Price Power

Michael Berliner was a former growth lead at Masterclass and Calm, and an absolute rockstar on onboarding, paywalls, and activation for subscription apps. 

Michael built one of the first quiz-style onboarding flows at Masterclass, which lifted conversion 20%, then revamped Calm's self-serve flow for a 50% conversion lift and a 30% lift in return retention!

How does more friction increase conversion?! We get into why it works, when it doesn't, and the paywall tactics Michael tests with clients today, from trial timelines to reverse trials to hiding the monthly plan.

Timestamps
00:00 Intro
01:00 Michael's background at MasterClass and Calm
01:50 MasterClass's quiz flow and the 20% conversion lift
04:20 Personalizing onboarding question by question
05:30 Can smaller apps personalize with less data?
06:00 Branching flows vs. simple personalization
10:30 Calm: a 50% lift from rebuilding the self-serve flow
11:40 Calm's activation flow and the 30% retention lift
14:00 Why ad click to trial start is the most important part of your app
16:00 Trial start rate vs. trial quality
17:25 Where to put account creation
20:55 Why long quiz flows only work for some categories
22:40 Matching onboarding to the acquisition source
23:55 Putting part of the core product inside onboarding
26:05 Setup, aha and habit moments
28:20 Building several small aha moments
32:25 How onboarding has changed over the last 5 to 10 years
36:25 Blinkist's trial timeline screen
40:20 Hiding the monthly plan behind "See all plans"
42:50 Matching plans to user intent
46:00 Reverse trials after someone closes the paywall
48:40 Upsells right after purchase
50:00 A discount right after the trial starts
52:15 The paywall best practice Michael disagrees with
53:55 Why your A/B tests are lying to you
56:40 Why copy and design tests rarely move paywalls
57:20 Where to find Michael

What you'll learn:
• Why a longer quiz onboarding beat a "Get Started, go to checkout" flow at Masterclass
• How to decide between one cohesive flow and separate branches per segment
• Why splitting traffic into segments can quietly kill your experiment velocity
• Why install to trial start shapes activation and retention downstream
• How to place account creation so users see a reason to give you their email
• Which product categories need intent-building onboarding and which need speed
• How Duolingo and Tolan use a slice of the core product before the paywall
• Why the Blinkist trial timeline raised trial starts and cut complaints
• How hiding the monthly plan pushed Mojo's annual share from 60% to 80%
• What to show users who close the paywall, and what to sell users who just paid

Key Takeaways:
• Friction that builds intent raises conversion. Masterclass added a quiz flow inspired by Noom and conversion rose 20%, while product leadership insisted it couldn't work.
• Start with one flow and earn your segments. If a segment gets 30% of traffic, a 10% win there is a 3% win overall, so only split when use cases truly differ, like sleep vs general meditation.
• Install to trial start sets up everything after it. The average install to trial rate is about 6%, and well-optimized onboarding can convert 2 to 3 times better than average.
• The paywall is the most anxious moment in your app. Blinkist's trial timeline screen lifted trial starts 23% and cut complaints 55%, because telling people how the trial works builds trust.
• Defaults decide plan mix. Mojo moved the monthly plan behind "See all plans" and annual share went from 60% to 80%. Keep that button visible so price-sensitive users can still find monthly.
• A "no" at the paywall still signals intent. In one case study, users who closed the paywall converted at 0.5% on freemium, and a 7-day reverse trial lifted that 10x.

Links & Resources
Otter Growth: https://ottergrowth.com
Michael's "Ideal Flow" presentation: https://ottergrowth.com/boa
Michael Berliner on LinkedIn: https://www.linkedin.com/in/berlinermichael/
Botsi: https://botsi.com

We talk with An Nayal, founder of Flamme, and now Agniverse, a UGC creator agency. An grew Flamme with 200 tiktoks a day, leading to over 50M views, and 200k+ downloads! 

But still, he started with poor retention, didn't have monetization early on, and made tons of mistakes along the way. And that's mostly what we talk about! Wins are nice, but our mistakes are what we learn the most from.

Timestamps

00:00 Cold open
00:12 Intro
02:22 Launching Sparks and getting into TechCrunch Disrupt
03:27 The trademark notice that forced a rename
04:27 Zero retention and talking to dozens of couples
05:11 ChatGPT launches and Flame becomes an AI couples coach
06:32 Early traction: throwing everything at the wall
08:26 Why manual user outreach still matters
10:00 Second TechCrunch feature, still 4.2% day-30 retention
10:59 Finding the daily question in Mixpanel
12:06 Partner triggers and the carrot-and-stick system
13:34 Adding monetization after launching free
14:53 Charging per couple vs. per person
15:45 How Flame measured retention
16:44 Pricing by Twitter FOMO and competitor research
18:12 The trap of small-sample A/B tests
19:13 Three distribution phases: PR, paid, creators
20:44 The RAG fundraising mistake
21:37 Going all in on TikTok with a dozen iPhone 8s
23:01 Why organic is a fair testing ground
24:06 50M views, 200K installs, and why memes don't convert
24:50 Structuring UGC creative that converts
26:32 Referral programs and the K-factor ceiling
30:43 Why An decided to sell Flame
31:34 Six months of rejections from dating and mental health players
33:00 The LinkedIn message that led to an offer
34:36 Inside the performance-based deal structure
35:53 Founder burnout
36:43 The two walls: building and distribution
38:22 Adniverse: creators vs. influencers
39:11 Handler: engineering virality with AI
41:15 Consistency and loving the process

