Home / Paywalls / Video script
Download raw .md

Your Paywall's Real Enemy Isn't Price

The psychology of paywalls that convert by removing fear

Format: target 8–12 min | Audience: builders shipping subscription apps Core thesis (say it 3 times): Your paywall isn't fighting "is this worth $10." It's fighting "will this app trap me." The best paywalls sell the exit.

Every number is fact-checked against research/paywalls-research.md (confidence tiers inside) and our own Mobbin audit (mobbin/paywalls-mobbin-audit.md, n=63 paywall screens from top iOS apps, sampled August 2026). Margin notes marked ️ are not spoken.


[0:00 – 0:45] COLD OPEN · ~110w

ON SCREEN: A generic "Start Free Trial" button. A thumb hovers. Taps. Cut to iOS Settings → Subscriptions → Cancel.

More than half of the people who cancel a three-day free trial cancel it on day zero. The same day they started it.

They tap "start free trial" — and before your welcome email even lands, they've gone into settings and shut off the renewal.

That's not indecision. That's a defusal. Your trial is a bomb they've learned to disarm.

ON SCREEN: $86 → $219.

And here's why. Ask people what they spend on subscriptions each month, and in one survey of a thousand consumers the average guess was 86 dollars. Their itemized, actual average: 219.

Your paywall isn't fighting "is this worth ten dollars." It's fighting "will this app trap me."

Day-0 stat: RevenueCat State of Subscription Apps 2026, 115k apps — telemetry, cite on screen. $86/$219: C+R Research 2022, n=1,000, self-report — the words "in one survey" are load-bearing.


[0:45 – 1:20] ROADMAP · ~90w

ON SCREEN: Three numbered cards.

Three parts.

One — the price psychology that's actually backed by research: why "free" is not a price, why "$0.27 a day" works, and exactly where both of those break.

Two — free trials. The industry benchmarks and the only two randomized experiments ever run disagree — and knowing why will change how you read every stats report forever.

Three — we sampled sixty-three paywalls from top iOS apps ourselves. I'll show you what the best ones do differently, and then we'll rebuild a bad paywall into a good one, element by element.


[1:20 – 3:40] ACT ONE — "FREE" IS A CATEGORY, NOT A PRICE · ~330w

ON SCREEN: A Hershey's Kiss priced 1¢. A Lindt truffle priced 15¢.

Two chocolates on a table. A Hershey's Kiss for one cent. A Lindt truffle — objectively the better chocolate — for fifteen.

In this experiment, seventy-three percent took the truffle. Fourteen cents more for way better chocolate? Easy.

Now cut both prices by a single penny. Kiss: free. Truffle: fourteen cents.

Sixty-nine percent switch to the Kiss.

ON SCREEN: "FREE ≠ cheap. FREE = a different category."

One cent flipped the room. That's Shampanier, Mazar and Ariely, 2007, in Marketing Science. Their conclusion: zero isn't a low price. Zero is a different mental category — when something is free, your brain skips the cost-benefit math entirely, because there's no downside to calculate.

Which is why "7 days free" and "first week for 99 cents" are not 99 cents apart. They're in different categories. And why the phrase "$0.00 due today" is so potent — it's a true statement that keeps the whole transaction inside the free category.

ON SCREEN: "85¢/day → 52% · $300/year → 30%"

Second tool: temporal reframing. In 1998, John Gourville asked people to support a charity at "85 cents a day" — 52 percent said yes. Framed as "$300 a year" — the same money — 30 percent. The daily frame makes your brain fetch daily-sized comparisons: coffee, parking. The yearly frame fetches vacations and rent.

But here are the two catches nobody tells you.

Catch one: Gourville's own follow-up found the effect reverses at higher amounts. "$0.27 a day" reads like pennies. "$4.29 a day" reads like lunch, every day, forever — at that magnitude the yearly frame actually performs better.

Catch two: Apple requires the actual billed price and term to be the prominent element on your paywall. Per-day framing can decorate the price. It cannot legally be the price. Creators telling you to show only the per-day number are handing you a rejection.

One chocolate experiment variant used 26¢/27¢ in the paper; the 1¢/15¢/14¢ version with 73%/69% is verified in the brief. The "$1/day vs $350/year" stimulus circulating online is Atlas & Bartels 2018, not Gourville — don't blend them.


[3:40 – 5:50] ACT TWO — THE TRIAL-LENGTH SPLIT SCREEN · ~330w

ON SCREEN: Split screen. LEFT: "Benchmark report: longest trials convert best — 45.7%." RIGHT: "RCT, n=337,724: 7-day beat 30-day."

