# Research Brief — Pricing-Page Psychology
### 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:** choice overload (Iyengar jam study, Scheibehenne meta, Chernev moderators) is fully covered in `decision-fatigue-research-brief.md` — do not rehash here. The relevant bridge: Chernev's "a dominant option makes overload disappear" is the empirical case for a highlighted "Most Popular" tier.

---

# ⚠️ PART 1 — CORRECTIONS

The standard tellings of six pricing-page staples needed changing. In each case the corrected version is either more accurate *and* more powerful, or is your honesty beat.

### 1. The Economist decoy — a classroom demo, not a field result ⭐ THE HONESTY BEAT

**What actually happened** — Ariely, *Predictably Irrational* (2008), ch. 1. The pricing on economist.com was real (Ariely reproduces the actual ad), but **the experiment was a survey of 100 MIT Sloan MBA students**, never a field test. Web-only $59 / print-only $125 / print+web $125 → **16 / 0 / 84**. Decoy removed, different 100 students → **68 / 32**. `SOLID` on the numbers, **but The Economist never measured anything. Nobody knows what the decoy did to actual subscriptions.**

Three more layers, all verified:
- Frederick, Lee & Baskin (2014, JMR, footnote 8) note the demo "is identical to one discussed in Kivetz, Netzer, and Srinivasan (2004)," whose version produced a milder **43% → 72%**. Ariely's 32→84 is the flashy outlier retelling.
- **Ariely credibility caveat:** the 2012 PNAS honesty paper was retracted (Sept 2021) after Data Colada showed the insurance data was fabricated; Duke's investigation (disclosed 2024) concluded the data was falsified but found no evidence Ariely *knowingly* falsified it. The Economist demo is unrelated, but if you lean on Ariely, expect comments. Coherent arbitrariness (§2) is co-authored with Loewenstein & Prelec — safer.
- And the kicker below: the effect the demo illustrates **largely fails with real products** (Correction 2).

**Safe wording:** *"Ariely showed the ad to a hundred MBA students... and when he deleted the middle option for another hundred students, preferences flipped. It's a great demo. It is not a field experiment — and what happened next in the research is the part nobody tells you."*

### 2. The decoy effect largely fails outside numeric tables — get this exactly right

**Frederick, Lee & Baskin (2014), JMR 51(4): 487–507 — "The Limits of Attraction."** `SOLID` — full primary PDF read. **38 studies total.** The scoreboard, from their own General Discussion:
- Fully abstract numeric-matrix stimuli: significant attraction effect in **4 of 5** studies
- Numeric + perceptual/verbal representation: **2 of 5**
- At least one attribute directly experienced or shown (photos, tasted drinks, real mints, popcorn, fruit): **0 of 27** — and one significant *reversal* ("repulsion effect"; e.g., bottled water target 70%→52% with decoy added, p<.01)

The smoking-gun manipulation: **the same gambles** produced a decoy effect when probability was a number (Study 2b, n=791: target 21%→37%, p<.001) and nothing when it was a pie chart (34% vs 35%). Study 3b: n=4,033 via Google Surveys, same split. Verbatim: attraction occurs "when stimuli [are] represented numerically, but not otherwise"; boundary conditions "seem to be so restrictive that its practical validity should be questioned."

**Huber, Payne & Puto's own 2014 response** ("Let's Be Honest About the Attraction Effect," JMR 51(4): 520–525) `SOLID` — concedes a lot:
- The 1982 paper was "designed as a *demonstration study*"; "We did not set out to suggest a tool for marketing practice"; stimuli were iteratively tuned until choice sets "generated significant and replicable demonstrations."
- Five inhibiting conditions (verbatim list): **(1) strong prior trade-offs, (2) inability to identify the dominance relationship quickly and easily, (3) cross-respondent value heterogeneity, (4) strong dislike of the decoy, or (5) strong liking for the decoy.**
- "We suspect that the asymmetric dominance effect occurs rarely in the marketplace today."
- Huber's own unpublished conjoint test: 586 respondents × 20 choices, ~4,000 dominated choice sets — "he could not detect any consistent increase of the target's share."
- The pivot that saves your Act Three: *"we more often observe successful use of **compromise** in the marketplace"* — see §4.

