (personalized feed, matched recommendations, generated plan) are often perceived as low-quality because users assume no work was done. Slowing down to show the process, "Analyzing your responses... Building your plan... Personalizing your experience", increases perceived value of the output and trust in the product.

*Backfires when:* Labor illusion with no underlying value is still worthless. If the output doesn't actually match what the process claimed to do, the disappointment is worse for having created high expectations.

**Zeigarnik Effect.** People remember and feel more compelled by incomplete tasks than completed ones.

*Use it:* Onboarding completion mechanics rely on this principle, partial progress bars create a pull to complete. Re-engagement emails that reference an incomplete action ("You set up your account but haven't done X") leverage the Zeigarnik effect directly. Incomplete content experiences (mid-way through a course, partially completed a goal) create a psychological obligation to return.

*Backfires when:* Incomplete tasks that users genuinely don't care about completing don't create tension, they create guilt or frustration. The incomplete task must matter to the user's own goals.

**Progressive Disclosure.** Show users only what they need for this moment. Reveal complexity as they're ready for it.

*Use it:* One of the most common onboarding design errors is front-loading features "so users know what they're getting." Users don't need to know everything during onboarding, they need to experience the one thing that will make them return. Feature education should follow activation and demonstrated value, not precede it.

*Backfires when:* Progressive disclosure that hides features users need from the start creates frustration. Disclose progressively relative to what users need, not relative to what's advanced from your perspective.

**Cognitive Load.** The more mental effort a user must spend to complete a task, the more likely they are to abandon it.

*Use it:* High cognitive load during onboarding is one of the most common and most fixable causes of D1 drop-off. Every unnecessary field, ambiguous label, or unclear next step adds load. Checkout flows with too many decisions lose users to decision paralysis. The goal is to minimize the processing required at each step, particularly at high-stakes moments.

*Backfires when:* Simplifying by hiding complexity vs. simplifying by removing it. Hiding it (accordion menus, "advanced settings") is sometimes necessary but does not reduce load if users need to find it anyway.

**Chunking.** People remember and process information much better when it's broken into meaningful groups.

*Use it:* Multi-step onboarding flows should be chunked into phases with clear progress markers. Long forms convert better when grouped into themed sections. Feature discovery is more effective when capabilities are introduced in logical clusters rather than individual function by function. Navigation with chunked categories (Stripe's approach: Products, Solutions, Developers) is easier to scan than exhaustive flat lists.

*Backfires when:* Chunks that are arbitrary or too large lose their benefit. Each chunk should represent a meaningful unit the user can hold in mind as a single concept.

**Hick's Law.** The more choices users face, the longer decisions take and the more likely they are to quit.

*Use it:* Audit any screen where users must choose from multiple options. Pricing pages with more than three tiers confuse. Navigation with more than five items dilutes. Onboarding dropdowns with 15+ options are conversion destroyers. The fix is usually progressive disclosure, smart defaults, or reducing options entirely.

*Backfires when:* Removing too many choices can also frustrate users who have a specific need not covered by the reduced set. The goal is not minimum options, it is minimum effort to find the right option.

**Occam's Razor.** The simpler solution is almost always better, not because simplicity is elegant, but because complexity has a cost.

*Use it:* The most common violation of Occam's razor in growth work is adding steps, features, or messages to fix a problem that the right simplification would solve better. Conversion optimization that adds more copy to a confusing page is almost never as effective as removing the confusion.

*Backfires when:* Simplicity that obscures necessary complexity is not better. Some decisions genuinely require detail. Occam's razor means do not add complexity unnecessarily, not that all complexity is bad.

**Familiarity Bias.** Users prefer experiences that feel like things they already know how to use.

*Use it:* Onboarding that uses familiar UI conventions (standard form behavior, recognizable icons, platform-native navigation patterns) reduces learning time and drop-off. Products entering a new category should borrow familiarity from adjacent categories users already understand. Superhuman's use of familiar keyboard shortcuts is a direct application.

*Backfires when:* Familiarity bias can make users resistant to genuinely better experiences that require a new mental model. Introduce novel patterns with explicit guidance and patience, don't assume good design is self-evident.

**Mental Model.** Users come with existing beliefs about how things work, and they'll use your product through that lens whether you design for it or not.

*Use it:* Every product is navigated through users' prior mental models. Understanding what those models are, what analogies users reach for to describe your product, is critical for onboarding copy, feature naming, and UX decisions. Products that work against users' existing models pay a heavy onboarding cost.

*Backfires when:* Users' existing mental models can be wrong in ways that lead to misuse. Simply designing for the existing model may not be enough, sometimes you need to explicitly correct it.

**Empathy Gap.** People systematically underestimate how much emotions drive their own and others' behavior.

*Use it:* Ad creative that leads with emotional resonance rather than functional features tends to outperform because it meets users in the emotional state they're actually in. Onboarding designed by product teams who know the product well routinely underestimates how disorienting the first session is for a completely new user. Churn surveys that ask "why did you cancel?" get rational answers, but the actual cancellation often happened in a moment of frustration, which is harder to articulate.

*Backfires when:* Designing for the emotional user and forgetting the rational one. Some decisions, especially high-consideration purchases, involve both, ads can lead with emotion, but the paywall must also satisfy rational scrutiny.

**Nudge.** Small, context-appropriate design cues can shift user behavior significantly without restricting choice.

*Use it:* The most effective nudges in growth work are invisible, they don't feel like persuasion because they match the user's own intentions and make the right path easiest. Airbnb nudging users to add dates and guests (making search results better) rather than forcing them is a benign nudge. Amazon defaulting to subscription (and discounting it) is a commercial nudge. The ethical distinction matters for trust.

*Backfires when:* Nudges that serve your goals at the expense of users' generate reactance when discovered. Design nudges toward outcomes the user actually wants.

**Priming.** What users see or experience just before a key decision affects how they make that decision.

*Use it:* The experience immediately preceding a key conversion moment is often as important as the conversion moment itself. Onboarding screens before the paywall, imagery before the permission request, copy before the commitment ask, all prime the user for or against conversion. Most products treat these screens as functional steps; they should be treated as priming opportunities.

*Backfires when:* Priming that misrepresents the experience is manipulation. Prime toward the genuine experience, not an idealized version that will create disappointment.

**Goal Gradient Effect.** Users accelerate toward a goal as they get closer to reaching it.

*Use it:* Progress indicators in onboarding increase completion rates. Trial paywalls that represent the trial as an in-progress journey (not an evaluation period) accelerate conversion. Streak mechanics drive return behavior because each streak day represents progress toward a milestone. Artificial headstarts, giving users a pre-populated state, pre-stamped progress, or "you're already X% of the way there", activate the effect without requiring prior behavior.

*Backfires when:* Progress bars that feel irrelevant to user goals don't produce the effect. The milestone must matter to the user. "Complete your profile" only activates goal gradient if users care about having a complete profile.

**Peak-End Rule.** Users don't average their experience, they judge it by how they felt at the peak moment and how it ended.

*Use it:* The end of the onboarding flow, the end of the first session, the end of the upgrade flow, and the end of the cancellation flow all disproportionately determine how users remember their experience and whether they return. Zapier showing confetti after upgrade is a deliberate peak-end investment. Adobe's ambiguous cancellation confirmation is a peak-end failure.

*Backfires when:* Peak moments can be negative too. A frustrating paywall, a confusing checkout, a harsh cancellation flow, these define how the product is remembered and discussed. Fix these before adding positive peaks.

### Mostly relevant at: paywall

**Social Proof.** When users are uncertain, they look to what others have done as the safest guide for what to do themselves.

*Use it:* Social proof operates at every stage of the funnel. At the top: aggregate numbers ("10 million users"), category logos, and review ratings. At conversion: "Most popular" plan labels, testimonials near the paywall, review quotes at the point of hesitation. In retention: activity feeds that show what peers are doing. In ads: user count, specific user outcomes, press logos.

*Backfires when:* Social proof that is clearly aspirational rather than typical backfires. "Results may not be typical" disclaimers undermine the proof. The most credible social proof is the most representative, not the most impressive.

**Scarcity.** People value things more when they're genuinely limited, by time, quantity, or access.

*Use it:* Limited-time trial pricing ("This offer expires Sunday"), early-access positioning, and limited cohort enrollments all use legitimate scarcity. The line to watch is whether the scarcity is real. Countdown timers that reset, "limited spots" that are always available, and expiring offers that reappear destroy credibility faster than they convert.

*Backfires when:* Fake scarcity is a short-term tactic with long-term costs. Users who discover it feel manipulated. Many markets now have consumer protection regulations around false urgency claims.

**Loss Aversion.** Losing something hurts roughly twice as much as gaining something equivalent feels good.

*Use it:* Trial-to-paid conversion messaging should be framed around what users will lose, not what they'll gain. Retention messaging for users approaching churn should reference their accumulated investment. Streak mechanics derive most of their motivational power from loss aversion, users don't want to lose a 14-day streak, even if they're not that attached to the daily behavior itself.

*Backfires when:* Overusing loss framing creates anxiety and association with negative emotions. Use it at conversion-critical moments, not as the primary tone of all product communication.

**Endowment Effect.** Users value something more when they feel it belongs to them, even before they've paid for it.

*Use it:* Trials that get users to invest in and personalize the product before the paywall create endowment. The user is now protecting something that feels theirs, not just deciding whether to buy a product. Trello's deferred account creation (create boards and cards, then confirm email to avoid losing them) is a direct application. Blinkist's onboarding gets users to build a personal library before presenting the paywall.