What you'll learn:
• Why press gets you a launch spike but never a growth channel
• How to find the one feature your retained users come back for
• How a partner-triggered "carrot and stick" loop lifted day-30 retention to 50%+
• Why charging for features that used to be free backfires, and what to paywall instead
• Why Flame charged per couple instead of per person
• Why small apps should fix distribution before running pricing experiments
• Why memes convert at 0.05–0.1% while product-focused videos hit 0.5–1%
• Why K-factor is capped in a two-person product
• How a "no ask here" LinkedIn message turned into an acquisition offer
• The difference between creators and influencers in a UGC strategy

Key Takeaways:

• Follow the retained users, not the roadmap. The AI coach got the press, but retained users came back for the daily question. Flame stopped building new features and A/B tested that one until it was ten times better.
• In a multiplayer app, users re-engage each other. A morning nudge ("extend your streak") and an evening warning ("you'll break their heart") turned partner activity into the retention engine.
• Paywall new value, not existing features. Flame kept the free product free and put newly requested features behind the paywall.
• Charge the household. One partner usually drives the purchase, and if either partner stops using a two-person product, both churn.
• Not all views are equal. Flame's first 20 to 30 million views came from memes that barely converted. Videos showing the product solving a real problem converted 5 to 10 times better.
• Cash incentives produce low-intent referrals. Nobody tells a friend about an app to earn five dollars. Ask for the referral at the aha moment.

Links & Resources

Agniverse: https://agniverse.co/
Handler: https://gethandler.ai/
An Nayal's website: https://annayal.com/
An Nayal on X: https://x.com/consumerxai
An Nayal on LinkedIn: linkedin.com/in/annayal

We’re talking with Gabe Kwakyi today, who was actually our first guest! He is a mobile app OG, was previously the CEO of Lingvano, the #1 sign language app, and is now doing consulting, advising and fractional head of growth for apps. As well as building some stuff on his own.

This is a pretty cool episode because Gabe has been testing ChatGPT ads, and we get to learn from his real experience about how they work, what works today, and what the future could look like.

Gabe has run ChatGPT Ads tests across his own portfolio of apps and several client apps since the self-serve platform launched in May 2026. He walks through the setup, shares the context hints that flopped and the ones that drove conversions, and explains why an app icon beat every other creative he tried. He also makes the case that ChatGPT Ads will become a serious channel once OpenAI has enough conversion data; until then, most tests will fail.

What you’ll learn:
• How ChatGPT Ads are structured: campaigns, ad groups, ads, and the handful of targeting options available today
• What context hints are and why they’re the main lever you control
• Why ChatGPT Ads deliver decent CPMs, CPCs and CTRs but very few trials or purchases
• How Gabe’s test across eight apps turned out, and why most apps saw zero purchases
• Why users on ChatGPT’s free and Go plans may be harder to convert into subscribers
• Why OpenAI’s missing conversion data, not its targeting tech, is the real blocker
• How conversation “turns” affect whether a user is ready to see an ad
• Which context hint patterns worked: conversation context, user goals, rejected ideas, and the “layup insertion”
• Why keyword-style context hints only sometimes work
• Why a plain app icon outperformed screenshots and Meta creative
• What URL parameters ChatGPT passes and what the “oppref” parameter is
• How to tailor your onboarding to ChatGPT traffic intent
• How to tell whether ChatGPT Ads are lifting your organic ChatGPT referrals
• When it’s worth testing now and when to wait
• Gabe’s feature wish list for the ChatGPT Ads team

Key Takeaways:

Context hints are your creative. ChatGPT Ads give you almost no creative surface, so the context hint does the work that ad creative does on Meta. Gabe spends his testing budget there first and treats headline and copy as secondary.

Clicks are cheap; buyers are rare. Across Gabe’s eight-app test, CTR averaged 1.68% and install rate 3.36%, but purchase conversion averaged 0.14%. Most apps saw zero purchases. Industry reports match: one review of 35 case studies found 22 with zero conversions.

The blocker is data, not targeting. Meta became a great performance channel because every advertiser sent conversion events back. ChatGPT Ads launched in May 2026 and hasn’t collected enough postbacks to know which users buy. That will change as more advertisers integrate their MMPs.

Target the moment, not the topic. “Users who want date ideas” brings volume. “User keeps rejecting every date idea and wants better ones” brings people who still have the problem and are open to a new solution. Gabe’s best results came from these frustrated, self-qualifying moments.

