# Research Brief — Mobile App Paywall Design
### Fact-check reference for the YouTube script

**Confidence key**
`SOLID` — primary source located, figure verified
`CONTESTED` — real source exists, but methodology or interpretation is disputed
`SHAKY` — widely repeated, source is weak, absent, or circular
`BLACKLISTED` — never repeat this claim

**Cross-reference note:** choice overload, defaults psychology (Madrian & Shea, Johnson & Goldstein), Hick's Law, 7±2, social login, and form-field research are already covered in `decision-fatigue-research-brief.md`. This brief does not rehash them; it points to them where the paywall application depends on that spine (plan defaults → that brief's §5; number of plans → its §1 Chernev moderators).

---

# ⚠️ PART 1 — CORRECTIONS

Six popular claims in paywall content that are wrong, mis-told, or missing their caveats.

### 1. "Hard paywalls convert 5x better — so use a hard paywall"

The 5x number is real but the advice is a category error. RevenueCat's State of Subscription Apps 2026 (115,000+ apps, $16B tracked revenue, 1B+ transactions): median download-to-paid at day 35 is **10.7% for hard-paywall apps vs 2.1% for freemium** — and coverage of the report adds an ~8x revenue-per-install gap at day 60. `SOLID` as telemetry — https://www.revenuecat.com/state-of-subscription-apps

**But this is correlational, not causal.** Apps don't randomize their access model. Hard-paywall apps self-select: they tend to be utility/health apps with instant clear value; freemium apps are chasing scale, network effects, and word of mouth. RevenueCat itself frames freemium as "the right call when free users drive word of mouth, network effects, or long-term brand scale," and notes year-one retention for the two models is nearly identical. **No randomized test of hard vs soft paywall at the same app has ever been published.** The honest statement: hard paywalls monetize a smaller funnel harder. Whether that's more *total* revenue depends on what free users are worth — which the metric doesn't measure. (Same category-vs-store trap as Borle et al. 2005 in the decision-fatigue brief §8.)

---

### 2. The Economist "decoy pricing" story — The Economist never ran it

The most repeated pricing story on the internet: web-only $59, print-only $125, print+web $125; the "useless" middle option supposedly shifted everyone to the $125 bundle and boosted revenue 43%.

| What's claimed | Reality |
|---|---|
| "The Economist's famous pricing experiment" | **The Economist ran no experiment.** Dan Ariely saw the ad and ran a **hypothetical classroom survey on 100 MIT students** (*Predictably Irrational*, 2008). No purchases, no revenue, n=100 students. |
| "The decoy effect is settled science" | The lab effect (Huber, Payne & Puto 1982) is real **with abstract numeric stimuli**. Frederick, Lee & Baskin, "The Limits of Attraction," *JMR* 51(4), 2014, found the effect **largely disappears with realistic products** — they "failed to replicate several of the results most frequently cited as evidence." Yang & Lynn (2014, same JMR issue) reported the same. | 

`SOLID` (that the correction is right) — https://journals.sagepub.com/doi/10.1509/jmr.12.0061
**On camera:** "add a decoy plan" is advice built on a classroom survey plus a lab effect that dies on contact with real products. If a decoy tier works on your paywall, that's your A/B test's finding, not psychology's.

---

### 3. "Longer trials convert better" — the telemetry and the RCTs point opposite ways

RevenueCat 2025 reported the *highest* trial-to-paid conversion for the *longest* trials (17–32 days: 45.7%). Creator content quotes this as "make your trial longer." But trial length in that dataset is **chosen by the app**, not randomized — confident apps with sticky products pick long trials.