*Backfires when:* Endowment combined with a terrible free experience doesn't work. Users have to value what they're being asked to keep. If the trial didn't deliver real value, nothing to protect.

**Sunk Cost Effect.** Users who have already invested time, effort, or money in something are reluctant to abandon it, even when abandoning would be rational.

*Use it:* Products that get users to invest early create retention through sunk cost. This compounds with the endowment effect and investment loops. Duolingo's gem wagering mechanic uses sunk cost explicitly: you bet gems on your streak, and the fear of losing what you wagered keeps you engaged.

*Backfires when:* Sunk cost used manipulatively, trapping users who no longer get value, generates resentment and negative word-of-mouth. The investment must correlate with genuine product value, not just manufactured switching costs.

**Fitts's Law.** The larger and closer a target is, the faster and more accurately users can interact with it.

*Use it:* Amazon devotes a third of the purchase screen to the subscription option while the one-time purchase radio button is tiny. This is Fitts's law in deliberate commercial use. Conversely, cancellation links in settings menus are often made intentionally small and distant from natural user paths. Your primary conversion action should be the most physically accessible element on the screen.

*Backfires when:* Deliberate friction-via-size on exit or cancellation actions is a dark pattern with growing regulatory scrutiny in many markets. Design for guidance, not entrapment.

**Reactance.** When users feel their freedom is being restricted, they push back, often doing the opposite of what you want.

*Use it:* Any forced or pressured conversion attempt risks reactance. This is the mechanism behind "hidden exit" paywalls backfiring, users don't just leave, they leave angry and review-negative. The inverse is also true: giving users genuine control over how they experience the product builds goodwill that makes conversion more likely.

*Backfires when:* Reactance doesn't always announce itself. It shows up as churn, negative reviews, and social media complaints rather than as explicit refusal. Measure reactance by watching what happens to retention after conversion-forcing interventions.

**Second-Order Effect.** Every product decision creates downstream consequences that weren't the original intention, and these secondary effects can be larger than the primary one.

*Use it:* Any growth lever being pulled should be evaluated for its most likely second-order effect. Adding notifications to improve D7 retention, does it produce genuine return or just reactive opens that look like retention? Offering a deep discount to convert trials, does it attract price-sensitive users who churn faster?

*Backfires when:* Second-order effects are often invisible until they've been running for long enough to show up in cohort data. Build lagging metrics into your measurement from the start.

### Mostly relevant at: pricing

**Anchoring Bias.** The first price a user sees shapes everything they judge as expensive or cheap afterward.

*Use it:* Pricing page architecture is an anchoring decision. If your most expensive plan appears first or largest, everything else feels cheaper. Annual vs. monthly comparison ("$8/month billed annually vs. $15/month") uses the lower number as the anchor even though the annual commitment is larger. In ads, leading with the original price before the discounted price uses anchoring to increase perceived value.

*Backfires when:* Anchors that are implausible get discounted. If users sense the anchor is manufactured, the effect reverses and breeds distrust.

**Framing.** How information is presented matters as much as the information itself, sometimes more.

*Use it:* "Less than $1 a day" vs. "$27/month" is the same price framed differently, and the first consistently outperforms in B2C contexts. "Try free for 14 days" vs. "Start your free trial" frames the same offer differently, the first leads with duration and lowered commitment. Free trial expiration emails framed as "your access ends in 3 days" outperform "your trial is 10 days old" because one is a loss, the other is a fact.

*Backfires when:* Framing that misrepresents the actual offer erodes trust when users discover the reality. Frame the truth more favorably, don't create false impressions.

**Default Bias.** Users tend not to change from whatever option is presented as the default, even when changing would benefit them.

*Use it:* Defaulting to annual billing, defaulting to the recommended plan, and pre-selecting opt-in behaviors all leverage this bias. In onboarding, sensible defaults reduce cognitive load and completion time. In settings, defaults shape behavior because most users never change them.

*Backfires when:* Exploitative defaults (pre-checked opt-ins for paid upgrades, auto-renewal without clear disclosure) create legal exposure and destroy trust when discovered. Use defaults to guide good decisions, not to trap users.

**Von Restorff Effect.** When multiple similar items are presented, the one that looks different will be noticed, remembered, and chosen most often.

*Use it:* Pricing page "recommended" or "most popular" plan highlighting is the most common growth application. Making one plan visually distinct, a highlighted border, a contrasting background, a "Best value" tag, consistently directs the majority of choices toward that plan. In ad creative, a single distinctive visual element within a predictable category breaks pattern recognition and captures attention. In onboarding, a visually isolated action is more likely to be taken than one presented alongside other options.

*Backfires when:* Don't rely exclusively on color for distinction, users with color vision deficiencies may not perceive the difference. Combine color with size, shape, or text treatment.

**Centre-Stage Effect.** When choices are laid out spatially, users disproportionately pick whichever option sits in the middle.

*Use it:* Pricing tier layout is the clearest growth application: the plan placed in the middle of a three-tier row captures disproportionate selection share even when its value proposition is identical to a plan on either side. The same effect shows up in feature comparison tables, onboarding option pickers, and any UI presenting a small set of parallel choices side by side.

*Backfires when:* The effect is weak or absent when options differ sharply in price or value, users will still pick the outlier that actually fits their need. Centre-stage placement nudges close calls, it does not override a clear mismatch.

**Decoy Effect.** Adding a third, strategically inferior option makes your preferred option look significantly more attractive.

*Use it:* Three-tier pricing structures are nearly always decoy structures. The middle or premium tier is the target. The other tiers are designed to make it look attractive. On paywall screens, the "decoy" tier is usually the lowest option, priced just below the target in a way that makes the jump look worthwhile.

*Backfires when:* A decoy that's obviously manufactured destroys trust. The inferior option must feel like a real offering, not a planted straw man.

**Hyperbolic Discounting.** People consistently value a smaller reward now over a larger reward later, far more steeply than a rational discount rate would justify.

*Use it:* Immediate, smaller incentives outperform larger, delayed ones in acquisition and conversion offers. A discount applied at checkout today beats a larger discount promised next billing cycle. Trial structures that deliver value immediately (instant access, first result same session) convert better than structures that defer the payoff, even when the deferred version is objectively richer. Annual plans are harder to sell on total savings alone because the saving is realized in the future while the cost is paid now.

*Backfires when:* Leaning too hard on immediate incentives trains users to wait for a deal rather than convert at full price. Use urgency and immediacy for a specific offer window, not as the default state of the paywall.

---

## Pricing and packaging, field notes

Practitioner case claims from operators who ran the tests. Treat each as a strong prior from someone who did it, not as platform-wide data.

### Pricing & Packaging
1. Test-priority order (Parizek, Mojo): price by market -> highest-LTV plan duration -> plan ratio -> trial/intro offer -> paywall design LAST. Pricing/packaging moves more per unit of effort than design.
2. The variant that made the most money was wrong to ship (Mojo): +100% price beat +50% on new revenue but 7-day cancellations spiked; shipped +50%. Judge price tests on 7-day cancels, not month-one revenue. Proxy validated vs real renewals (+/-).
3. Monthly-to-annual ratio fix took annual mix 5% -> 50% (Layfield, Codecademy). Price annual at avg monthly retention + 1-2 months. Monthly LTV ~$85-90, annual ~$300.
4. "Not a yearly product" app: ratio fix + an onboarding screen setting expected usage duration -> annual 16% -> 39% AND paid conversion 6.6% -> 13.7% (Tideman).
5. Four cases where CUTTING price is right (Gamsriegler): wrong mental category, never localized, mission fit, category parity gap. Only with a bad trial/purchase CR too.
6. Gate the speed, not the capability (Muir, MyFitnessPal case): the free feature that only saves time is the one to paywall; clawing back later is far harder.
7. Payback window decides which plan you push (Lais): bootstrapped ~7 days -> annual hard; funded 60-90 days; biggest apps 6-12 months. Watch cohorts 45-60 days.
8. Sell a new high-touch offer as one-time purchase first, first unit free; repeat rate decides tier vs bundle (Kwakyi, Lingvano).
9. App-to-web is not a grand slam: iOS-US-only entitlement, ejection friction eats the take-rate gain, especially vs 15% small-business cut (Kwakyi).
10. Big offers in bursts, never permanent: a permanent rich offer is a one-way door of expectations (Scarpaci).

### The Paywall
11. Hide monthly behind a second screen: annual 60% -> 80%, overall conversion HELD (Parizek, Mojo). Second win: yearly's monthly-equivalent price next to real monthly price. "Layout tests only get big when they change which plan people see first."
12. Broad targeting needs a price ladder, not one point: high-price no-trial paywall -> three-plan trial paywall -> 50-70% off end-of-onboarding/home-screen (Burke).
13. THE TRIAL-TIMELINE PAYWALL ONLY WORKS FOR HABIT APPS (Muir): every client copied Blinkist's timeline; it sells a routine/relationship. Utilities: state job + price.
14. Three surfaces every user passes: checkout, payment retries/win-back, cancel flow. 1% on checkout = 1% on revenue, no segment dilution (Layfield).