Use your app icon. The ad image is a small square next to the answer. Screenshots and text-heavy banners are unreadable there. An icon looks like a product recommendation, which is what users expect.

Testing is cheap, so treat it as learning. The minimum budget is $25 a day and setup takes minutes with an MMP tracking link. Gabe recommends testing every couple of months so you’re ready when performance turns, and skipping it if you need a new channel to work this quarter.

Check your organic referrals. One advertiser spent $600 with no attributed conversions but saw organic ChatGPT referrals rise. If you test, watch untracked traffic and “how did you hear about us” answers.

Links & Resources
• Gabe Kwakyi on LinkedIn: https://www.linkedin.com/in/gabrielkwakyi/
• Rhythmic apps: https://rhythmicapps.com/

Timestamps
00:00 Why Gabe started testing ChatGPT Ads
01:45 How the ChatGPT Ads platform works
03:15 What context hints are
07:30 Great clicks, low conversions: results from eight apps
10:00 Free and Go users, and the missing conversion data
13:00 Is ChatGPT closer to search or social?
15:30 Conversation turns and the right moment for an ad
17:30 Rejected ideas and the “layup insertion”
20:45 Why campaigns are so volatile
22:00 More context hint categories to test
26:00 Personalizing onboarding by traffic intent
28:30 URL parameters and tracking
30:00 Why the app icon beats every other creative
35:15 Is it worth testing right now?
38:30 The organic referral halo
41:00 Gabe’s feature wish list
46:15 Have fun with it: the return of real marketing

Jacob Rushfinn, CEO of Botsi, has spent a decade in consumer apps figuring out how to convert users. He compiled 13 principles for your onboarding flow that will make a difference. These are all from experimentation and research from top apps like Opal, Finch, Headway, Tiimo, Tolan, Zing, and more.

What you'll learn:

• Why the first screen is a drop-off point most teams never measure
• How Opal, Rise Science, and Headway hook users before they tap anything
• Why story-style intro screens beat any UI pattern you invent
• Why simple string personalization beats complex personalization almost every time
• How to make quiz questions feel like a conversation instead of a form
• Why letting people build something (a pet, a name, a journal entry) raises commitment
• How to show users the math on the value they'll get
• How to build trust across the whole flow: answer the biggest fear, lead with emotion, back it with numbers
• When to prime permission prompts and when to just fire the native one
• Why there is no excuse left to skip video, with examples from Lingokids, BoldVoice, and Ladder
• How to turn loading screens into selling screens
• How to make the free trial impossible to miss, and what to do when users close the paywall anyway

Key Takeaways:
- Your first screen is a drop-off point. It can lose 10 to 30 percent of new users before they tap anything. A plain welcome, get started, log in screen works for big brands; everyone else needs a hook that's interesting and easy, like Opal's glowing rock or Finch's "Hatch a new pet" button.

- Ask for the account when intent is highest. Tiimo moved account creation from right before the paywall to right after the welcome screen and lifted both account creation and trial starts. It also stops you from asking for an email and a payment back to back. Test it, or skip account creation until after the paywall.

- Say the user's goal back to them. Ask the goal, ask why it matters, then repeat their words on the next screen, in the plan recap, and in the paywall headline. Zing, Tiimo, and Impulse all do this, and it rarely fails to lift conversion.

- Let people build something. People value what they put work into. Finch has you design a bird, which is fun, low-effort, and raises no privacy concerns. Add presets and shuffle buttons so nobody stalls on a blank text field.

- Show them the math. Turn inputs into a concrete number: Opal's five years of your life, Rise's sleep debt, Headway's 38 book summaries a month. When users hit the paywall, they should already know what they're getting.

- Build trust before you ask for money. Most apps cram trust into the paywall. Better apps spread it across the flow: answer the user's biggest fear (privacy, for Flo), lead with emotion, back it with a number, and repeat your strongest social proof.

- Make the trial impossible to miss. People don't read paywalls. They see a price and leave, even when the trial is visible. Trial reminders, choose-your-trial screens, and multi-page paywalls force users to notice it. When they close the paywall, follow up with a second offer.

- Onboarding doesn't end at the paywall. Don't drop a new subscriber onto an empty home screen. Imprint sends you straight into your first lesson; Balance auto-starts a sleep meditation for people who picked sleep. This is what separates good trial starts from good trial conversion.

Links & Resources
- Price Power Podcast: https://pricepowerpodcast.com
- Botsi onboarding flow library: https://botsi.com/resources/flows
- Retention.blog: https://retention.blog
- Jacob on LinkedIn: https://www.linkedin.com/in/jacob-rushfinn/

Timestamps
00:00 Why onboarding decides whether your app is worth paying for
00:45 1. Hook them on the first screen
03:00 2. Borrow UI patterns people already know
04:00 3. Ask for account creation when intent is highest
05:30 4. Ask the user's goal, then say it back to them
07:05 5. Make onboarding questions feel like a conversation
08:15 6. Let people build something
09:30 7. Show them the math
10:30 8. Build trust before you ask for money
12:48 9. Earn the permission opt-in
14:15 10. Use video, because it's cheap now
16:05 11. Turn loading screens into selling screens
18:00 12. Make the trial impossible to miss
20:20 13. Onboarding doesn't end at the paywall
22:13 Recap of all 13
24:15 What to test first and which metrics to watch

Philip Pages, founder of Redux Payments, explains why moving off Apple's in-app purchases quietly strips away a retention machine most founders never knew they had, how failed payments actually work, and when payment orchestration is worth the extra complexity.