The only randomized evidence says the opposite for acquisition:

- **Yoganarasimhan, Hema et al., "Design and Evaluation of Optimal Free Trials," *Management Science* (2023; working paper 2020).** Major anonymous SaaS firm, **n = 337,724 users randomized to 7 / 14 / 30-day trials**, outcomes tracked ~2 years. **7-day trial was the best uniform policy: +5.59% subscriptions**, ~+7.9% revenue, and better retention (~+6.4%) vs the 30-day baseline. 14-day was statistically indistinguishable from 30-day. Personalized assignment beat uniform only modestly (≈6.8% vs 5.6% lift) and was fragile. `SOLID` — https://arxiv.org/abs/2006.13420 · https://pubsonline.informs.org/doi/10.1287/mnsc.2022.4507
- **Zhang & Duan (2025), *Frontiers in Psychology* 16:1568868.** Image-editing freemium SaaS, **n = 680,588 users, 190 countries, 3-day vs 7-day randomized**, 2-year window. 7-day: trial adoption +11.1%; immediate conversion +7.3% (n.s.); **delayed conversion +42.4%; overall conversion +20.9%**. Mechanism: longer trials increased premium-feature engagement, but users who *finished their task* during the trial were less likely to subscribe (satiation/demand cannibalization). `SOLID` — https://pmc.ncbi.nlm.nih.gov/articles/PMC12217587/

**Reconciliation for the script:** short trials force the subscribe decision while motivation is hot (and both RCTs are consistent with 7 > 30); the telemetry's "long trials convert great" is survivorship — apps that can afford a 30-day trial already have retention. Saying "the data says longer trials convert better" cites a selection effect as a causal one.

---

### 4. The Blinkist "transparent paywall" story — real, but get the provenance and the metric right

The famous case: Blinkist's paywall told trial users exactly when they'd be charged and promised a reminder before trial end.

| Often said | Reality |
|---|---|
| "Blinkist increased conversions 23%" | It was **+23% trial sign-ups** (paywall → trial start), not trial-to-paid. Self-reported **4% higher trial retention** is the only downstream number. |
| "A famous study" | It's a **self-reported internal A/B test**, first written up by Blinkist designer **Jaycee Day** ("Solving Trial Conversions," UX Planet, 2021), popularized by Growth.Design's teaching case and Purchasely's interview with Blinkist's Eveline Moczko. **No sample size, duration, or significance ever published.** |
| Other numbers | Push-notification opt-in **6% → 74%**; customer complaints **−55%**; user research found "forgetting to cancel and being charged" was the #1 hesitation, and 33% of cancellations happened immediately after trial start. |

`CONTESTED` — credible, consistent across three tellings, but vendor-published with no methodology. Say "Blinkist reports," not "a study found." — https://growth.design/case-studies/trial-paywall-challenge · https://www.purchasely.com/blog/using-transparency-to-increase-your-conversion-rate-with-eveline-moczko-blinkist

---

### 5. "Freemium converts at 2–5%" stated as a law

The range circulates with rotating attribution (sometimes pinned on Kumar's HBR piece, which is a strategy article, not a benchmark study). What's actually sourced:

| Figure | Source | Confidence |
|---|---|---|
| Mobile freemium download-to-paid (D35) median: **2.1%**; overall all-model median 2.0% | RevenueCat 2026, 115k apps | `SOLID` (vendor telemetry, huge n) |
| Dropbox ~**4%** of free users paying (called exceptional) | outside estimates, never a filing | `CONTESTED` |
| Evernote conversion by tenure: **1.5% month 1 → 7% year 1 → 11% year 2** | Phil Libin conference remarks, ~2011 | `CONTESTED` — self-reported, 15 years old |
| Kumar, "Making Freemium Work," *HBR* May 2014 | framework: free tier too generous → no one upgrades; too stingy → no acquisition; treat the free/premium line as a dial you re-tune, not a launch decision | `SOLID` as attributed argument, contains no benchmark table — https://hbr.org/2014/05/making-freemium-work |