### Onboarding -> The Ask
15. Mimo: paywall promised "professional developer," ad/onboarding promised a kids' game; users doubted a year of content. Fix the promise chain, not the paywall (Gamsriegler).
16. Prove value before the paywall, not during the trial: ~75% of traffic gone before day-1 retention. Scripted wins: PictureThis template scan, recipe import demo, quiz-answers "read on you" (Burke).
17. AI/conversational onboarding: 3x longer, MORE completions, day-14 ARPU >10x (Gen Z audience). Win = interactivity, not AI; judge on day-14 ARPU (Tideman).
18. One promise repeated at every touchpoint: ad line = store page = onboarding = paywall headline; swapping "Unlock Premium" for the winning ad's promise moved trials AND subs. Org chart (UA vs product) is why it's rare (de Baillenx).
19. Onboarding answers can decide who sees the expensive tier (Muir, Duolingo Max read; speculative on their internals, tactic stands).
20. AI apps: sell the output without generating it (pre-made examples in onboarding, confirm-the-brief step before first real generation) (Muir).
21. Gate referral rewards behind the action that predicts retention, not signup (Scarpaci; Blue Apron vs HelloFresh).

### Trials, Offers & Second Chances
22. Mojo: 3-day vs 7-day trial, no trial-to-paid difference -> ship 3 for faster payback (Parizek).
23. Trial on the yearly plan ONLY, as a third annual push lever; trial appetite is regional (LATAM values it more); removing trial for new users LOST at Mojo despite community case studies. "A tactic can be real and still wrong for you."
24. Trial starts are a vanity metric: longer trial -> more starts, fewer payers (procrastination). Judge on paid conversions per install (Tideman).
25. >10% purchase rates with no trial at all in high-WTP categories (AI, dating, fitness); day-one purchase event is the strongest ad-algo signal (Burke).
27. Exit-intent ~50% discounts on paywall dismiss now carry a BIG share of total subscriptions; paid trials replacing free also rising (Lais). Trial-start signal weakening as conversions route around trials.

### Reading the Test
34. Judge pricing/plan-mix tests on revenue per allocated user, the only deciding row ($60 x 5.0% = $3.00 beats $40 x 6.5% = $2.60); run every 6 months (Layfield).
35. Split trial opt-in into two questions first: % seeing the paywall vs % converting on it; fix whichever half is weak (Gamsriegler).
38. Pricing-page/paywall views are the strongest predictive variable for user value models; instrument them like revenue (Galiza).
42. Ads as an upstream lab for paywall copy: 15 promise variants as hooks, read in ~7 days on cold traffic, top 2-3 graduate to the paywall test; specific wording beats vague consistently (de Baillenx).
---

## Paywall redesign case studies

Every one of these is a bundled redesign. Never attribute a bundle's lift to a single element.

### RevenueCat four redesigns (agency A/B cases)
1. Crypto fintech promo paywall: price-chart discount story + "Back to $15.99/month" marker + trader language -> view-to-sub 2.7% -> 3.24% (+20%).
2. Driver-license prep onboarding paywall: trial TOGGLE + short form + weekly price anchor ($14.99/wk) vs monthly at $9.99/wk (40% savings) + review-first -> +17% ARPU.
3. Party game onboarding paywall: toggle + short form + monthly REMOVED to "View All Plans" + "SAVE 83%" + $12.33/wk yearly vs $29.99/wk weekly -> +31% install-to-trial, +64% revenue. (BUNDLED: per-period framing cannot claim the lift alone.)
4. Food app onboarding paywall: toggle + short form + reviews + impact fact -> +72% install-to-trial. NOTE: Apple began rejecting trial toggles Feb 2026 (article's own FYI); toggle wins here are Android-only advice now.

### Superwall five patterns (pattern census of $1M+/mo apps, with revenue figures)
1. Anchor & decoy: MacroFactor ($2.3M/mo), Calm ($4M), SCRL ($2M, "SAVE 85%").
2. Value stack: MyFitnessPal ($13M), ChatOn AI ($5M), ClassDojo ($1.3M).
3. Social proof engine: Flo ($9M, testimonial on paywall), Yazio ($3.3M), Speak ($2.8M, 5M users + 4.8 stars/140k reviews).
4. Soft commitment (risk reversal): Strava ($11M, day 1/28/30), Cal AI ($2M, "No Payment Due Now" + reminder), Lose It! ($3.3M, whole paywall = how the trial works).
5. Now-or-never offer: Captions, Finch, Yazio - big-contrast limited discounts (Flo's 44%-off gift sequence is this family).

### RoastMyApp "51 Paywall Experiments"
Nature: an experiment IDEA CATALOG with named-app example screenshots, NO outcome data attached - a breadth inventory, not evidence. Use as an experiment-idea catalog for a CRO backlog:
- #4 "see all plans" button, #17 highest-LTV plan as the no-brainer deal, #18 trial on only 1 plan (Tolan/Mojo pattern), #16 yearly vs higher-priced yearly (decoy at duration level).
- #20 "7-day free reverse trial that starts automatically without IAP" - pictured example: Ladder. Direct validation of the decline-path pattern as a known play.
- #39 "end-your-trial-now discount for direct purchase" - names Flo's convert-early tactic as a recognized experiment (still no public number).
- #23 weekly-yearly anchor combo, #24 design-your-trial (Ahead), #34-38 exit-offer cascades (doc itself flags Apple-review risk), #42 offer during app uninstall.
- Useful reframes: #30 "7-day transformation timeline to set realistic expectations" (the time-expectation idea in the wild), #15 post-purchase buyer's-remorse flow.
---

## What the teardowns found

Counts and sequences observed across the corpus, by vertical. They tell you what apps do and how commonly, which is a different claim from what works. Use them to know where you sit relative to the field, and to spot the places where the whole field is doing something unexamined.

### Health and fitness, 23 flows

**Early account creation as value exchange, matched to arrival intent.** 70% of apps (16 of 23) place signup before or bundled with the first personalization questions. The one late-signup cluster is Weight & Nutrition Coaching, where the plan does the convincing first. Account creation appears framed as "Create your account to build your health profile". Health users arrive with high intent to invest in themselves, so the early ask is earned here in a way it is not in wellness or education. This is what scopes the cross-vertical "delay the signup ask" pattern: delay it where arrival intent is low, not everywhere.

**An honest counter or named phases, not a vague progress bar.** Fitbod puts a real numbered counter on every screen ("3 of 9"). Named phases work the same way. Users carry health-form anxiety into this vertical's onboarding, so knowing exactly how much is left matters more here than anywhere. Precision lowers perceived effort more than any vague bar.

**Read the user's own answers back to them, verbatim.** Peloton's recommendation screen names the exact picks: "because you like Meditation," "45-60 mins." Personalization has three layers, say it is personalized, say how, say why. The more specific the callback to what the user actually typed, the stronger the trust and the stickiness.

**The reason for the question sits in its own header or subheader, and serves the user.** Runna ties its workout-frequency question directly to injury risk. Needed wherever it is unclear what answering gets the user. The test is whose interest the stated reason serves: "so we can personalise your plan" is the app's need dressed up; "so we don't give you a plan that injures you" is the user's.

**Reassurance specific enough that the user could check it.** Strava's "will not appear on your profile" and Gentler Streak's "will never leave this device" score the highest trust ratings in the dataset. "We take privacy seriously" is noise. A concrete, checkable claim placed next to the field it protects is what lands, and in this vertical the fields are the most sensitive data a consumer app asks for.

**Age/DOB is the vertical's nearest universal.** 83% of apps (19 of 23) ask for it, more than any other question type. Prevalence, not endorsement: this tells you where you sit relative to the field, not that you should ask. Apply the triangulation test in the critical checks before adding it.

**Medical care & access routes by treatment or service selection, not by stated motivation.** None of GoodRx, Ada, Hims, Hers or Zocdoc ask for a goal or motivation at all; they ask what you need. A different personalization mechanism from the rest of the vertical. These are transactional jobs, the user knows what they came for, and a motivation question delays it.

**Segment signature: height and weight cluster in Weight & Nutrition Coaching.** Height appears in 4 of 6 apps in that segment against 3 of 17 everywhere else. The questions exist because the calorie and metabolic math needs them, not because they personalise. A useful test for your own flow: which of your questions is load-bearing for a calculation, and which is habit.

**Anti-pattern: a sensitive permission pre-toggled to on.** Flo arrives with Apple Health write access already toggled on, over some of the most sensitive data in the vertical. The missing active yes is itself what reads as a loss of control. The user notices the choice they never got to make, and the trust cost lands on the whole flow.

**Anti-pattern: questions hidden inside a fake "Analyzing..." screen.** Noom. It works once. After the first user notices the progress is theatre, the cost is paid on every subsequent trust ask in the flow, including the paywall.

**"Is this for you or someone else?" is a Cycle & Reproductive Health signature, not a general pattern.** Clue and Flo both ask it, both before birthdate; no app in any other segment asks it at all. The trust asks a flow can make grow out of what the product actually does. Cycle tracking has a real proxy-use case; most verticals do not, and borrowing the question without the use case is friction.

**A multi-screen consent stack is warranted only by a genuine triage job.** Ada is the only app in the set with three separate consent screens (age-gate/ToS, data-privacy, research opt-in), and the only one doing actual symptom triage. GoodRx, Hims and Hers fold consent into account creation as boilerplate. Ada's stack is not over-caution, it is the flow matching its liability. Copying it without the triage job buys three screens of friction and no trust.

**Attach a real number to the notification ask, framed as progress toward the user's goal.** Yazio cites a 78% opt-in rate. Runna promises 2x. Aaptiv, 3x. The device recurs across two of the five segments. An unframed ask reads as marketing and gets a reflex no. A specific statistic makes the ask feel safer, not just more persuasive, because it shows other people said yes and got something back.

**Earn a system permission before requesting it.** Aaptiv requests tracking permission on open, before a single value screen has loaded. Where a permission sits in the sequence matters as much as how it is framed. An ask before any value has been shown draws a reflex denial, and with system permissions you rarely get a second attempt.