Okay. How long should your free trial be?

The benchmark data — RevenueCat, a hundred fifteen thousand apps — shows the longest trials converting best. Trials of seventeen to thirty-two days convert at nearly forty-six percent, the highest of any bucket. So: longer trial, better conversion. Right?

Here's the problem, and it's the most useful thing in this video. Apps choose their own trial length. Who picks a 30-day trial? Products so sticky they know you'll still be using them in week four. The long trials aren't causing the conversion — confident products are choosing long trials. That's a selection effect wearing a causation costume.

ON SCREEN: "The only two randomized experiments:"

Because when researchers actually randomized it, the answer flipped.

A major SaaS company let researchers randomly assign 337,724 users to a 7, 14, or 30-day trial. Published in Management Science. The 7-day trial won — 5.6 percent more subscriptions, nearly 8 percent more revenue, and better retention two years out. Fourteen days performed the same as thirty.

A second randomized experiment — 680,588 users at a freemium software product, 3-day versus 7-day — found 7 beat 3, with 21 percent higher overall conversion. So the evidence brackets it from both sides: seven beats thirty, and seven beats three.

When a benchmark and an experiment disagree, believe the experiment.

ON SCREEN: "46.5% of apps now run ≤4-day trials."

Meanwhile the industry is stampeding the other direction — nearly half of subscription apps now run trials of four days or less, even though those convert worst in the same benchmark reports everyone quotes. And those ultra-short trials are exactly the ones getting defused on day zero.

One wrinkle worth keeping: in that second experiment, users who finished their whole task during the trial were less likely to subscribe. If your product solves an occasional problem, a long trial lets people extract the value and leave. Match the trial to your product's natural usage cycle — not to a benchmark chart.

Yoganarasimhan et al., Mgmt Science 2023: +5.59% subs, +7.91% revenue, ~+6.4% retention. Zhang & Duan 2025, Frontiers in Psychology: +20.9% overall, +42.4% delayed conversion. Both SOLID. Do not name RevenueCat's numbers as causal claims.


[5:50 – 8:10] ACT THREE — THE TRANSPARENCY EQUILIBRIUM (Blinkist + our data) · ~340w

ON SCREEN: Blinkist's trial paywall with the Today / Day 5 / Day 7 timeline.

Now the real-world example, because one company saw all of this coming eight years ago.

Around 2018, Blinkist ran user research on why people wouldn't start their free trial. The number-one hesitation wasn't price. It was: "I'll forget to cancel and get charged." A third of their cancellations happened immediately after starting the trial — day-zero defusals, before anyone had a name for it.

So they rebuilt the paywall to attack the fear instead of the price objection. A timeline: today — full access. Day five — we send you a reminder. Day seven — first charge. That reminder email — the thing every growth marketer would call conversion suicide — became a selling point.

Blinkist reports trial starts went up twenty-three percent. Complaints dropped by half. Push-notification opt-in went from six percent to seventy-four — because the notification now had a job the user wanted done.

Self-reported internal test, no n or significance published — "Blinkist reports" is the required phrasing. Numbers: +23% trial signups, −55% complaints, 6%→74% opt-in, 4% higher trial retention.

ON SCREEN: "We sampled 63 paywalls (top iOS apps, Aug 2026)." Counts animate in.

So is transparency the norm now? We pulled sixty-three paywalls from top iOS apps and tallied them ourselves. Here's what the field actually looks like.

Ninety percent of multi-plan paywalls preselect a plan — the default does the choosing. Fifty-seven percent reframe the annual price per month or per week. But only one in eleven trial paywalls shows a Blinkist-style timeline. Discount anchors outnumber social proof two-and-a-half to one. Sixty-three percent make the close button deliberately subtle — and on six percent it's gone entirely.

And a detail I love: the most common price ending on these paywalls isn't .99 anymore. It's .98 — forty-eight percent versus forty-one. The over-learned trick got one cent weirder.

Read that list again, though. The tools top apps lean on are pressure tools — defaults, anchors, vanishing exits. The fear-removal tools that Blinkist validated are still rare. That's your opening.

All figures from our audit file; on screen cite as "n=63 paywall screens, top iOS apps via Mobbin, Aug 2026 — curated sample, not the app economy." Preselection: 35/39 multi-plan screens; timeline 3/32 trial screens = 9%; anchors 56% vs social proof 22%; close subtle 40/63, absent 4/63.


[8:10 – 10:50] ACT FOUR — THE TEARDOWN · ~400w

ON SCREEN: Both mockups side by side (interactive demo). A sleep app called "Drift." Stay locked here; highlight each element as named.