**On camera:** give the 2.1% median with the source and year; don't say "industry standard 2–5%."

---

### 6. "$0.27/day framing" — the research is real, the boundary condition never gets mentioned

The pennies-a-day tactic has an actual paper behind it (see §1 below) — but Gourville's own follow-up shows the effect **reverses at higher magnitudes**: "$1/day" beats "$365/year," but when the daily number gets big enough to stop resembling petty cash, the **aggregate frame wins** (Gourville 2003, *Marketing Letters* 14(2)). A $95.88/year app shown as "$0.26/day" is inside the effect; a $30/week app shown as "$4.29/day" is arguably outside it — $4.29/day retrieves comparisons like lunch, not pennies. `SOLID` — https://link.springer.com/article/10.1023/A:1025467002310

Also: Apple's review practice requires the **actual billed price and term to be the prominent element** — per-day framing can decorate the paywall but cannot legally *be* the price display (see §9). Creator advice that says "show only the per-day price" is a rejection waiting to happen.

---

# PART 2 — THE VERIFIED SPINE

## §1 — Temporal price reframing (the "$0.27/day" literature)

**Gourville (1998), "Pennies-a-Day: The Effect of Temporal Reframing on Transaction Evaluation," *JCR* 24(4): 395–408.** Framing a charitable donation as **"85 cents per day" got 52% compliance vs 30%** for the economically identical "$300 per year" — a 22-point swing from wording alone. Mechanism (two-step): the daily frame triggers retrieval of trivial daily expenses (coffee, parking) as the comparison set; the aggregate frame retrieves large infrequent expenses. `SOLID` — https://academic.oup.com/jcr/article-abstract/24/4/395/1797969

**Boundary (Gourville 2003, *Marketing Letters*):** effect reverses at high magnitudes — see Correction 6. `SOLID`

**Provenance beat:** the tactic predates the paper — Time magazine and competitors field-tested per-issue vs per-year framing in the early 1980s, with per-issue offers reportedly 10–40% more effective (recounted by Dholakia, *Psychology Today*, 2019, citing industry tests). `CONTESTED` — industry lore, but good for showing "$0.27/day" is a 40-year-old magazine tactic. — https://www.psychologytoday.com/us/blog/the-science-behind-behavior/201904/the-powerful-influence-pennies-day-price-offers

**Atlas & Bartels (2018), "Periodic Pricing and Perceived Contract Benefits," *JCR* 45(2).** Peer-reviewed follow-up: periodic (per-day) framing doesn't just shrink perceived cost — it shifts attention toward the **ongoing stream of benefits**, increasing willingness to maintain subscriptions. The modern academic footing for per-day framing on subscription products specifically. `SOLID` — https://home.uchicago.edu/bartels/papers/Atlas-Bartels-2018-JCR.pdf

**Vendor cases (label clearly):** Mailchimp reportedly +21% annual commitment from "$29.99/month billed annually" vs "$359.88/year"; Mojo (via RevenueCat's pricing-psychology writeup) +45% new revenue per paywall impression from monthly-equivalent framing of annual billing. `CONTESTED` — single-company tests, no methodology published. — https://www.revenuecat.com/blog/growth/subscription-pricing-psychology-how-to-influence-purchasing-decisions

**Price endings — the solid one:** Anderson & Simester (2003), "Effects of $9 Price Endings on Retail Sales," *QME* 1(1): 93–110. Three field experiments (real catalogs, real purchases): **$9 endings raised demand in all three**, famously including demand *rising* when a dress was repriced from $34 → $39, strongest for new items, weaker when "Sale" cues were present. The rare pricing-psychology claim with genuine field-experimental backing — and why $9.99/$29.99 endings on paywalls aren't just cargo cult. `SOLID` — https://link.springer.com/article/10.1023/A:1023581927405