### Education, 20 flows

**Transactional apps skip onboarding entirely.** Photomath, Gauth, Course Hero and Quizlet ask zero questions. When the job is to get an answer fast, every onboarding question lowers completion.

**Fifteen questions is the observed ceiling, and habit is the question everyone long adds.** No app in the set exceeds 15 onboarding questions. Every app asking four or more includes at least one habit-formation question. Prevalence, not endorsement, but a useful bound. If your flow is past 15 you are outside the entire observed field, and if it is long without a habit question you are long for reasons the field does not share.

**Four questions define the floor across every segment.** Content and path (70%), goal and motivation (65%), daily time commitment (45%), attribution (40%). These are what the field converges on. Treat them as the set to justify against, not the set to copy: each still has to change something the learner sees.

**Prior knowledge is a sequencing signal, not a universal best practice.** Asked by 80% of language apps and 100% of coding apps, and by no segment without curriculum sequencing. The question earns its place only where a curriculum actually branches on the answer; elsewhere it is a quiz the product never reads.

**Microlearning: psychographic questions used to convert, not to personalise.** Self-perception prompts such as "do you relate to this?" placed ahead of the paywall. No other segment does this. The answers barely change the content. What they change is how invested the user feels by the time the price appears. Worth naming honestly in an audit: it is a conversion mechanic wearing personalization's clothes.

**Attribution is asked where paid acquisition is heaviest, and asked in the wrong place.** Language apps ask it at 60%, consistent with survey-based attribution replacing lost platform signal. Every app that asks it places it inside onboarding. The rare row where the whole field looks wrong: completion data favours a post-signup or post-paywall placement, and the question personalises nothing, so it is pure cost paid at the most fragile point in the flow.

**Session timing is underused relative to its retention link.** Only 20% of apps ask when the learner prefers to study. The gap worth testing into. Fixed-time cues are one of the better-evidenced retention devices in learning, and four in five apps in the set never ask.

**Three high-signal question types are absent across the entire set.** Not one of the 20 apps asks about past failure, social context, or context of use. The clearest unexamined space in the vertical. Every app asks what the learner wants and none asks why the last attempt failed, who else is involved, or where they will actually be using it, the three things that most predict whether this attempt goes the same way.

**Behavioural baseline questions are nearly absent; the field runs on stated preference.** Only Imprint asks about actual prior behaviour. Stated preference is the weakest input a personalization engine can run on, and it is what almost the whole vertical uses.

**No app in the set uses onboarding as a differentiator.** Within each segment the question set and order are near-identical across competitors. Read this as the opportunity row. The entire field converged on the same flow without testing it, which means onboarding is uncontested ground in a vertical where acquisition costs are contested hard.

### Sports, 18 flows

**Fan apps: team or entity selection is the dominant personalization signal.** Present in 100% of fan engagement and 100% of news & scores apps. The user picks who they follow and every downstream surface personalises off it. One question does the work a whole quiz does elsewhere. If you are fan-facing and running a multi-question quiz, this is the row to test against it.

**Fan apps: three core questions is the convergence point.** Median of 3 core questions for both fan-facing segments against 6 for team management. No fan app exceeds 5. Fan intent is immediate and shallow, the user came to check a score or follow a team. Every question past the third is spending goodwill the app has not yet earned.

**Demographics collection tracks regulatory exposure, not personalization intent.** 0% of news & scores apps collect demographics. The team management and fan engagement apps that do are driven by registration and age-gate obligations. Before adding a demographic question, check whether the apps you are copying ask it because it personalises something or because a regulator made them. Mostly it is the latter.

**Team management: collects roughly twice the data and personalises less with it.** 6 core questions on average against 3 for fan apps, and almost none of that data changes what the user is shown. The clearest self-audit in this dataset: list every question, then name the screen each one changes. Questions that change nothing are operational, and belong after activation rather than before it.

**Fan apps: finish personalization before requesting registration.** 7 of 11 fan-facing apps complete full personalization before asking for an account, and one requires no account at all. A fan has already told you the team they follow, which is enough to show them something worth having; the account ask lands better once they have seen it.

**Identity collection is exclusive to team management, and warranted there.** 100% of team management apps collect identity during onboarding against 0% of fan engagement apps. The product genuinely cannot function without knowing who is on which roster. The general rule this illustrates: a trust ask is earned by what the product does, so if you cannot name the feature that breaks without the answer, you have not earned the question.

**Fan apps: notification setup goes last, after content selection is complete.** 50% of apps configure notifications during onboarding, and every one of them places it after the user has chosen what to follow. Once the user has picked a team, the notification ask stops being a request for attention and becomes the delivery mechanism for the thing they just asked for.

---

## Named examples, principle by principle

The apps we would show someone for each principle, and the ones that get it wrong in an instructive way. Store listings and flows change, so recheck before quoting one.

### 1. Store screenshots are billboards
Done-well bar: each screenshot lands one message in half a second; captions sell the value, not the feature; the set survives with no text read at all.
- Best example: Tolan. First frame = one benefit line in giant type + one character, zero UI chrome ("Tolan is your friendly guide through life"); every frame keeps the formula.
- Instructive failure, incomprehensible: Whoop. "Know yourself at the molecular level" over a dense biomarker panel - legible words, no landable message.
- Instructive failure, feature-verbs: TrainingPeaks. "Plan training. Sync workouts. Find coaches." - perfectly legible in half a second and says nothing about the outcome.
- "Mobile banking made easy" over a visibly dense dashboard - the caption promises what the creative disproves.
- Captured Sep 2026; listings rotate, recheck before reuse.

### 2. Every question shows its value
Done-well bar: the reason sits on the same screen as the ask, written as the user's benefit; on significant asks (birthday, gender, contacts) it also carries a checkable privacy line.
- Best example: Strava, birthday screen. Three concrete reasons (performance analysis, leaderboard filtering, youth safety) plus "your birthday or age will not appear on your profile" - a falsifiable privacy claim, not vague reassurance.
- Runna, workout-frequency question. The reason is protective ("at most once more than you currently run, to reduce injury risk"): the app limits the user for the user.
- RISE, energy-levels question. Mechanism reason ("we use factors beyond sleep to predict daily energy") - specific, though framed as what the app consumes rather than what the user gains.
- Lifesum. The same reason line repeated on every field; scored the strongest reason-giving rating in the wellness set. Observations: bare questions are the NORM across the corpus - the reason subheader is the exception, so this is cheap differentiation. Strava's pattern (reason + privacy line) is the ceiling on sensitive asks.

### 3. Prime the notification ask
Done-well bar: primer before the OS dialog showing the actual notification they'll receive, a real number on the benefit, framed in their goal; ceiling adds per-type toggles and a graceful no that preserves the one-shot OS prompt.
- Best example: Runna. Real notification preview with real content ("Easy Run 4km: 30-40m"), "2x more likely to finish your plan" - number + preview + goal framing in one screen.
- Flo. The primer IS a value moment: shows their predicted period date and the actual reminder they'd get, then tells them "tap Allow on the next screen" - best OS-dialog setup in the set.
- Zero. "26% more likely to build a fasting habit" + Maybe Later - the graceful no.
- Headway. Per-type toggles with a reason per type (mailing-list pattern): opting in feels like choosing, not surrendering. Rare in the corpus. Observations: Yazio personalizes the primer with the user's goal date; Insight Timer literally previews the OS dialog with an arrow.

### 4. Make it feel built for them
Done-well bar: a result computed from their actual answers, cited back in specifics, with a comparison anchor, a mechanism link, and a confirm/adjust step so the user co-signs the output.
- Best example: RISE, sleep-need screen. "Your sleep need is 10h 30m" computed from their patterns, the "Most People" comparison band, "How do you calculate it?", "Adjust sleep need", and a THAT LOOKS RIGHT confirm. The only example in the corpus with all five elements.
- Elevate, starting EPQ. Real per-skill scores from an actual assessment - the strongest edtech version of a computed result.
- MacroFactor-style calorie plan. Name callback + "2,930 daily calories - 1.2 kg by April 8" + "How we make recommendations."
- Yazio, goal projection. "70.0 kg by April 7" with a chart; weaker because the benefits list under it is generic claims, not callbacks. Observations: the confirm/adjust step is the least-shipped element and the highest believability lever: verification converts personalization from claim to belief. Peloton's "because you like Meditation" tags are the canonical recommendation-side callback. Noom's "analyzing" screen that hides extra questions is the anti-pattern: it works once, then every loading state reads as suspect.

### 5. Put the ask where the leverage is (account creation)
Done-well bar: signup lands after a real value moment as a trade ("save your plan"), or deliberately first as a commitment device; never a cold wall before any value. Ceiling adds SHRINKING the ask: email-only or one-tap SSO first, profile later (Duolingo asks name then email progressively; RISE asks email alone).
- Best example (pattern): Duolingo. First lesson before any wall, soft dismissible wall, hard wall later; the A/B behind it is reported as roughly +20% next-day retention (secondary sources; some render it as DAU).
- Balance / Happier. Hold registration until "your plan is ready."
- RISE. Waits until it can show a real Sleep Debt figure before asking for an email. Observations: deferred registration is the most common wellness placement (9/18) but only a plurality - front-loading persists.