Alright — let's rebuild one. Fictional sleep app, Drift. The before screen commits only sins we actually found in the wild. Same product, same prices, both screens.

HIGHLIGHT: the plan cards

Plans. Before: three identical cards — weekly, monthly, yearly — none selected. We know from the forms video what a wall of unmade decisions does. And the "MOST POPULAR" badge is sitting on the monthly plan — which is the highest-margin plan for the app and the worst deal per month for the user. Users can do division. A badge that contradicts the visible math doesn't steer people — it teaches them the screen is lying.

After: two plans. Yearly preselected — like 54 percent of multi-plan paywalls we sampled. The badge sits where the math points.

HIGHLIGHT: the price line

The price. Before: "$59.98/year." One number, one frame — vacation-sized.

After: "$4.99 a month, billed annually — $59.98 a year." Per-month framing to shrink it — that's Gourville — with the true billed price right there, prominent, which keeps Apple happy and keeps trust intact. Almost nobody in our sample hides the real number anymore. The apps that got caught doing it aren't in the sample.

HIGHLIGHT: the countdown timer — then delete it

The timer. Before: "80% OFF — expires in 4:59." If the discount is real, say when it ends. If it resets every session, congratulations, you've installed a lie detector that always goes off. Deleted.

HIGHLIGHT: the trial timeline

In its place, the thing only nine percent of trial paywalls have: the timeline. Today, full access. Day five, we remind you. Day seven, billing starts. You are literally putting your cancellation window on a billboard — and that's the point. It's the answer to the number-one objection, printed on the screen.

HIGHLIGHT: the CTA

The button. Before: "CONTINUE" — continue to what? After: "Start my free week — $0.00 due today." True statement. Zero-price category. The brain math never leaves "free."

HIGHLIGHT: the close button

And the exit. Before: a ghost-grey x hiding in the corner. After: a visible one, plus "Not now" in plain text. A hidden exit tells users you expect to win by trapping them. A visible one says the product expects to win on merit. That signal is social proof.

ON SCREEN: Fear-added vs fear-removed counter: 6 → 0 / 0 → 6.

Count it up. The before screen adds fear in six places. The after removes it in six — and shows more true pricing information, not less.


[10:50 – 11:30] CLOSE · ~120w

ON SCREEN: The after screen. Hold.

Users didn't get cheaper. They got burned. They guess 86, they're paying 219, and they've learned to defuse trials on day zero in self-defense.

So here's the rule: the best paywalls sell the exit. Every element that makes leaving feel safe — the visible price, the reminder promise, the cancel-anytime that's actually true — is an element that makes starting feel safe.

Open your paywall tonight and count: how many elements add fear, how many remove it? If the timers outnumber the reassurances — now you know exactly what to fix.

[CTA / outro]

---

APPENDIX A — Optional expansion beats

A1. Hard vs soft paywalls: the category-vs-store trap (+75 sec) — insert after Act Two

The "hard paywalls convert 5x better" stat (10.7% vs 2.1% download-to-paid at day 35, RevenueCat) is real telemetry and a real category error — access models are self-selected, and free-user value goes unmeasured. The measured version of the cost comes from news: when publishers field-tested paywalls, visits fell 51%, with young readers nearly vanishing (Chiou & Tucker 2013); the NYT's metered paywall cut heavy-user engagement 57% while lifting print+digital revenue (Pattabhiramaiah et al. 2019). Paywall conversion is the category metric; audience formation is the store metric — same trap as Borle et al. in the forms video. ️ This is the strongest cut material — restore first if running long.

A2. The weekly-preselect counter-cluster (+30 sec) — insert in Act Three after the preselection stat

A cluster in our sample (Liven, Moonly, Lovi, TIDE) preselects the expensive weekly plan while dangling yearly as "best offer" — betting on subscribers who don't do the math staying subscribed at $5/week. It's the exact inverse of the annual-default wisdom, and a live example of a default chosen for the metric, not the user (forms video, §ethics).

A3. Reverse trials & the endowment effect: the honest take (+40 sec) — insert at end of Act Two

"Give features then take them away — loss aversion!" has zero direct peer-reviewed tests in software. The endowment effect is real (coffee mugs, 1990), but the closest real evidence — the 680k-user experiment — found satiation can dominate: users who finish their task in-trial subscribe less. Frame reverse trials as plausible and untested; nobody else will.