## §2 — The zero-price effect: why "free trial" is categorically different from "$0.99"

**Shampanier, Mazar & Ariely (2007), "Zero as a Special Price: The True Value of Free Products," *Marketing Science* 26(6): 742–757.** The clean pair of conditions: Hershey's Kiss 1¢ vs Lindt truffle 15¢ → **73% chose the truffle**. Drop both prices by one cent — Kiss free, truffle 14¢ → **69% chose the Kiss**. A one-cent change flipped the majority. Free isn't a low price; it's a different category — zero-price options get an affective benefit boost, not just a cost reduction, and the effect survives real (non-hypothetical) purchase conditions in the paper. `SOLID` — https://pubsonline.informs.org/doi/10.1287/mksc.1060.0254 · PDF: https://web.mit.edu/ariely/www/MIT/Papers/zero.pdf

**Paywall translation:** "Try 7 days free" vs "First week $0.99" is not a $0.99 difference — it's a category switch. This is the legitimate science behind trial-first paywall CTAs and behind why "$0.00 due today" phrasing (Blinkist-style) is potent: it keeps the transaction in the zero-price category while stating the truth.

## §3 — Trial design: the causal evidence

(Full numbers in Correction 3.) The stack, in one table:

| Study | Design | n | Headline result | Confidence |
|---|---|---|---|---|
| Yoganarasimhan et al., *Mgmt Science* 2023 | RCT 7/14/30-day, SaaS | 337,724 | 7-day best: **+5.59% subs, +7.91% revenue**, better 2-yr retention; 14 ≈ 30 | `SOLID` |
| Zhang & Duan, *Frontiers in Psych* 2025 | RCT 3 vs 7-day, freemium SaaS | 680,588 | 7-day: **+20.9% overall conversion**, driven by **+42.4% delayed** conversion; satiation effect if users finish their task in-trial | `SOLID` |
| RevenueCat 2026 | telemetry | 115k apps | 5–9 day trials convert 37.4% vs 25.5% for ≤4 days; **46.5% of apps now use ≤4-day trials** (share rising) | `SOLID` as description, `BLACKLISTED` as causal claim |

**The uncomfortable detail worth airing:** the industry is migrating *toward* ultra-short trials (≤4 days now the plurality) even though those show the *lowest* median conversion in the same dataset — and **"55% of all 3-day trial cancellations happen on Day 0"** (RevenueCat 2026): users start the trial and immediately cancel to avoid the forgotten-charge trap. That behavior is the Blinkist research finding showing up in telemetry, eight years later.

**Trial start timing:** across categories, ~78–90% of trials start on Day 0 (Business 89.9%, Health & Fitness 82.1%, Productivity 78.0%). The paywall decision effectively happens at install. `SOLID` (RevenueCat 2026)

## §4 — Hard vs soft paywalls, and what hardening does to audiences

**Mobile telemetry** — Correction 1 numbers (10.7% vs 2.1% D35; retention converges by year one). `SOLID` as telemetry.

**The news-industry natural experiments are the closest thing to causal evidence on hardening a paywall:**

- **Chiou & Tucker (2013), "Paywalls and the demand for news," *Information Economics and Policy* 25(2): 61–69.** A media publisher field-tested paywalls at local news sites; comScore panel data. **Visits fell 51%** after the paywall, with a **far larger collapse among 18–24-year-olds** (widely summarized as near-total loss of that cohort). `SOLID` — https://www.sciencedirect.com/science/article/abs/pii/S0167624513000097 · PDF: https://www.oxy.edu/sites/default/files/assets/Economics/Chiou/chiou_and_tucker_paywalls.pdf
- **Pattabhiramaiah, Sriram & Manchanda (2019), "Paywalls: Monetizing Online Content," *Journal of Marketing* 83(2): 19–36.** NYT metered paywall (March 2011): **unique visitors −16.8%**; **heavy users −57.2%** vs light users −11.3% — the meter, by design, taxes your most engaged readers. Offsetting spillover: **+1–4% print circulation** (driven by print+digital bundling). `SOLID` — https://journals.sagepub.com/doi/10.1177/0022242918815163 · summary: https://www.ama.org/2019/03/07/before-you-put-up-a-paywall-read-this-study/

**App translation (novel-angle material):** paywall conversion is the *category* metric; installs, engagement, word-of-mouth are the *store* metric (see decision-fatigue brief §8, Borle et al.). Hardening reliably improves the first and taxes the second, and the news literature actually measured the tax: −51% traffic, youngest users first, heaviest users hit hardest.