#### Where seven strong apps put the account ask

The seven captures map the full placement spectrum, not one pattern:
- Duolingo (best example, deferred): lesson first, then "Don't lose your progress! Let's create a profile" with LATER; signup itself is progressive (name, then email). The reported ~20% next-day-retention A/B owner.
- RISE (post-value save): email asked only after Sleep Debt is on screen, "to save your progress," privacy link inline.
- Lifesum (pre-plan gate): quiz complete -> "Let's get you your personalized health plan" -> signup -> "setting everything up." Signup framed as the key to the reveal.
- Meditopia (pre-trial gate, 93% progress): "Complete your profile to save your preferences and access your program," then trial explainer.
- Brilliant (pre-paywall): "Sign up for free to discover your learning path," then the premium screen.
- Fabulous (mid-flow with purpose): email required mid-onboarding with an explicit no-motivational-emails opt-out, adjacent to the signed commitment contract.
- Flo (post-premium): account creation AFTER premium is granted, wrapped in social proof (#1 women's health app, 90% prediction accuracy claim). Observation: placement is a spectrum, and the dead pattern is the cold wall with nothing traded. Pre-paywall placement also arms the decline path: an email captured before the paywall is the winback channel for decliners.

### 6. Show the trial timeline (paywall anxiety)
Done-well bar: trial start, reminder day, and charge day visible with REAL DATES, the reminder promised (and sent), cancellation named as easy with the steps shown.
- Best example: Blinkist. The originator (+23% trial starts, complaints -55%): Today/Day 5/ Day 7 with the charge date, per-month price framing, and "How can I cancel" answered on the same screen.
- Duolingo (the ceiling): lets the user PICK which day the reminder arrives (2 or 3 days before, with dates). Added friction that signals transparency; Duolingo reported it boosted conversions (RevenueCat 2026, Anmol Tiwari).
- Headspace: timeline + $69.99 ($5.83/month) per-period framing on the same screen - doubles as a price-in-small-units example.
- Strava: 30-day version with "you won't be charged until Feb 6" date specificity.
- Also filed: Insight Timer (cancel-in-15-seconds copy), Brilliant, Tolan (dated reminder promise + "easy to cancel, no penalties or fees"). Also filed: Lingokids (guide + dated reminder), Duolingo Super trial paywall ("We'll notify you 2 days before" + "no penalties or fees" + No Thanks). Tolan is the floor done well - one plain dated heads-up. Blinkist adds the full map. Duolingo is the ceiling: choosing WHICH day the reminder arrives is simultaneously a control device and a micro-commitment. The pattern sells a ROUTINE - works for habit products, not utilities (Alice Muir). Copy the promise only if you're selling the same thing.

### 7. The default does the selling
DONE = the recommended plan preselected with visual weight and a badge, the alternative priced to make it obvious. DONE WELL = lead with the ONE plan you want sold and put the rest behind a "See all plans" tap (the Mojo pattern: annual share 60->80% with conversion held). Buyers overwhelmingly take whatever you lead with (82% per Superwall; mix inverts 37->63% on preselection alone).
- Annual-forward + "See all plans" in the store already: Fitbod, Blinkist, Imprint paywalls all lead with one plan and tuck the rest behind a link.
- Best example: Ladder. Annual preselected with checkmark + "Best value + all team access" badge, both plans shown per-month ($20.83 vs $43.98) so the monthly anchors the annual, CTA reinforces ("Continue - save 50%").
- Tolan. Yearly default with "Most popular, save 73%", $1.35/wk vs $4.99/wk, and the free trial attached to the YEARLY plan only - trial availability as a third annual push (the Mojo pattern from the Price Power playbook).
- Duolingo Super family upsell: family plan preselected against individual - the default applied to tier upsell, not just duration.
- WARNING: Apple began rejecting toggle paywalls under Guideline 3.1.2 in early 2026 (RevenueCat dates it Feb 2026; no announcement). Observations: Mojo moved monthly one screen deeper - annual went 60->80% with conversion flat. "Layout tests only get big when they change which plan people see first." The trial-toggle is the emerging control-pattern to watch.

### 8. Price in small units
Done-well ladder: DONE = the annual plan carries a per-month/per-week equivalent. DONE WELL = the equivalent sits NEXT TO the alternative's real price so the gap is visible without arithmetic (the Mojo pattern).
- Picked on scan-clarity and single-variable purity.
- BoldVoice (web-checkout paywall): the fuller-transparency variant - annual card carries BOTH units ($149.99/year AND $12.49/month), reframe repeated at the CTA, "No trial period" labeled on monthly (trial as a visible annual perk - a second lever, so slightly less pure as a framing teach). Near-tied gold; defensible in either order.
- Fitbod. "$79.99/yr billed yearly" + the explicit reframe "That's only $6.67 per month!" on the CTA - the pattern said out loud; best verbal version.
- Imprint: "7-day free trial, then $99.99 per year ($8.33/month)."
- Headspace: "$69.99 ($5.83/month)" on the trial-timeline screen.
- Observations: the oft-cited $12.33/wk-vs- $29.99/wk breakdown comes from a BUNDLED RevenueCat redesign (party game app: +31% install-to-trial, +64% revenue - framing cannot claim the lift alone; the +20% case was a separate crypto promo paywall). The clean directional evidence is Mojo: yearly's monthly-equivalent shown against the real monthly price raised annual share. Cite the bundle honestly or cite Mojo.

### 9. Design the decline path
Done-well bar: a closed paywall triggers a next step instead of a dead end - on iOS in 2026 that means a reverse trial (auto-grant temporary premium, no purchase UI) or a DELAYED offer (end of onboarding / home screen / lifecycle), not an instant exit-discount popup. Platform reality: Apple has been rejecting instant paywall-abandonment offers (citing Guideline 5.6, inconsistently, no written ban) and killed toggle paywalls under 3.1.2 in early 2026 (RevenueCat dates it Feb 2026). Instant exit discounts remain viable on Android and web.
- Reverse trial (the compliant iOS play): dismissers get premium free for days, no card, no offer screen; loss aversion converts. 0.4% -> 4.5% freemium conversion for ONE app, a practitioner case (Steve P. Young, App Masters) quoted inside RevenueCat's 2026 report, not platform data and no control described; Ladder and Strava run versions. The compliance read is our inference (no purchase flow shown), not an Apple ruling.
- Best example (grant-first family): MasterClass. Rare pattern in the wild; that rarity is part of the point. Renewal terms bill the upgraded tier's full price in year two - loss aversion built into the mechanics.
- Convert-early trial offer: Flo. Three-beat sequence - premium granted, "a gift just for you" box reveal (variable reward), then "turn your free trial into 44% OFF FOREVER" one-time offer. A discount for committing DURING the trial, not on decline. No public A/B; adjacent evidence: paid intro offers convert higher-quality subscribers (Bagdasaryan, RevenueCat 2026).
- Sunlitt: the classic one-time 50% offer screen ("close this page and it's gone forever") - clean visual for the pattern Apple now rejects when fired instantly on paywall dismiss.
- Ladder: "Welcome gift - you've unlocked the Ladder Widget" - giving VALUE instead of discount as the post-paywall beat.
- Observations: exit-intent discounts carried a big share of subscriptions (Day30) - which is exactly why Apple moved.

### 10. The commitment pact
Done-well bar: an explicit commitment moment sealed with a physical act (press-and- hold), written in the user's own goal terms, placed after buy-in and before the ask.
- Best example: Fabulous, "John's Contract" - the contract is PERSONALIZED (their name, their goal), sealed with press-and-hold, and cites precommitment research on the screen.
- Headway: "Commitment pact" card + press-and-hold fingerprint; generic text. Evidence (no public app A/B yet): stickK platform data - 78% goal achievement with stakes + accountability vs 35% without; implementation-intentions meta-analysis (Gollwitzer & Sheeran, 94 studies, d=.65) - forming an explicit commitment roughly doubles follow-through; commitment-contract RCTs in smoking cessation show durable abstinence gains. Prevalence: rare and new - a differentiator.

---

---

## Benchmarks

Two sources carry almost every number here. RevenueCat's State of Subscription Apps 2026 covers
about 115,000 apps and reports medians and percentiles, so its figures are what a typical app
sees. Adapty's State of In-App Subscriptions 2026 covers about 16,000 apps, mostly iOS, and
reports averages, so its figures usually run higher on the same metric. A 6% and an 11% for the same
step is median against mean plus sample, not a contradiction. Where the two genuinely disagree,
the last part of this section says so.

Check the referent before repeating any of these. Several are shares of a subset rather than
rates, and they are misquoted constantly. And place the app before you quote a number: the
cross-vertical median is freemium-weighted and will understate a hard-paywall, iOS, AI or
high-priced app by several times.

### Place the app first: the five axes that move the numbers

**Access model is the single biggest swing.** Realised install-to-paid at day 35 (RevenueCat 2026):

| Access model | Bottom 10% | Median | Top 25% | Top 10% |
|---|---|---|---|---|
| Hard paywall (nothing usable without paying) | 4.2% | 10.7% | 20.0% | 38.7% |
| Freemium (a free path exists) | 0.3% | 2.1% | 4.5% | 8.2% |

Hard converts about 5x freemium at the median, and the hard-paywall floor is twice the freemium
median. A gated free tier plus an onboarding paywall is still freemium: benchmark it at the
upper freemium range, not the hard-paywall median. Getting this wrong inflates the expected
rate by roughly 8x. Year-one retention is the same under both models (about 27% hard, 28%
freemium on annual), so access model is a conversion decision, not a retention decision.

**Category is the store category.** Both reports cut by store category, so a wellness, sleep or
meditation app reads the Health and Fitness rows, a journaling or habit app usually Lifestyle or
Productivity, a kids app Education, and a fintech app Finance or Business. Where a category has
no row for a metric, use the cross-vertical median and say so.