Philip has looked inside hundreds of app payment accounts, and most of them are leaving real money on the table. He walks through the difference between a customer who wants to churn and a card that simply had no funds that morning; why recovered subscribers stick around for another three to six billing cycles; and the single setting that quietly blows up Stripe accounts. Jacob and Philip also get tactical on the app-to-web paywall pattern that converts, the retry schedule that recovers cards without triggering the card networks, and a meditation app that got back six figures by changing nothing but the timing of its retries.

What you'll learn

• Why Apple's in-app purchase system silently recovers failed payments, and exactly what you give up when you leave it
• Why most failed payments have nothing to do with a customer wanting to cancel
• How recovered subscribers behave once you win them back (they stay another three to six billing cycles)
• Why the "30% down to 5%" margin pitch for web billing is misleading
• How to tell whether your app is even ready to test web billing
• Why older, Facebook-native audiences convert better on web-to-app flows than younger ones
• How to set up web payments without blowing up your Stripe or Paddle account
• What chargebacks, early fraud warnings, and approval rates actually do to your account health
• How merchant-of-record providers take ownership of your transactions, and what that costs you later
• What payment orchestration is, and the revenue point where it starts paying for itself
• How one meditation app recovered over $150K by fixing retry timing by locale
• How long you should really retry a failed card, and why spamming it backfires
• The compliant two-button app-to-web paywall pattern that didn't exist a year ago

Key takeaways

• Apple runs a retention machine you never see. Behind the scenes, Apple retries cards near paydays, fixes bank issues, and serves in-app prompts for up to 60 days. It feels like magic, but it is a system. Move to web and that system is now your job.

• Most failed payments are not people trying to leave. Insufficient funds at one moment in time and random bank rejections make up the bulk of failures. Treating every failure as a lost cause writes off customers who already paid you for months.

• The margin math is messier than the pitch. Providers love to say you'll save 25 to 27 percent. Philip says heavily discount that. Even so, saving 15 percent and reinvesting it in growth can compound fast, so the channel is still worth testing.

• Don't touch web billing before product-market fit. Under a million in ARR with no PMF, the only job is finding PMF. Optimizing payments early is a distraction. Post-PMF, commit real budget or don't bother.

• Cancel your subscriptions. The number one way apps wreck their Stripe account is leaving failed subscriptions open. They stack, they collect chargebacks, and one subscription that should have had a single chargeback ends up with seven.

• Retry timing beats retry volume. One app had 64 percent of failures come from insufficient funds. Retrying by local payday instead of at random recovered over $150K. Brute-forcing the card just gets you flagged by the networks.

• Orchestration is where big brands quietly win. Netflix, Spotify, and other large subscription companies route payments across multiple processors to lift approval rates. Most smaller apps have never even considered it.

• The app-to-web playbook is new, and it works. Two CTA buttons, a rebuilt mini-onboarding before the paywall, and Apple Pay on by default. Philip's word for the pattern Jacob describes: gold. It did not exist a year ago.

Links & resources
• Philip Pages on LinkedIn: https://www.linkedin.com/in/philip-pages-a881b5139/
• Redux Payments: https://www.reduxpayments.com/

0:00 Intro
02:00 Intro and Philip's first company (peaked near $3.5M ARR, then shrank)
04:11 Apple's hidden failed-payment machine, and what you lose moving to web
07:00 What a failed payment actually is, and why most aren't churn
11:05 The real margin math behind leaving the App Store
12:16 Why web billing is hitting critical mass now
17:20 Matching the right user to the right payment method
21:33 Getting started: providers, merchant of record, subscription management
25:47 How to not blow up your Stripe account
28:43 Chargeback alerts and fighting chargebacks
30:40 Merchant of record: owning the transaction, and getting out
34:27 Payment orchestration and cascading across processors
39:06 When web billing is a mistake (pre product-market fit)
44:47 App-to-web compliance and the two-button paywall playbook
52:32 A failed-payment recovery teardown (the six-figure meditation app)
59:02 How long to retry a failed card
01:03:00 The scariest account story, and what Redux Payments does

George Deglin, CEO and co-founder of OneSignal, explains why calendar-based lifecycle marketing holds subscription apps back, why SaaS dashboards are losing their place as the primary interface, and what autonomous lifecycle marketing looks like in practice.

George walks through OneSignal's autonomy ladder, from L0 (a traditional SaaS dashboard) to L4 (a self-improving agent). He explains how each level builds customer trust, why behavioral triggers outperform scheduled messages, why email remains overlooked by consumer apps, and why companies worried about token costs should build features that cost more—not less.