## §5 — Plan architecture, defaults, and badges

**What paywalls actually look like (RevenueCat 2026, `SOLID` as telemetry):**
- **Two plans is the most common configuration** (41–60% of paywalls by category); 3+ plans is a minority (6–27%). The Chernev moderators (decision-fatigue brief §1) predict exactly this: uncertain preferences + time pressure → small sets with a dominant option.
- Subscription revenue mix: **monthly 42% · yearly 34% · weekly 20%** — the "annual is where all the money is" line is false for mobile. Weekly dominates Gaming (82%); annual dominates Health & Fitness (68%); Productivity is 77% monthly.
- Most common price points: **$5/week, $10/month, $30–40/year** (NA: $9.99/mo, $39.99/yr).

**Preselected plan defaults:** vendor guidance (Adapty, Superwall, RevenueCat) converges on preselecting the longest plan; circulating vendor anecdotes claim plan-mix inversions (e.g., ~37% → ~63% yearly selection; "82% of non-closing viewers kept the default"). `CONTESTED` — no published methodology for any of these numbers; the *mechanism* is the best-replicated finding in the adjacent brief (defaults read as advice — Madrian & Shea, §5 there). Use the peer-reviewed defaults literature for the claim, vendor anecdotes as illustration only.

**"Most popular" badges:** no peer-reviewed test on paywalls exists. The nearest systematic evidence cuts the other way: DoWhatWorks' review of ~15 large-company A/B tests found social-proof additions on plan-selection pages **tended to lose** (Dropbox logos, Glofox testimonials among the losers). `CONTESTED` — https://www.dowhatworks.io/blog/should-you-add-social-proof-on-pricing-pages. A "most popular" badge is better understood as a default/dominant-option marker (Chernev: a dominant option dissolves overload) than as social proof — which also predicts when it fails: when it contradicts the visible price logic.

## §6 — Endowment, loss aversion, and the "reverse trial"

**The honest state:** the claim "having-then-losing features converts better than never-having (endowment effect)" has **no direct peer-reviewed test in software**. The endowment effect itself is real (Kahneman, Knetsch & Thaler 1990 — mugs, WTA ≈ 2×WTP), but every "free trials convert 2–5x better than freemium because loss aversion" article is blog arithmetic transplanting a coffee-mug lab result onto SaaS funnels. `SHAKY` as a mechanism claim for paywalls.

**What actually exists:**
- Zhang & Duan (2025): more premium-feature *engagement* during trial → higher conversion, **but** users who completed their job-to-be-done in-trial converted *less* (satiation beats endowment when the need is episodic). `SOLID` — the closest peer-reviewed look at the mechanism, and it complicates the loss-aversion story rather than confirming it.
- Trial vs no-trial telemetry gaps (RevenueCat) are confounded by everything in Correction 1.

**On camera:** frame reverse trials as "plausible, loss-aversion-flavored, untested" — an honesty beat no other creator will deliver.

## §7 — What subscribers don't know they're paying (the transparency case)

- **C+R Research (May 2022, n=1,000 US consumers, self-report):** asked to guess their monthly subscription spend, average guess **$86**; their itemized actual average **$219** — ~2.5x underestimate. **42%** said they'd forgotten they were paying for at least one unused subscription. `CONTESTED` — market-research vendor, small n, self-report; but directionally corroborated by banking-data studies. — https://www.crresearch.com/blog/subscription-service-statistics-and-costs/ · https://www.cnbc.com/2022/06/02/consumers-spend-133-more-monthly-on-subscriptions-than-they-realize.html
- This is the demand-side context for Blinkist's finding (forgotten-charge fear = #1 trial hesitation) and RevenueCat's Day-0 cancellation stat. Three independent source types — user research, telemetry, consumer survey — triangulate the same anxiety.