**Price tier is relative to the category's range.** The reports do not publish the dollar cut
points behind "high, mid and low-priced". Place the app against the interquartile ranges below
(Adapty 2025 data): inside the range is mid-priced, above it high, below it low.

| Plan | North America | Europe | Latin America | Middle East and Africa | Asia-Pacific |
|---|---|---|---|---|---|
| Weekly | $6 to $9 | $6 to $10 | $5 to $8 | $5 to $8 | $5 to $8 |
| Monthly | $7 to $10 | $8 to $18 | $4 to $7 | $4 to $7 | $4 to $7 |
| Annual | $28 to $50 | $28 to $55 | $22 to $45 | $22 to $42 | $25 to $48 |

Annual by category in North America: Productivity $40 to $65, Health and Fitness $30 to $45,
Education $30 to $45. An annual plan under $28 in North America or Europe is almost certainly
underpriced; median subscription prices rose 7% to 10% between 2023 and 2025.

**Paywall placement and trial interact.** Install-to-paid (Adapty 2026, averages):

| Placement | With trial | Without trial |
|---|---|---|
| In onboarding | 1.35% | 0.82% |
| In-app, later | 0.89% | 0.76% |

The placement advantage is almost entirely a trial interaction: onboarding plus trial is about
2x in-app plus trial, and without a trial the two placements nearly tie. In-app with no trial
is the weakest configuration in the data and the one teams default to when they do not want to
be pushy.

**Platform.** iOS converts about 3x Android overall and 6x on weekly plans. Install-to-paid by
plan (Adapty 2026): weekly 1.55% iOS against 0.26% Android, monthly 0.49% against 0.17%, annual
0.53% against 0.14%. RevenueCat's day-35 figure is 2.6% App Store against 0.9% Google Play,
while trial-to-paid is at parity (32.6% against 32.5%): the store premium is all at the top of
the funnel. Blended numbers hide it.

**AI or not.** AI apps start trials at about half the average rate on Adapty's data (5.3%
against 10.9%; Adapty's funnel section gives the same all-app step as 11.2%) and convert trials a little worse (20.6% against 25.6%), then earn 41% more per
payer in year one ($30.16 against $21.37, RevenueCat) with weaker retention (annual 21.1%
against 30.7%) and higher refunds (4.2% against 3.5%). RevenueCat's own trial-start rate runs
the other way (8.5% AI against 5.6% non-AI), which is a sample difference; either way the AI
lever is install-to-trial, not trial-to-paid.

**Price tier.** Higher-priced apps convert better at every step, not worse. Install-to-trial
median 8.9% high-priced against 5.4% mid and 4.4% low; install-to-paid 2.8% against 2.0% and
1.4% (RevenueCat 2026). On weekly plans the pattern is monotonic across price buckets, 5.4% to
9.9% install-to-trial (Adapty). High-priced apps also lose more at the first renewal (37%
against 24% on annual), and survivors then behave identically by the third renewal.

### Install to trial

The share of installs that start a trial within 30 days (RevenueCat 2026, medians):

| Cut | Median | Top 25% | Top 10% |
|---|---|---|---|
| All categories | 6.2% | not published | 20.3% |
| Business | 9.1% | 16.2% | not published |
| Health and Fitness | 6.9% | not published | 23%+ |
| Education | 6.5% | not published | not published |
| Utilities | 6.5% | not published | not published |
| Gaming | 4.4% | not published | not published |
| Travel | 4.1% | not published | not published |
| Media and Entertainment | 4.0% | not published | not published |
| North America | 7.1% | 15.0% | about 25% |
| Asia-Pacific | 5.7% | 13.5% | not published |
| Western Europe | 5.0% | not published | not published |
| India, South-East Asia, Latin America, Middle East and Africa | 3.0% to 3.7% | not published | not published |
| High-priced | 8.9% | 16.5% | 27.0% |
| Mid-priced | 5.4% | 11.6% | not published |
| Low-priced | 4.4% | 10.3% | not published |

Adapty's averages for the same step: 11.2% globally, 14.5% in North America, 7.6% to 10.2%
elsewhere. By plan and price bucket (Adapty 2026, install-to-trial):

| Price bucket | Weekly | Monthly | Annual |
|---|---|---|---|
| Low | 5.4% | 0.6% | 2.6% |
| Lower mid | 6.8% | 0.6% | 2.2% |
| Upper mid | 9.8% | 0.3% | 1.8% |
| High | 9.9% | 0.5% | 3.4% |

Monthly is the structural dead zone: 0.3% to 0.6% install-to-trial, and a shrinking share of revenue. Weekly's advantage over monthly is 15x to 20x at the upper price buckets, which is the
mechanical reason weekly plus trial produces the highest twelve-month LTV. Annual converts to
trial 2x to 3x better than monthly despite the larger ask, because it attracts higher intent.

### Trial to paid

| Cut | Median | Top 25% | Source |
|---|---|---|---|
| All categories | about 32% | not published | RevenueCat 2026 |
| Travel | 43.5% | 62.4%+ | RevenueCat 2026 |
| Health and Fitness | 37.7% | 51.4%+ | RevenueCat 2026 |
| Shopping | about 35% | not published | RevenueCat 2026 |
| Education | about 33% | not published | RevenueCat 2026 |
| Gaming | 25.0% | 39.8%+ | RevenueCat 2026 |
| Photo and Video | 22.2% | 33.1%+ | RevenueCat 2026 |
| North America | 34.2% | not published | RevenueCat 2026 |
| Asia-Pacific | 31.9% | not published | RevenueCat 2026 |
| Western Europe | 29.7% | not published | RevenueCat 2026 |
| Latin America, Middle East and Africa | 20% to 23% | not published | RevenueCat 2026 |
| India and South-East Asia | 15.2% | 25% | RevenueCat 2026 |
| Weekly plans | 27.8% | not published | Adapty 2026, average |
| Monthly plans | 45.0% | not published | Adapty 2026, average |
| Annual plans | 32.1% | not published | Adapty 2026, average |
| AI apps | 20.6% | not published | Adapty 2026, average; 25.6% for all apps |

**Trial length, three answers.** RevenueCat's trial-to-paid by length: 17 to 32 days 42.5%
(top quartile 59.4%), 10 to 16 days 35.4%, 5 to 9 days 37.4% (top quartile 52.8%), four days
or under 25.5% (top quartile 38.5%). Longer trials convert 1.7x the shortest, and 46.5% of apps
run the short ones, up 4.4 points in a year. Adapty's best one-year-LTV configuration is a
weekly $5.99 plan with a 3-day trial and its worst conversion configuration is a monthly $19.99
plan with a 7-day trial, which is a statement about single configurations, not a length test.
Mojo tested 3 against 7 days and found no trial-to-paid difference, so shipped 3 for faster payback. Length is a
conversion lever on RevenueCat's data and unproven as an LTV lever; test it, and read the
result on revenue per install rather than trial-to-paid alone.

**Cancellation timing.** Share of trial cancellations that happen on Day 0 and by Day 1
(RevenueCat 2026): 3-day trials 55.4% and 84%, 7-day 39.8% and 64%, 14-day 35.7%, 30-day
31.1%. These are shares of cancellations, not shares of trials. Over half of 3-day-trial
cancellations happen within hours, which is users protecting themselves against the fear an
honest trial timeline removes.

### Install to paid

RevenueCat's realised install-to-paid at day 35, beyond the access-model split above: Health
and Fitness 2.9%, Business 2.6%, Education 2.5%, Gaming 1.0%. Older cross-vertical percentiles
(RevenueCat, 2024): bottom quartile 0.6%, median 1.7%, top quartile 5% to 6%.
Adapty's direct install-to-paid with no trial is 4.9% averaged across plan types, and 1.1% to
2.8% on weekly plans by price bucket, the only plan type where direct purchase competes with
trial-gated conversion.

### Timing: when trials start and when people pay

Share of trial starts on Day 0: about 90% across most categories on both reports (Adapty,
North America: Entertainment 93.7%, Photo and Video 92.5%, Graphics and Design 92.4%, Utilities
91.6%, Productivity 89.6%, Lifestyle 88.6%, Health and Fitness 87.2%). Education is the
exception: 71.3% globally and 69.1% in North America, with 23.5% (26.4% in North America) of its trial starts after day 31. After day 3, trial starts drop
below 5% in every category (RevenueCat).

Four different Day-0 shares circulate and they are not interchangeable: the share of trial
starts on Day 0 (about 90%), the Health and Fitness trial-start share (82.1%, RevenueCat), the
share of paid conversions on Day 0 (about a third, RevenueCat) and the category range for paid
conversions (15.6% to 71.4%, Adapty, below). Purchases are slower than trial starts. Share of paid conversions on Day 0 (Adapty, North
America): Photo and Video 71.4%, Graphics and Design 62.7%, Productivity 47.8%, Entertainment
45.8%, Lifestyle 41.5%, Health and Fitness 29.7%, Utilities 17.9%, Education 15.6%. Education
and Health and Fitness peak in days 4 to 14 (55% and 37% of their paid conversions), which is
trials expiring. RevenueCat puts about a third of all paid conversions on Day 0 and more than
60% inside a week, with a hard-paywall spike on days 4 to 7 (25.7%, the 7-day trial expiring)
and a longer freemium tail (22.9% after week 6 against 15.3%). Set the conversion window by
access model: a day-7 readout on a freemium app cuts off a fifth of its revenue.

### Trial or no trial, by category

Twelve-month LTV per subscriber with and without a trial (Adapty 2026, global averages): annual
about $50 with a trial against $36 without; weekly $55 against $46; monthly $36 against $29.
Weekly plus trial grows from $7.40 on Day 0 to $54.50 over the year. First renewal on weekly
plans is 59.2% with a trial against 37.0% without, and the gap narrows after that: the trial
effect is front-loaded.