What you'll learn:

• Why lifecycle messaging should respond to product behavior instead of a marketing calendar

• Why behavioral triggers outperform scheduled sends by 4x to 9x

• How product and lifecycle-team silos weaken message quality

• Why email is an underused, inexpensive retention and win-back channel

• What George means when he says "dashboards are dying"

• What happened when OneSignal asked 20 customers whether they preferred its dashboard or Claude, Gemini, or ChatGPT

How the L0-to-L4 autonomy ladder moves from assistance to recommendations and independent execution

• Why most companies pursuing ML personalization are optimizing the final 5% before mastering the basics

• Whether AI could mean fewer lifecycle-marketing hires

• How AI inbox filtering will raise the bar for message relevance

• Where app-to-web billing and RCS payment flows could go next

• Why OneSignal won't charge separately for AI—and why George favors expensive AI use cases today

Key takeaways:

• Calendar sends are the default—and the problem. Moving from scheduled messages to behavioral triggers is a larger opportunity than adding AI-generated personalization to a weak foundation.

• Email is inexpensive and still reaches users after they uninstall. The main obstacle is organizational coordination, not technology.

• Dashboards may fade, but product intelligence remains valuable. Users will increasingly operate software through agents while vendors expose more configuration through APIs.

• The autonomy ladder is also a trust ladder: assistance, proactive recommendations, delegated execution, and finally guarded autonomy that knows when to ask for help.

• Complex personalization can move a mature program from 95% to 100%, but many teams are overlooking the fundamentals.

• AI may eliminate tedious lifecycle work and reduce some hiring, while also making effective lifecycle marketing affordable for more companies.

• Today's expensive AI capability may become inexpensive as model costs fall, so George argues for building ahead of the cost curve.

Links & resources

• OneSignal: https://onesignal.com
• George Deglin on LinkedIn: https://linkedin.com/in/gdeglin
• George's essay on autonomous lifecycle marketing: https://onesignal.com/blog/the-future-of-lifecycle-marketing-is-autonomous/

Chapters

00:00 Cold open and episode introduction
01:00 What separates apps that win at lifecycle marketing
02:25 The most underused channel in consumer apps
03:15 Why teams default to paid media over email
04:30 Lifecycle marketing as an extension of the product
05:35 Retention, LTV, and who owns it
06:55 Change, provocative claims, and pushback
07:55 "Dashboards are dying"
08:05 Where the idea started with browser and computer use tools
10:10 Asking 20 customers: dashboard or agent?
11:15 Does every tool become a transactional layer?
12:40 Headless software and configuration APIs
14:50 The visual feedback loop problem
16:40 Previewing an email campaign inside Claude
18:05 The autonomy ladder, L0 to L4
20:00 Moving from assistance to recommendations
23:50 What customers are doing with L1
25:00 Saving customers hours on reports
26:00 L2: the recommendation engine
27:30 Democratizing customer engagement
28:40 Will competitors learn from my data?
30:30 AI and one to one personalization
32:20 Why triggered messages perform 4x to 9x better
34:15 Personalization versus segmentation
35:00 The 95% to 100% problem
36:25 Will AI replace marketers?
40:05 Why cheaper lifecycle marketing creates more companies
41:25 Does easier sending lower quality?
43:00 AI filtering on the receiving end
44:50 Pricing, packaging, and subscriptions
46:20 App to web billing
49:10 RCS payments inside the conversation
51:30 Token costs and why OneSignal won't charge for AI
54:30 Wrap

Michal, product growth lead at Applica Agency, explains why moving a paywall to the point of peak anticipation beat waiting for the aha moment, how trial length changes trial starts rather than trial conversion, and why the same paywall test produced opposite results on paid and organic traffic.

Michal walks through three client teardowns. A client went from a buried feature-gate paywall to an onboarding paywall and grew ARR 9x; trial start rate moved from roughly 3% to 16%. Beducated, a sex education web funnel, lifted revenue per user 47% on Meta traffic and 27% on organic after Applica discovered the two channels wanted completely different plan structures. Alux, a wealth and finance content app charging $149 a year, improved day one retention 22% and install-to-paid conversion 26% by mining users' own free-text answers out of Mixpanel and rewriting the onboarding in their words.

What you'll learn:

• Why anticipation beats the "aha moment" as a buying trigger
• What a paywall exit-intent survey revealed about price objections
• Why device buyers felt they were paying twice for a subscription
• How trial duration moves trial starts but leaves conversion untouched
• Why a 14-day trial beat every discount as an exit-intent downsell
• How that downsell came to drive 10-15% of total revenue
• When per-placement paywalls are worth building, and when they're not
• How often to paywall a retained free user base
• Why one A/B test can land positive and negative at the same time
• Why plan length is really about how far ahead users can picture themselves
• How a weekly plan works as the web's version of a free trial
• Why "Start learning" beat "Subscribe now" for conversion in 2026
• How to cheaply validate a post-purchase upsell before building one
• How to turn Mixpanel free-text answers into paywall copy
• Why fewer options in session one improved day one and day seven retention

Key takeaways:

Anticipation peaks before first use, not after. Users had just paid $300 for a device and were hopeful — more hopeful than after their first session, which feels like a mild shock. Waiting for the aha moment meant waiting for intent to decay.