## §8 — The rules of the game (Apple) and the regulatory whiplash

**Apple App Review Guideline 3.1.2 + review practice** (`SOLID` for the guideline text; `CONTESTED` where noted):
- Auto-renewable subscriptions must deliver ongoing value; minimum period 7 days, available across the user's devices.
- Paywalls must show the **actual billed price and period prominently**; rejections for insufficient price prominence are routine (the specific "use ≥16pt" advice is developer-community lore, **not** written Apple policy — label it as such).
- A "free trial" label must be backed by a real StoreKit introductory offer (duration, price, and what's lost at trial end disclosed) — not a home-rolled timer.
- Terms of Use and Privacy Policy links must be in the app's paywall UI; a **Restore Purchases** path is expected.
- **Hard paywalls are allowed.** "Apple rejects hard paywalls" is a myth; what gets rejected is hiding price, fake trials, and missing links.
— https://developer.apple.com/app-store/review/guidelines/ · rejection patterns: https://www.revenuecat.com/blog/growth/the-ultimate-guide-to-app-store-rejections

**FTC "click-to-cancel":** the Negative Option Rule (finalized Oct 2024 — separate consent for auto-renewal, cancellation as easy as signup) was **vacated in its entirety by the Eighth Circuit on July 8, 2025** — six days before its compliance deadline — on procedural grounds (FTC skipped the required preliminary regulatory analysis under FTC Act §22). ROSCA and state auto-renewal laws (notably California's) still apply, and the FTC has moved to restart rulemaking. `SOLID` — https://www.cooley.com/news/insight/2025/2025-07-11-click-to-cancel-just-got-cancelled-eighth-circuit-vacates-entirety-of-ftcs-negative-option-rule — check status again before publication; this is live.

---

# PART 3 — HOOK CANDIDATES

1. **"One cent flipped the room."** At 1¢ vs 15¢, 73% chose the Lindt truffle. Cut both prices by a single cent — free vs 14¢ — and 69% chose the Hershey's Kiss instead. (Shampanier, Mazar & Ariely 2007, *Marketing Science*.) `SOLID`
2. **"52% vs 30% — for the same $300."** Ask for "85 cents a day" and half say yes; ask for "$300 a year" and less than a third do. (Gourville 1998, *JCR*.) `SOLID`
3. **"337,724 people were randomized into a trial-length experiment. The 7-day trial beat the 30-day — on subscriptions, revenue, AND two-year retention."** (+5.59% subs, +7.91% revenue; Yoganarasimhan et al., *Management Science*.) `SOLID`
4. **"People guess they spend $86 a month on subscriptions. The real number averages $219."** (C+R Research 2022 — label as a consumer survey, n=1,000.) `CONTESTED` but honest if labeled.
5. **"The news industry hardened its paywalls and measured what apps won't: traffic fell 51%, and the youngest readers almost entirely vanished."** (Chiou & Tucker 2013.) `SOLID`
6. **"55% of 3-day-trial cancellations happen the day the trial starts."** Users have learned to defuse your trial before it can charge them. (RevenueCat 2026, 115k apps.) `SOLID` (telemetry)