The category cut reverses the average in half the categories. Trials lift twelve-month LTV in
Utilities (+85.1%) and Health and Fitness (+63.6%) and reduce it in Productivity (-13.7%),
Lifestyle (-21.2%), Entertainment and Graphics and Design. Trial adoption is 82% in Education,
77% in Health and Fitness and 57% in Lifestyle, about seven in ten apps overall. Refunds run
higher on trial plans: monthly trial 14.9% and annual trial 15.5% of revenue, against 4% to
8.5% on the same plans without a trial.

### Retention, renewal and churn, to the point where it bears on the paywall

Year-one retention by plan (RevenueCat 2026, medians): yearly about 28% (range 20% to 40%),
monthly about 8%, weekly 1% to 2%. Annual churn in year one is about 72%, up from 56% the year
before, and 35% of all annual cancellations happen in month one. The first renewal is the
retention event: weekly renewal jumps from about 45% at the first to 75%+ after, monthly from
55% to 78%+, annual from 35% to 65%+, in every category, geography and price tier. Adapty's
twelve-month subscriber retention on trial plans: annual 19.9% (nothing happens until the first
bill), monthly 14.2%, weekly 5.5% (about 3.8% without a trial). Weekly LTV is conversion
volume times price, not longevity.

### Plan structure and price

Weekly plans are 55.5% of all app revenue on Adapty's data, up from 43.3% two years earlier;
monthly's share is shrinking. Health and Fitness runs the opposite way: annual is 60.6% of its
revenue and growing. Annual's share of all subscription durations fell from 41.4% to 33.6% in a
year (RevenueCat), while prices held. Annual-dominant apps earn $0.46 revenue per install
against $0.32 for weekly-dominant and $0.29 for monthly-dominant (RevenueCat, day 60), while
weekly plus trial produces the highest LTV per payer (Adapty): per-install counts the
non-payers, per-payer does not, and the right metric depends on acquisition cost. High-price
apps earn about 3x the twelve-month LTV of low-price apps, and median subscription prices grew
7% to 10% between 2023 and 2025, so a price untouched for two years is now below market.
Intro offers: only 9.3% of apps use promotional pricing at all; the median intro discount is
50.1%; hobby-tier apps take 65% to 99% of new payers through intro offers while the
top-performing tier takes 0% to 10%.

### What paywalls actually look like

Adoption across RevenueCat's 2026 paywall census: two plans shown 41% to 60% of paywalls by
category (the dominant structure), one plan 20% to 40%, three or more 6% to 27%. Highlighted
pricing 74.5%, more than one plan 59.2%, free-trial messaging 54.0%, cancel-anytime assurance
30% to 35%, testimonials 10% to 17%, countdown timers 1.4%, progress bars 0.2%. Most paywalls
scroll (59% to 76%) and roughly half are text-dense. The dominant CTA is "Continue", which is
slightly ambiguous about what happens next. Among high-intent buyers, 82% take whichever plan
is preselected, and switching the default from monthly to annual inverts the mix from 37% to
63% annual (Superwall network data; shares of buyers, not of all users). Only 9.3% of apps use
promotional offers, so a well-built decline path is still uncommon.

### Where to spend testing effort

Share of experiments of each type that improve LTV (Adapty 2026): localisation 62.3%, trial
structure 59.6%, plan duration 58.7%, price 45.5%, visual or copy changes
34.6%. That is win rate, not lift size; Adapty separately estimates pricing tests deliver up to
80% uplift against 30% for visual changes. Apps running 50 or more experiments earn 18.7x the
revenue of single-experiment apps, which is correlational: heavy testers are better-resourced
apps. Most teams spend their testing budget on the lowest-yielding category because it is the
easiest to ship.

### Web and in-app

In-app paywalls convert 45% better than web paywalls (1.60% against 1.10%, Adapty) and hold a
higher LTV throughout the year even after the store fee; web subscribers retain better through
month three because they are more deliberate buyers. A true web-to-app funnel is a different
thing from a web paywall: Reading.com ran its in-app onboarding identically on web and the web
version converted 50% better to trial and 30% better to paid in a controlled comparison, and

### The store step: no public number is trustworthy, and you do not need one

No published benchmark for store-page-view to install carries a disclosed denominator and a
sample size at the same time. The vendor figures that circulate disagree with themselves by 3x
between editions and mix impression and page-view denominators. Apple and Google both compute
this benchmark for you: App Store Connect (Analytics, Benchmarks) and the Play Console show
your listing against the 25th, 50th and 75th percentile of your peer group, set by category,
business model and download tier. Neither publishes the values. The one defensible public
figure is paid: an Apple Search Ads tap converts to an install about 56% of the time at the
median (AppTweak, 2025 data), the highest-intent click there is.

### The counterintuitive one

The paywall configuration with the lowest Day-0 conversion often produces the highest
twelve-month LTV (Adapty 2026). Optimising the conversion moment can work against the
relationship. Judge every paywall test on revenue per user over a renewal cycle, with first
renewal and refunds beside conversion.

---

## Evidence from outside the industry, and the pattern it supports

Controlled studies that put a number on the mechanism under a pattern. None of these is an app conversion result. Quote the referent as written.

| Pattern | The study and the numbers | What it means for a flow |
|---|---|---|
| Every question shows its reason | Langer, Blank and Chanowitz 1978. Asking to cut a copier queue for 5 pages: 60% agreed with no reason, 93% with an empty reason ("because I have to make copies"), 94% with a real one. For 20 pages: 24%, 24%, 42%. | On a small ask any reason works. On a big ask an empty reason performs exactly like none and only a real one moves anything. A permission or a payment is the big ask. |
| Value before the ask, small yes before big yes | Freedman and Fraser 1966. Homeowners asked cold to put a large ugly road-safety sign on the lawn: 16.7% agreed. Asked two weeks after agreeing to a 3-inch window sticker: 76.0%. | A small prior commitment more than quadrupled agreement to a large one. It is the mechanism under micro-commitments and under asking for the account after the plan. |
| The primer screen is only as good as what it says | Woolf and Edwards 2021, 107 randomised trials of contacting people before a request. Overall odds ratio 1.33 for responding; 1.09, not distinguishable from nothing, in the low-bias trials. | An extra screen before the ask buys nothing by itself. The content of the screen, the user's own reason, is the whole effect. No published test compares a push primer against none. |
| Make the choice explicit at the ask | Keller, Harlam, Loewenstein and Volpp 2011, CVS field experiment, 11,182 members. A plain opt-in box for automatic refills: 12.3% enrolled. A forced choice with the alternative spelled out ("I prefer to order my own refills"): 21.9%. | Spelling out what declining costs, and making the user actually choose, nearly doubled take-up. This is an active-choice design inside the request, not a primer before it. |
| Start people with progress already earned | Nunes and Dreze 2006, car wash loyalty cards. A 10-stamp card issued with two stamps already on it was completed by 34% of customers; an 8-stamp blank card requiring the same eight purchases, by 19%. | Endowed progress. It is the mechanism under reverse trials, pre-filled progress bars and "you have already started" screens. |
| An explicit commitment in the user's own terms | Gollwitzer and Sheeran, meta-analysis of 94 studies on implementation intentions, effect size d = .65. Separately, stickK reports 78% goal achievement with stakes and accountability against 35% without, on a self-selected platform population. | Forming an explicit "when, where, how" commitment roughly doubles follow-through on goals in general. It is not a conversion effect and should not be translated into a percentage; it is why a signed, personalised commitment screen has good priors and why no app has published a test of one yet. |
| Value before the ask, in both directions | Duolingo moved account creation after the first lesson and reported a roughly 20% lift in next-day retention, A/B tested. Pinterest gated content behind signup and reported a 50% conversion lift. | These do not conflict. Pinterest's feed is instant value on first scroll, so its gate sat after value. Where value needs a plan built first, the ask belongs after the plan. The rule is where value lands relative to the ask, not "always defer". |

## Numbers to stop quoting

Figures that circulate in this space and do not hold up as usually stated. An audit that cites one loses the room. Each line says what the number actually is.