Also Yang & Lynn (2014, JMR 51(4): 508–513): **91 attempts, 23 product classes → only 11 reliable attraction effects**; pictures and meaningful verbal descriptions reduced it to chance. `SOLID`. 2020s consensus: real but fragile, flips with presentation format (Cataldo & Cohen 2019 — by-attribute layouts yes, by-alternative layouts no/reverse), killed by time pressure (Pettibone 2012). Note: a SaaS pricing page with numeric feature counts in a comparison table is closer to the format where it *works*; three photographed products is the format where it doesn't. That nuance is the useful takeaway.

**Original for the record:** Huber, Payne & Puto (1982, JCR 9(1)): six categories, all text/numeric two-attribute stimuli; average target share increase **+9.2 percentage points** (n=153, secondary-verified) — not the "30% sales boost" of blog legend.

### 3. SSN anchoring ("coherent arbitrariness") — real paper, contested replication

Ariely, Loewenstein & Prelec (2003, QJE 118(1)). **55 MIT Sloan MBA students**, six products (avg retail ≈$70), incentive-compatible BDM auctions. Top vs bottom SSN **quintile** WTP: cordless keyboard **$55.64 vs $16.09** (3.46×); correlations r = .32–.52 across products. `SOLID` — Table verified against the authors' own reprint.
- **Misquote trap:** the famous "57–107% more" is above-median vs below-median, per product. The quintile comparison is the ~2.2–3.5× claim. Don't mix them.
- **Replication asterisk (say it):** Fudenberg, Levine & Maniadis (2012, AEJ:Micro): anchor–valuation correlations −0.11 to 0.21, only one significant; quintile ratios 1.1–1.4 vs Ariely's 2.2–3.45. Maniadis, Tufano & List (2014, AER 104(1)): direct replication of the annoying-sounds anchoring, n=116 — effect ~half the original size and **p = 0.253**. Bergman et al. (2010, Econ Letters): found anchoring (~+45%) but smaller, decreasing with cognitive ability. Counter-critique: Data Colada #7 argues the "failure" is overstated. 
- **The defensible line:** anchoring on *estimates* is bulletproof (§1 below); anchoring on *willingness to pay* is real but probably one-half to one-third the size of the famous demo. `CONTESTED` as a WTP effect.

### 4. "Remove the dollar sign" — one lunch menu, and the overall test was not significant

Yang, Kimes & Sessarego (2009), *Cornell Hospitality Report* 9(8). Full report read. St. Andrew's Café (Culinary Institute of America, Hyde Park NY), **lunch only, Aug–Nov 2007, final n = 201 checks**, randomized by table. Three formats: "$20.00", "20.", and scripted "twenty dollars".
- **The overall ANCOVA effect of price format was NOT statistically significant.** One linear contrast was: numeral-only parties spent **$3.70 more than the restaurant average** (p<.05). The widely quoted "≈8% more per person" ($23.00→$24.87) is a derived figure the authors explicitly say "was not the subject of our statistical tests."
- Confound: the no-$ menu also dropped cents and rounded ("20."), so sign-removal is entangled with rounding.
- The scripted "twenty dollars" format — which the authors predicted would win — **tied for worst**, contradicting the folk "spell it out" advice AND their own hypothesis.
- `CONTESTED` as a finding; `BLACKLISTED` as a law. Safe wording: *"one Cornell field study, one restaurant, 201 checks, and the headline number wasn't even the tested one."*