APPENDIX B — Title, thumbnail, chapters

Titles 1. Your Paywall's Real Enemy Isn't Price ← strongest: names the reframe 2. Why Users Cancel Your Trial in the First Hour (55% Do) 3. I Sampled 63 Paywalls From Top Apps — The Best Ones Sell the Exit 4. The Free Trial Science Every App Gets Backwards

Thumbnail: Split. LEFT: paywall with countdown timer + tiny x, red tint, "TRAP". RIGHT: timeline paywall (Today→Day 5→Day 7), green tint, "SAFE". Or: a big "$0.00" with a lit bomb-fuse being snipped.

Chapters

0:00  Users defuse your trial on day zero
0:45  What we're covering
1:20  One cent flipped the room ("free" is a category)
2:40  $0.27/day: the paper, the boundary, the Apple ceiling
3:40  Trial length: benchmarks vs the only two real experiments
5:50  Blinkist put the cancel window on a billboard
7:00  We sampled 63 paywalls — what top apps actually do
8:10  Teardown: rebuilding Drift's paywall
10:50 The rule: sell the exit

APPENDIX C — Ranked cut list

Baseline ~11:00 as written (≈1,650 spoken words @150wpm). Work down until at target.

# Cut Saves Cost
1 The .98-vs-.99 detail (Act Three) 0:15 A charming but nonessential flourish. Keep if possible — it's very tweetable.
2 Anderson & Simester teaser is already absent from the main spine — nothing to cut
3 The satiation wrinkle (end of Act Two) 0:25 Loses the "match trial to usage cycle" advice — the most actionable line in Act Two.
4 Blinkist's push-notification number (Act Three) 0:10 Small; keep the +23% and −55%.
5 The close-button beat (Act Four) 0:25 Loses the "exit as signal" idea that the Close depends on — if cut, also soften the Close.
6 Catch two / Apple ceiling (Act One) 0:20 Loses the compliance payoff; the per-day beat still works.

Below ~9:00 you're cutting teaching. If you need 8:00, split: Part 1 = psychology + trials (Acts 1–2), Part 2 = Blinkist + data + teardown (Acts 3–4).


APPENDIX D — Production notes

Every number, with its source

Claim Source Tier
55% of 3-day-trial cancellations on Day 0 RevenueCat State of Subscription Apps 2026 (115k apps) SOLID telemetry — cite year on screen
$86 guessed vs $219 actual; 42% forgot ≥1 sub C+R Research 2022, n=1,000, self-report CONTESTED — must say "one survey"
73% truffle → 69% Kiss on a 1¢ change Shampanier, Mazar & Ariely 2007, Marketing Science 26(6) SOLID
52% vs 30% (85¢/day vs $300/yr) Gourville 1998, JCR 24(4) SOLID
Per-day reverses at high magnitude Gourville 2003, Marketing Letters 14(2) SOLID
Apple price-prominence rule App Review Guideline 3.1.2 + documented rejection patterns SOLID (the "16pt" folklore is NOT stated)
17–32 day trials 45.7% t2p RevenueCat 2025 SOLID as description, never causal
RCT #1: 7-day +5.59% subs, +7.91% rev, better 2-yr retention, n=337,724 Yoganarasimhan et al., Management Science 2023 SOLID
RCT #2: 7 vs 3-day, +20.9% overall, +42.4% delayed, n=680,588; satiation effect Zhang & Duan 2025, Frontiers in Psychology 16:1568868 SOLID
46.5% of apps ≤4-day trials; ≤4-day converts worst RevenueCat 2026 SOLID telemetry
Blinkist: fear = #1 hesitation; 33% cancel at trial start; +23% trial signups; −55% complaints; 6%→74% opt-in Blinkist via Growth.Design + Purchasely interview CONTESTED — "Blinkist reports" phrasing mandatory
Audit: 90% preselect; 54% annual; 57% reframe; 9% timeline; 56% anchors vs 22% social proof; 63% subtle close, 6% absent; .98 48% vs .99 41%; 36% of special-offer screens run countdowns Our Mobbin audit, n=63, Aug 2026 Original data — state sample honestly

Delivery notes - The Act Two "selection effect wearing a causation costume" beat is the video's signature credibility moment — drop graphics, talk to camera. - Say "Blinkist reports" and "in one survey" exactly — those phrasings carry the honesty. - The teardown's timer-deletion is the emotional peak; give it a beat of silence after "always goes off." - Never say: Economist decoy story, "longer trials convert better," "hard paywalls convert 5x so go hard," "endowment effect proves reverse trials," any unattributed "$133/month wasted." Full blacklist in the research brief.

Demo spec — see demo-spec.md in this folder. The interactive before/after is the on-screen artifact for Act Four; walkthrough order there matches the highlight order here.