Trial length moves trial starts, not trial conversion. A longer window lowers cancel-anxiety and gets more people to begin; conversion rate holds steady. Trials went 3 days to 7, then added 14 days as an exit-intent downsell at the same conversion rate.

Survey the people who say no. Fewer than 10% of people who closed the paywall cited price. The top reasons were not feeling ready and feeling they'd already paid via the hardware, which redirected the roadmap from discounting to trial duration.

Discounting isn't the only downsell. The winning offer was more time, not less money, priced identically to the original, converting at the same rate.

A single test can produce two opposite results. Beducated's first test looked ambiguous until Applica split it by campaign ID: Meta and organic-influencer traffic needed completely different plan structures. Cold traffic needed a cheap entry point; warm traffic didn't.

Plan length reflects how far ahead users can picture themselves. Removing Beducated's one-month plan for organic traffic didn't hurt conversion — rare. Trusting the influencer, users moved straight to three-month and annual plans.

Settled best practices are worth retesting. A CTA test Michal almost skipped — swapping "Subscribe now" for "Start learning" — lifted conversion 20-50%, a reminder that even settled wins deserve a rerun.

Your users already wrote your best copy. Alux asked users what they wanted to achieve and by when; Applica pulled every Mixpanel answer, split by converters vs. non-converters, and rewrote onboarding in converters' own words.

Focus beats choice in a first session. Alux's home screen offered too many paths; Applica cut it to one goal-based block until 50% completion. Day one and day seven retention both improved.

Links and resources Applica: https://applica.agency 
Applica case studies: https://applica.agency/case-studies 
Michal on LinkedIn: https://www.linkedin.com/in/michael-bardin-60b224291/
Pulsetto: https://pulsetto.tech 
Beducated: https://beducated.com 
Alux: https://alux.com 
Botsi: https://botsi.com

Timestamps
01:00 Intro and what Pulsetto actually is
03:30 The paywall was buried behind feature gates
04:30 Moving the paywall to the end of the onboarding quiz
07:00 Why the industry fetishizes the aha moment
08:00 The math nobody runs: if they never see it, they can't buy
09:00 The result: 9x ARR
09:30 Trial start rate was 3%, trial conversion was 80%
10:30 The exit-intent survey and what people actually said
11:45 Trial length moves starts, not conversion
14:00 RevenueCat Paywall Builder and getting independent of developers
15:00 The downsell that was more time instead of less money
17:30 When unique per-placement paywalls are worth building
20:00 How often to show a paywall to free users
22:30 Trial start rate goes from 3% to 16%
23:30 Beducated and the web-to-web funnel
24:30 +47% RPU on paid, +27% on organic
25:00 One test, two completely different results
27:00 Splitting the analysis by campaign ID
28:00 Removing the monthly plan and nothing happened
29:45 The weekly plan as the web version of a trial
32:30 The CTA test Michal did not want to run
35:00 Post-purchase upsells: PDF first, then Beducated Duo
37:00 Why stacking two upsells kills the second one
39:20 Alux, a $149 content app that could not lower prices
41:00 Mining Mixpanel for what users wrote in their own words
43:30 Converter language is the best marketing copy you have
45:00 Too many options in the first session
47:00 Anything completed beats no completion
50:45 +22% day one retention, +26% install to paid
51:00 Retention problems are activation problems
52:00 Wrap and credits

Jonathan Parra, founder of Tapas Growth, explains why app category predicts test results better than the app itself, how to sequence design, packaging, and price tests, and why the ugly paywall keeps winning.

Jonathan has designed close to 4,700 paywalls. He walks through the testing order he uses with clients, the five paywall placements every new app should ship before optimizing anything, and the exit questionnaire that replaced his old discount ladder. He also gets specific on numbers: a healthy app loses half its trial starts, win-back campaigns aimed at those cancelers convert at 5 to 6 percent, and removing a free plan can push conversion from 2 percent to 12 percent while gutting your traffic.

What you'll learn:
• Why app category, not app quality, is the first thing Jonathan looks at when predicting a test outcome
• How product polish and a clear ICP change the size of the win you can expect
• Why he turns down clients he doesn't think he can make money for
• How to decide between freemium and a hard paywall using your marginal cost per free user
• Why AI apps with real inference costs should start with a hard paywall and a 3 to 7 day trial
• How to gate the expensive part of your product and leave the cheap part free
• Why design tests come before packaging tests, and packaging before price
• How a design winner sets up a price increase that doubles ARPU
• What changed in his testing workflow now that LLMs can crunch the data
• How device signals like battery level and network type get used as demand scores

Key Takeaways:

• Marginal cost decides your monetization model. If a free user costs you nothing, keep them and monetize later. If every action fires an LLM call or streams video, a hard paywall with a short trial is the honest answer. The middle path is gating the expensive feature and leaving the cheap one open, like charging for photo-to-macros and giving away water logging.