### 5. Charm pricing — the real field numbers are stranger and better than the blog version

Anderson & Simester (2003, QME 1: 93–110), primary PDF read. Mail-order women's-clothing catalogs, real randomized field experiments:
- Pilot: the same dress at **$34 → 16 units, $39 → 21 units, $44 → 17 units**. The $39 price outsold the $5-cheaper price. Across 4 dresses: 66 units at $9-endings vs 46 ($5 lower) and 45 ($5 higher) — **≈+40%**, while a $10 price spread did nothing. (Tiny unit counts — say "in a small pilot.")
- Full experiments: Study 1 (60,000 catalogs) **≈+35%**; Study 2 (62,500) **≈+15%** overall, **+22% for new items**, ~+10% n.s. for established items; Study 3 (270,000) **≈+7%**, and a "Sale" cue alone (+0.215 coeff) beat the $9-ending effect. $9-endings added little when a Sale cue was already present. **$9.50 endings did nothing or hurt.**
- Their conclusion: 9-endings work as an **information cue** ("this is a deal"), strongest when customers lack other price information — not as arithmetic hypnosis. `SOLID`, but report it as conditional.
- **When charm pricing backfires:** Wadhwa & Zhang (2015, JCR 41(5)): rounded prices evaluate better for hedonic/feeling purchases (champagne at $40.00 beat $39.72/$40.28; camera-for-vacation vs camera-for-class flipped the effect; interactions p≤.004). Stiving (2000, Mgmt Sci): 0-endings signal quality — why luxury prices are round. Preregistered null: Escher et al. (2026, Frontiers, N=729): $9.99 vs $10.00 — no reliable effect on purchase intention. `SOLID` collectively.

### 6. "Good-better-best / 3 tiers is optimal" — no primary source exists

The legitimate neighbors are the compromise effect (§4 — middles gain *share*) and choice-model papers (Kivetz, Netzer & Srinivasan 2004 — formalizations, not tier-count prescriptions). Nothing establishes three tiers as revenue-optimal. HBR's "The Good-Better-Best Approach to Pricing" (Rafi Mohammed, 2018) is a practitioner playbook — its headline case (Allstate Your Choice Auto, 3.9M policies by 2008) is company-reported, no controlled test. `BLACKLISTED` as "research shows"; fine as "the compromise effect is why the pattern exists, and here's the practitioner rule of thumb, labeled as such."

---

# PART 2 — THE VERIFIED SPINE

## §1 — Anchoring: the robust foundation

**Tversky & Kahneman (1974, Science 185: 1124–1131).** Wheel of fortune, then estimate % of African countries in the UN: **median estimates 25 vs 45 for anchors 10 vs 65.** Verbatim detail worth using: **"Payoffs for accuracy did not reduce the anchoring effect."** `SOLID` — primary verified. Traps: the paper reports no n for this demo and never says the wheel was rigged (that's Kahneman's 2011 retelling).

**Northcraft & Neale (1987, OBHDP 39: 84–97) — experts anchor too.** Real Tucson house (appraised & listed at $74,900), 10-page MLS packet, real tours. Fake listing prices $65,900 / $71,900 / $77,900 / $83,900. 
- Amateurs (n=48): mean appraisals $63,571 → $72,196 across anchors (p<.001).
- **Professional agents** (n=21, ~7 yrs experience; ±12% conditions): **$67,811 vs $75,190 — a $7,379 swing from the same house** (p<.01). Experiment 2 (54 amateurs, 47 experts, $134,900 house): all effects p<.001.
- The denial data (the good part): **56.2% of amateurs but only 24.0% of experts** admitted considering the listing price (Exp 1); verbatim, experts' checklists "**flatly denied their use of listing price**." Agents claimed a 5% credibility window; ±12% anchors moved them anyway. `SOLID` — full primary PDF verified.
- ⚠️ The circulating "92% vs 56%" denial figures are wrong. Delicious meta-detail: Northcraft & Neale's own intro misquotes the 1974 wheel numbers (swapping anchor 65 and estimate 45). Even the anchoring paper anchored wrong on the anchoring paper.

**Replication status — Many Labs 1 (Klein et al. 2014, Social Psychology 45): anchoring is arguably the best-replicated effect in psychology.** 36 samples, N=6,344. Anchoring took **4 of the top 5 effect sizes in the whole project**: babies born/day **d = 2.42**, Mt. Everest d = 2.23, Chicago d = 1.79, SF→NYC d = 1.17; 94–100% of individual samples significant; replications came out *larger* than the originals. `SOLID`. Honest caveat: these are anchors on factual estimates, not prices — pair with §2's WTP asterisk.

**Incidental anchors** — Critcher & Gilovich (2008, JBDM 21): restaurant named "Studio 97" vs "Studio 17" → diners would spend **$32.84 vs $24.58** (p=.03); jersey #94 vs #54 shifted judged probabilities. `CONTESTED` — real paper, but n≈200 per study and p-values .03–.05. Say "suggestive."