| As usually quoted | What it actually is |
|---|---|
| "Users spend 3 to 6 seconds on a store page and 60% never scroll past the first image" | Attributed to StoreMaven; appears in no report or library. StoreMaven's real published figure is a localisation lift of 5.1% to 24.6%. The sourced versions are about 7 seconds on the page and about 90% never scrolling past the third screenshot. |
| "82% of purchases happen in the first session" | No such figure in either 2026 report. Day-0 share of paid conversions runs from 15.6% (Education) to 71.4% (Photo and Video); RevenueCat puts about a third of paid conversions on Day 0. The 82 is the Health and Fitness Day-0 share of trial starts. |
| "$12.33 a week beat $29.99 a week: framing lifted install-to-trial 31%" | A bundled RevenueCat redesign of a party-game paywall: layout, plan count, a trial toggle, a savings badge and the weekly framing all changed at once. No single element can claim the lift. The clean framing evidence is Mojo's monthly-equivalent shown next to the real monthly price. |
| "RevenueCat found reverse trials take conversion from 0.4% to 4.5%" | A practitioner case from Steve P. Young quoted inside RevenueCat's 2026 report, not platform data. One app, freemium conversion, no control described. |
| "Personalising the paywall with the user's first name lifts conversion up to 17%" | A vendor marketing claim with no study, sample or app behind it. Geo and goal personalisation are well supported; name insertion is not. |
| "Animated paywalls convert 2.9x" | One crypto app during seasonal discount promotions, with no control group by the author's own account. Often misattributed to a paywall vendor. |
| "46,000 paywalls with performance data" | Screenshot databases with estimated app-level revenue, not conversion data tied to a design. No public database connects a specific paywall design to a measured result. |
| "Yazio gets 78% push opt-in from its primer", "Zero 26%", "Runna 2x", "Aaptiv 3x" | The apps' own screen copy, with no comparison group. Fine as examples of the device, not as evidence it works. Airship's NHL case reports two different lifts for the same change. |
| "Priming a permission lifts opt-in 2-3x" | Vendor aggregate claims that trace to other blog posts, never to data. The closest real study (Tan et al. 2014) tests the reason string inside the system dialog on hypothetical prompts, not a primer screen. |
| "The trial-to-paid rate for AI apps is 8.5%" | That is RevenueCat's trial-start rate for AI apps (share of installs starting a trial), against 5.6% for non-AI. AI trial-to-paid is about 20.6% against 25.6%. |
| "Hard paywall apps convert 10.7% of trials" | RevenueCat's 10.7% is realised install-to-paid at day 35 for hard-paywall apps, not a trial-to-paid rate. Their own blog summary labels it loosely. |
| "Apple says 3% of downloads convert to paid" | From an Apple developer-session demo that runs on artificial data. Apple publishes no benchmark values; they exist only inside your own App Store Connect account. |
| "Duolingo's delayed signup lifted next-day retention 20%" | Directionally real and A/B tested, but the 20% and the 12% baseline come from secondary sources and it is ambiguous whether it is a relative lift or a jump to 20%. Say "roughly 20%, as reported" and do no arithmetic on it. |
| "Noom proved long quizzes convert" | Noom has never published a quiz-versus-no-quiz test. Its length is an assumption the industry copied, not a result. The one published long-flow number is Sesame Care's 25 steps lifting conversion 40%, reported second-hand by Ravi Mehta with no methodology. |
| "Onboarding paywalls with a trial convert 1.78%", or 1.78% beside 0.89 / 0.82 / 0.76 | Adapty publishes two versions of one metric: 1.78% on the report page, and 1.35% with 0.89 / 0.82 / 0.76 for the other three cells on the blog. Quote one set whole and never mix them: the four-cell 1.35% set wherever cells are compared (it is what the served benchmarks carry), 1.78% only as a stand-alone headline. |
| "The median app turns 5.5% of installs into a trial" | RevenueCat 2026's cross-vertical median is 6.2% (top decile 20.3%); the 5.5% was an early read of the chart. Adapty's 11.2% for the same step is an average over a smaller, iOS-heavy panel, not a conflicting median. |
| "Mojo's monthly-equivalent reframe lifted conversion 26% in Mexico" | The referent is new revenue per paywall impression, not conversion rate: +45% Brazil, +26% Mexico, +8% USA for the same change. |
| "35% of annual subscribers cancel in month one" | 35% is the share OF annual cancellations that happen in month one, not a cancellation rate. Year-one annual churn is about 72%. |
| "55% of three-day trials are cancelled on day zero" | 55.4% is the share OF three-day-trial cancellations that happen on Day 0, not the share of trials cancelled. |
| "82% of buyers take the default plan" | Superwall network data on buyers who did not close the paywall: 82% of those take the preselected plan. It is a share of buyers, not of all users who saw the paywall. |
| "Web checkout keeps about 27% more revenue than in-app purchase" | Gross fee arithmetic against Apple's 30% tier only (30% vs about 3-4% for a payment processor). Against the 15% small-business and second-year rate the gap is about 11-12 points, and neither figure counts the conversion lost when users leave the app to pay. A take-rate gain, not a revenue result. |

---

## The judgment calls an audit has to make

The questions a growth lead actually asks, answered from the data above so an audit can take a position instead of listing options.

**Hard paywall or freemium?** A hard paywall converts about 5x more installs to paid (10.7% against 2.1% at the median) and earns 8x the revenue per install, and year-one retention is the same either way (about 27% against 28% on annual). The fear that hard paywalls churn out does not show in the data. Two caveats: hard-paywall conversion fell year over year (12.1% to 10.7%), and if growth depends on network effects or virality, free users are the acquisition engine and the calculus flips. Most apps are neither pure case: a free tier plus an onboarding paywall is freemium with an uplift, and should be benchmarked as such, not as a hard paywall.

**Should every app run a trial?** No. Trials lift twelve-month LTV in Utilities (+85%) and Health and Fitness (+64%) and reduce it in Productivity (-14%), Lifestyle (-21%), Entertainment and Graphics and Design. Trial-structure tests carry the second-highest win rate of any test type, so in the four negative categories the right move is to test trial against no trial rather than default to one.

**Weekly or annual?** The two reports measure different things. Weekly plus trial produces the highest twelve-month subscriber LTV (about $54); annual-dominant apps earn about twice the revenue per install ($0.46 against $0.32), because per-install counts the non-payers. High acquisition cost pushes toward weekly plus trial, low toward annual. Health and Fitness is the structural exception where annual is 60.6% of revenue and growing. Monthly is a dead zone either way at 0.3% to 0.6% install-to-trial.

**When should the paywall appear?** For most categories, in the first session: about 90% of trial starts happen on Day 0 and intent decays in minutes. Rootd moved its paywall into onboarding and revenue went 5x. Education is the exception, with only about 70% of trials starting Day 0 and a quarter starting after a month, so a first-session hard wall there optimises for a moment its users do not have. "After the value moment" and "days later" are different claims.

**Is a long onboarding a problem?** Question count was never the variable; unexplained asks are. Across 89 flows question count varies sixfold within one vertical. Sesame Care added steps to 25 and conversion rose, because each step built decision confidence for a high-stakes purchase. Score a flow on unexplained asks and on whether each step raised the odds of reaching value, not on screens.

**Will a transparent trial timeline just teach people to cancel?** They already cancel: 55.4% of three-day-trial cancellations happen on Day 0 and 84% by Day 1, as self-protection. Blinkist made the timeline explicit and trial starts rose 23%, support complaints fell 55% and push opt-in went from 6% to 74%. The cancel-anxious users were not converting anyway; transparency converts the fence-sitters. One caveat: the timeline paywall sells a routine, so it fits habit products. A utility should state the job and the price.

**What should we test first?** In descending share of tests that improve LTV: localisation 62%, trial structure 60%, plan duration 59%, plan count 57%, price 46%, visual and copy 35%. Most teams run that list bottom-up because visual tests are easiest to ship. Layout tests only get big when they change which plan people see first.

**Show me one causal result.** Almost everything in onboarding is correlational and the canonical activation metrics were rally cries, not science. Two clean A/B results exist: Calm surfacing the daily-reminder prompt after the first meditation took reminder setting from under 1% to 40% with about 3x retention, and Blinkist's explicit trial timeline. Both are trust-and-transparency changes.

**Doesn't Pinterest disprove "value before the ask"?** Pinterest's gate sat after a taste of the feed, which is instant value. Where value is instant an early ask is already after value; where value needs a plan built first, the ask belongs after the plan, which is what Lifesum, MyFitnessPal and Yazio do. Sequencing is relative to when value lands, not to screen number.

**Web checkout or in-app?** In-app paywalls convert about 45% better (1.60% against 1.10%) and keep a higher LTV even after the store fee; web retains better in the first three months because its buyers are more deliberate. The exception is a real web-to-app funnel: Reading.com ran an identical onboarding on web and in-app and the web version converted 50% better to trial and 30% better to paid, controlled. The container changes the answer as much as the flow does.

**What about AI apps?** The problem is the top of the funnel, not monetisation. AI apps start trials at about half the average rate (5.3% against 10.9%), convert trials a little worse (20.6% against 25.6%) and then earn 41% more per payer in year one, with weaker retention (-31% on annual). The lever is paywall comprehension and trust before the trial, not trial-to-paid.

---

## Three things that are easy to get wrong

**Question count is not the lever.** Across the teardowns, question count varies sixfold inside
a single vertical and no segment opens the same way twice. There is no standard sequence to
copy. What the strong flows share is a reason attached to every ask.

**Onboarding completion rate is a distraction metric.** What predicts paying is whether the
steps taken raised the odds of reaching value. A user who completes 25 explained steps beats
one who completes 5 unexplained ones.

**The template travels, the trust signal does not.** The structural playbook is nearly
universal: annual default priced per month, a three-step trial timeline, two plans, a
strike-through anchor. How a category earns trust is not. That is the part to design rather
than copy.

---

## How to read the evidence in this document

Four kinds, and they do not carry equal weight.

**Published experiment results.** A/B tested, with a named company and a named source. The
strongest thing here. Where a figure comes from secondary reporting rather than the company
directly, it says so.

**Industry data.** The RevenueCat and Adapty annual reports, 2026 editions, covering roughly
115,000 and 16,000 apps respectively. Reliable for direction. Always check what a percentage is
a percentage of.

**Teardown prevalence.** 89 onboarding flows and 52 paywalls across health,
wellness, fitness, sports, finance, kids and education. This tells you what apps do and how
commonly they do it. Prevalence is not proof that a pattern works.

**Behavioural research.** Controlled studies from outside the industry, listed with their
referents in the research table. They explain why a pattern might work and put a real number
on the mechanism; they are not app conversion results and should never be quoted as one.

The mechanisms named throughout, loss aversion, sunk cost, the goal-gradient effect, the IKEA
effect, are standard behavioural psychology. They explain why a pattern might work; they are
not evidence that it did.

---

*Otter Growth. Product growth for consumer subscription apps.
[ottergrowth.com](https://ottergrowth.com)*