• Design, then packaging, then price. A design winner can double conversion rate. Once you have it, raising price walks conversion back toward where it started while ARPU stays doubled. Price testing first just trades conversion for revenue with no ceiling raised.

• The ugly paywall wins and you have to accept it. Jonathan is a trained UX designer and says CRO is a different game entirely. Dense, loud, in-your-face layouts beat minimal ones often enough that he stopped arguing with the data, especially in the companionship and character AI space.

• Ask instead of guessing. His old exit flow was a fixed ladder: extended trial, then 33 percent off. It cannibalized revenue from people who would have paid more. Now an exit questionnaire asks why they bailed, and the offer matches the answer. Price complaint gets a discount. Trial complaint gets a longer trial.

• Half your trials cancel, and nobody markets to them. Jonathan targets users with an active entitlement and auto-renewal switched off. Those campaigns convert at 5 to 6 percent, which adds 2.5 to 3 points to overall conversion. It's the largest high-intent audience most apps ignore.

• Discount depth is a sequencing decision. Don't open with 80 percent off. Save the steep offers for expired users and Black Friday. A downgrade to a cheaper tier often keeps the customer without cheapening the brand, and a first-year-only discount lets you rebill at full price later.

• Weekly-only pricing is a speed run. ARPU looks great and churn is brutal. Jonathan will use weekly plans as paid intro offers or for genuinely short-use ICPs, but apps that sell nothing else ride viral traffic until the cohorts stop stacking.

Links & Resources
• Tapas Growth: https://tapasgrowth.com/
• Jonathan Parra on X: https://x.com/jondeparra
• Jonathan Parra on LinkedIn: https://www.linkedin.com/in/jondeparra/
• Jonathan's guest post on Retention.blog: https://www.retention.blog/p/expert-paywall-tips

Timestamps
00:00 Intro: 4,700 paywalls and counting
01:00 What Jonathan got wrong early at Superwall
03:30 Predicting test results before you run them
05:30  Using category benchmarks to diagnose an app
07:00 The two times he was wrong, and working for free
09:30 Freemium vs hard paywall, decided by cost
13:00 Gating the expensive feature, freeing the cheap one
14:00 Test order: design, packaging, price
17:30 Demand scores from device attributes
18:30 Age-based price testing and why it's risky
20:30 What changed post-AI in the testing workflow
23:30 Why the ugly paywall wins
27:30 Building a real exit flow
28:30 The questionnaire that replaced the discount ladder
31:00 The exact questions he asks
34:30 The five paywalls every new app should ship
38:30 Trial cancelers: the 5 to 6 percent win-back
40:30 Downgrades, discount depth, and brand
42:00 Transaction abandon tactics
44:00 Winning back expired subscribers
48:30 Email, push, SMS, and where the ceiling is
51:30 Weekly plans and the TikTok wall
54:00 Biggest packaging win: multi-page paywalls

Opal rebuilt their onboarding to work like a chat thread instead of a quiz, and Jacob walks through the whole thing screen by screen. The rock you crack open on the first tap, the sign-in question that replaced the sign-in buttons, the moment they tell you you'll spend eighteen years of your life looking at your phone, the paywall, and the monthly plan they only offer you if you try to leave.

This is a solo episode, so it goes deeper on the screens than a conversation usually allows. An app at Opal's scale doesn't ship an onboarding redesign without testing it hard first, which makes it a useful thing to study. The question isn't whether it works. It's which pieces of it would work for you.

What you'll learn:

• Why the first screen of your app is probably leaking more users than your paywall is
• How Opal replaced the Log in / Sign up wall with a question
• Why more login options usually pay for themselves once you scale
• The phrasing trick that gets people to answer a demographic question honestly
• Why Opal asks about your screen time before asking for screen time permissions
• How they split their permission requests apart, and what they put in between
• The three-beat setup: 91 days this year, 18 years of your life, then the rescue
• How paywall copy pays off a goal the user selected five screens earlier
• Why the trial reminder screen has almost nothing to do with reminders
• What "design your trial" is really doing to the user's decision
• The math behind pulling monthly off your first paywall
• The exit-intent monthly offer that almost nobody runs
• Why "no payment due now" keeps showing up next to the CTA
• What the Law of Shitty Clickthroughs says about the future of quiz onboarding
• The screen-count test for whether chat onboarding fits your app

Links & resources

• Retention.blog full written breakdown: https://retention.blog
• Opal: https://opalapp.com/
• Andrew Chen, "The Law of Shitty Clickthroughs": https://andrewchen.com/the-law-of-shitty-clickthroughs/
• Botsi: https://botsi.com