---

# PART 4 — NOVEL-ANGLE CANDIDATES

1. **The RCT vs telemetry split-screen.** Every creator quotes RevenueCat ("longer trials convert better"); the only two large randomized experiments ever run say shorter wins for acquisition. Teaching viewers *why* both are true (endogeneity/survivorship) is a signature-move segment — and no paywall video on YouTube currently makes it.
2. **The Day-0 cancellation arms race.** Blinkist's 2018-era user research (fear of forgotten charges), RevenueCat's 2026 telemetry (55% of 3-day-trial cancels on Day 0), and C+R's $86-vs-$219 gap are the same phenomenon measured three ways across eight years: users now treat trials as adversarial. Transparency isn't ethics theater — it's the equilibrium response. Blinkist got there early and the numbers (even self-reported) moved in every direction you'd predict.
3. **Per-day framing is now a *regulated* design pattern.** Gourville gives the psychology, Gourville 2003 gives the boundary where it backfires, and Apple's price-prominence enforcement legally caps how far you can push it. "The $0.27/day trick has a 1998 paper, a failure condition, and a compliance ceiling" is a three-act beat nobody else has assembled.
4. **The category-vs-store metric transplant.** Borle et al. (decision-fatigue brief §8) showed cutting assortment improved every category and hurt the store. Paywall hardening is the same trap: conversion (category) up, audience formation (store) down — and the news-industry literature (−51% visits, −57% heavy-user engagement at NYT) is the measured version of the cost mobile apps never see in their dashboards.
5. **"Free" is a category, not a price** — using zero-price theory to explain why trial-first CTAs, "$0.00 due today" microcopy, and the trial/no-trial paywall split behave so differently from a 99-cent intro offer, then landing the honesty beat that the endowment-effect justification for reverse trials remains untested.

---

# PART 5 — THE BLACKLIST

Never say these on camera:

| Claim | Why |
|---|---|
| "The Economist tested decoy pricing" / "the decoy raised revenue 43%" | Ariely's hypothetical classroom survey, n=100 MIT students. The Economist ran nothing. Decoy effect fails to replicate with real products (Frederick, Lee & Baskin, *JMR* 2014). |
| "Hard paywalls convert 5x better, so go hard" | Correlational telemetry; access models are self-selected; year-1 retention converges; free-user value unmeasured. |
| "Longer free trials convert better" (as causal) | Selection effect in vendor data. Both large RCTs found 7-day beats 30-day. |
| "Studies show multi-page paywalls convert 37% better" | Superwall telemetry (40M opens, Feb–May 2026) is real but correlational; multi-page was 24% of opens, zero-transaction paywalls excluded, no app controls. "Superwall's data shows" is fine; "studies show" is not. |
| "Blinkist increased conversions 23%" (unqualified) | It was trial *sign-ups*, self-reported, no n or significance ever published. |
| "Freemium converts at 2–5%, that's the standard" | Rotating attribution, no primary benchmark. Use RevenueCat's 2.1% D35 median with the year. |
| "Loss aversion means reverse trials convert 2–5x better" | No study exists. Blog arithmetic on top of a coffee-mug experiment. |
| "The endowment effect is why free trials work" (as fact) | Closest peer-reviewed evidence (Zhang & Duan 2025) shows satiation can dominate — mechanism unsettled. |
| "Apple requires 16pt font for prices" | Community folklore. The written rule is prominence/legibility; 16pt is a workaround heuristic. |
| "Apple bans hard paywalls" | False. Compliant hard paywalls pass review daily. |
| "Delaying the close button boosts revenue by X%" | No published test anywhere — vendor blogs recommend the tactic with zero data, and refund/churn/review costs are unmeasured. Also squarely dark-pattern territory. |
| "Consumers waste $133/month" (unattributed) | Real number, but it's one 1,000-person self-report survey (C+R 2022). Attribute or drop. |
| "Annual plans are where all mobile revenue is" | RevenueCat 2026: monthly 42% > yearly 34% > weekly 20%; weekly is 82% of Gaming. |
| "Click-to-cancel is the law now" | Vacated by the Eighth Circuit, July 8, 2025. ROSCA/state law still bind — but the FTC rule is (currently) dead. Verify again pre-publish. |
| Any "average paywall converts at X%" without placement, trial, and category qualifiers | Adapty/Superwall/RevenueCat figures vary 5–10x by placement and trial config; a single number is meaningless. |