00:00 Intro: tired of quiz-style onboarding?
01:05 Opal's chat-style redesign
01:28 The rock you crack open
02:34 Why first-screen drop-off compounds
03:15 "Have we met before?" instead of Log in / Sign up
03:58 Login options and the data reassurance copy
04:22 The hybrid quiz/chat question style
05:31 "What best describes you?"
06:21 Easing into the screen time permission
07:44 91 days, 18 years, and the aha moment
09:07 Splitting the permission asks apart
10:00 The personalized pre-paywall screen
11:38 The fist bump commitment prompt
12:15 "Two plus hours" and a copy critique
13:23 Social proof and the "Reclaim my time" CTA
13:55 The trial reminder screen
15:47 Design your trial
16:13 The math on removing monthly
18:06 "Not ready for a year?"
18:27 The "no payment due now" checkbox
20:08 Post-paywall onboarding and gamification
22:15 The Law of Shitty Clickthroughs
23:27 Why chat UX works right now
23:54 Which apps should actually test this
25:19 Teaser: the onboarding and paywall library

Andrey Shakhtin, founder and CEO of FunnelFox, explains why web subscriptions convert and monetize better than the app store, how to stand up a minimum viable web-to-app test, and the payment risks that can freeze your revenue once you scale.

What you'll learn:

• Why full-funnel conversion (impression to purchase) runs roughly 2x higher on web than in-app, and how the app store install step explains the gap
• Why web LTV is about 2x in-app on annual plans and at least 50% higher on monthly
• The two structural reasons web LTV is higher: quiz funnels skew toward older, higher-willingness-to-pay buyers, and you control retention end to end
• How owning the payment stack lets you run custom cancellation flows and dunning (failed-payment recovery) that Apple and Google never expose
• Why deterministic post-ATT attribution makes web the fastest creative-testing loop you have
• Why a web funnel is the cheapest way to validate a product, sometimes before the app exists
• Why free trials quietly poison your Meta optimization, and how a $1-$5 paid trial fixes the signal
• The unit-economics benchmarks that matter: roughly 40-60% day-zero ROAS and a 6-month payback
• Why most "amazing-looking" funnels still fail at the paywall and checkout
• How to structure a paywall: outcome-based value, visualization, FOMO, price, then social proof
• The real difference between refunds, disputes, and chargebacks, and why only chargebacks threaten your account
•  How dispute-rate and VAMP (Visa's acquirer monitoring) thresholds can get your PSP to freeze recurring revenue
•  Why chargeback-alert services and external billing (payment orchestration) are close to mandatory at scale
•  How an easy refund path plus a 50% save offer lowers chargebacks and protects net revenue
•  How post-purchase upsells add roughly 20% LTV without triggering disputes

Key Takeaways:

A web funnel is the cheapest validation you have. Test demand, pricing, even a niche for a few hundred dollars a day before committing engineering. Some teams launch with no product at all, then build the app for the niche that converts.

Web's edge is deterministic measurement. With no ATT loss, every purchase ties to an exact creative, so you iterate on message and audience by real numbers instead of inferring from installs.

The full funnel converts about 2x on web. You skip the app store install decision, a friction point where lukewarm users drop before they ever see the offer. Most teams miss it because they optimize to installs.

Paid trials protect your ad signal. A $1-$5 charge proves the card works and sends Meta a real purchase event, so it optimizes toward payers instead of trial-tourists who never convert.

The paywall is where funnels die. After ~100 funnel reviews, it's the most under-built step. "Unlock all features" is not value. Lead with a specific outcome and date, and build it like a landing page.

Chargebacks can freeze everything. Refunds are harmless, but cross a provider's dispute or VAMP threshold and it can lock all stored recurring revenue. Make refunds easy and offer a 50% save to keep subscribers.

Links & Resources

FunnelFox: https://funnelfox.com
State of Web-to-App Subscriptions report: https://funnelfox.com/state-of-web2app-2026
Andrey Shakhtin on LinkedIn: https://www.linkedin.com/in/andrey-shakhtin/

Timestamps

00:00 Andrey's path: 16 years in mobile, from code to growth to FunnelFox
04:30 The minimum viable web-to-app experiment
08:00 What scale you need first, and validating a niche with no product
10:00 Why web should complement, not replace, your other channels
13:30 How web monetization differs from in-app
16:00 The report: web LTV and conversion roughly 2x in-app
16:40 Why web LTV is higher: older buyers and full retention control
19:00 Why the app store install step kills conversion
21:00 Organic vs. paid traffic funnels
22:30 Free trials vs. paid trials, and the Meta signal
25:30 Why web-to-app projects fail
28:30 Unit economics: day-zero ROAS and the 6-month payback
29:30 Diagnosing a broken funnel against benchmarks
36:00 How to structure a paywall that converts
41:30 Intro offers and the telecom playbook
44:30 Refunds, disputes, chargebacks, and VAMP explained
51:00 Defending your payment account: alerts and external billing
55:30 The refund hack: offer 50% instead of losing the subscriber
57:30 Upsells and lifting LTV ~20%
59:30 Biggest pricing win: GLP-1 and $800 order values
01:00:30 What FunnelFox does