---

# PART 6 — PRIMARY SOURCES

**Pricing psychology**
Gourville (1998), Pennies-a-Day, *JCR* 24(4) — https://academic.oup.com/jcr/article-abstract/24/4/395/1797969
Gourville (2003), magnitude boundary, *Marketing Letters* 14(2) — https://link.springer.com/article/10.1023/A:1025467002310
Atlas & Bartels (2018), Periodic Pricing, *JCR* 45(2) — https://home.uchicago.edu/bartels/papers/Atlas-Bartels-2018-JCR.pdf
Shampanier, Mazar & Ariely (2007), *Marketing Science* 26(6) — https://pubsonline.informs.org/doi/10.1287/mksc.1060.0254 · PDF https://web.mit.edu/ariely/www/MIT/Papers/zero.pdf
Anderson & Simester (2003), $9 endings, *QME* 1(1) — https://link.springer.com/article/10.1023/A:1023581927405
Frederick, Lee & Baskin (2014), The Limits of Attraction, *JMR* 51(4) — https://journals.sagepub.com/doi/10.1509/jmr.12.0061
Huber, Payne & Puto (1982), original attraction effect, *JCR* 9(1)

**Free trials**
Yoganarasimhan et al., Design and Evaluation of Optimal Free Trials, *Management Science* — https://pubsonline.informs.org/doi/10.1287/mnsc.2022.4507 · preprint https://arxiv.org/abs/2006.13420
Zhang & Duan (2025), *Frontiers in Psychology* 16:1568868 — https://pmc.ncbi.nlm.nih.gov/articles/PMC12217587/

**Freemium**
Kumar (2014), Making "Freemium" Work, *HBR* 92(5) — https://hbr.org/2014/05/making-freemium-work · PDF https://vineetkumars.github.io/Papers/Freemium_May2014_Kumar.pdf

**News paywalls**
Chiou & Tucker (2013), *Information Economics and Policy* 25(2) — https://www.sciencedirect.com/science/article/abs/pii/S0167624513000097 · PDF https://www.oxy.edu/sites/default/files/assets/Economics/Chiou/chiou_and_tucker_paywalls.pdf
Pattabhiramaiah, Sriram & Manchanda (2019), *Journal of Marketing* 83(2) — https://journals.sagepub.com/doi/10.1177/0022242918815163 · AMA summary https://www.ama.org/2019/03/07/before-you-put-up-a-paywall-read-this-study/

**Vendor telemetry (label as such on camera)**
RevenueCat, State of Subscription Apps 2026 (115k apps, $16B) — https://www.revenuecat.com/state-of-subscription-apps
RevenueCat, State of Subscription Apps 2025 — https://www.revenuecat.com/state-of-subscription-apps-2025
Superwall, multi-page onboarding paywalls (40M opens) — https://superwall.com/blog/new-postmulti-page-onboarding-paywalls-convert-37-better-than-single-page-heres-why
DoWhatWorks, social proof on pricing pages — https://www.dowhatworks.io/blog/should-you-add-social-proof-on-pricing-pages
C+R Research, subscription cost survey (2022) — https://www.crresearch.com/blog/subscription-service-statistics-and-costs/ · CNBC coverage https://www.cnbc.com/2022/06/02/consumers-spend-133-more-monthly-on-subscriptions-than-they-realize.html

**Case studies**
Growth.Design, Blinkist trial paywall — https://growth.design/case-studies/trial-paywall-challenge
Purchasely × Blinkist (Eveline Moczko interview) — https://www.purchasely.com/blog/using-transparency-to-increase-your-conversion-rate-with-eveline-moczko-blinkist

**Rules & regulation**
Apple App Store Review Guidelines (3.1.2) — https://developer.apple.com/app-store/review/guidelines/
RevenueCat, guide to App Store rejections — https://www.revenuecat.com/blog/growth/the-ultimate-guide-to-app-store-rejections
Eighth Circuit vacatur of FTC Negative Option Rule (July 8, 2025) — https://www.cooley.com/news/insight/2025/2025-07-11-click-to-cancel-just-got-cancelled-eighth-circuit-vacates-entirety-of-ftcs-negative-option-rule