**The purchase-limit soup study — usable only with a flag.** Wansink, Kent & Hoch (1998, JMR 35(1)): 3 Iowa supermarkets, "Campbell's 79¢" end-cap, 906 shoppers analyzed. Cans bought: **no limit 3.32 · "limit 4" 3.58 · "limit 12" 7.0** (p<.01) — the arbitrary "12" more than doubled purchases. Verified: **this paper is NOT retracted** (Crossref clean), predates the Food and Brand Lab era, and has heavyweight co-authors (Hoch). But lead author Wansink had ~13–18 retractions and resigned from Cornell after a misconduct finding. `CONTESTED` — if used, disclose the Wansink situation on camera; never replicated at scale.

**Reference prices ("was $X, now $Y"):** Urbany, Bearden & Weilbaker (1988, JCR 15(1)): plausible advertised reference prices raised perceived value — **and exaggerated, implausible ones had generally the same positive effects, even among skeptics.** `SOLID` direction (keep qualitative). This is the peer-reviewed basis for strikethrough pricing.

## §2 — Coherent arbitrariness (see Correction 3 for the full treatment)

Two bonus details, both verified: (a) coherence is the real finding — >95% of subjects valued the keyboard > trackball and rare wine > average wine; the *level* was arbitrary, the *structure* was orderly. Demand curves look rational even when built on sand. (b) A companion study with **77 MIT executives**: SSN–bid correlations ~.29 (p<.01), and **72% claimed the anchor didn't influence them** — pair with Northcraft's denying realtors for a "you won't feel it working" beat. `SOLID`.

## §3 — Decoy / asymmetric dominance (see Correction 2 — that IS this section)

## §4 — The compromise effect: the one that actually holds up

**Simonson (1989, JCR 16(2)) — the original.** Brands gain share when they become the middle option. Verified table data (via S&T 1992 reprint): calculator batteries — brand share **43% → 60%** when a more extreme option was added, *even when the extreme option was marked unavailable* (ns 119–126). `SOLID`. ⚠️ The freestanding "+17.5% average" figure circulates from secondary sources only — I could not verify it in the primary. Don't use it.

**Simonson & Tversky (1992, JMR 29(3): 281–295)** — full primary PDF read. 22 experiments, n=100–220 each, **real color pictures from a merchandise catalog** (note: pictures — and the effects still appeared, unlike the decoy). The Minolta camera set, exact:
- Two options (n=106): X-370 $169.99 **50%** / Maxxum 3000i $239.99 **50%**
- Add Maxxum 7000i $469.99 (n=115): **22% / 57% / 21%** — the mid camera's share vs the cheap one went from 50% to 72% (p<.01)
- Adding a premium option didn't sell the premium option — **it sold the middle.** `SOLID`
- Extremeness aversion, verbatim: "the attractiveness of an option is enhanced if it is an intermediate option in the choice set and is diminished if it is an extreme option."

**Replication status — this is the robust one.** Neumann, Böckenholt & Sinha (2016, JCP 26(2)): meta-analysis, **142 observations — extremeness aversion is robust; the same product is 29% to 88% more likely to be chosen when intermediate than when extreme.** `SOLID`. Plus Huber/Payne/Puto's 2014 line: compromise, not attraction, is what marketers successfully use. Boundaries: weakens with larger sets and by-alternative formats — it's a tilt (share shifts), not magic.

**Script architecture note:** this is the clean arc — the famous effect (decoy) is fragile; the boring cousin (compromise) is the one that survives meta-analysis and is why good-better-best exists.

## §5 — Center-stage effect: real, small, thin

Valenzuela & Raghubir (2009, JCP 19(2)): center gum chosen **50% vs 29.2% (left) vs 20.8% (right)** — but **n=48**. Mechanism: people infer the middle option is the popular one. Raghubir & Valenzuela (2006, OBHDP 99): *The Weakest Link* contestants in the two center positions reached the final **42.5%** and won **45%** of the time vs **17.5% / 10%** for the extremes. Rodway et al. (2012): center preference replicates in horizontal and vertical arrays (incl. real socks). `CONTESTED` overall — a handful of labs, modest ns, moderated by array size and goal. ⚠️ There is no Wimbledon study — that's a false memory in circulation; it's a TV game show. Safe use: "middle placement gets a measurable tilt in small studies" — supports putting the plan you want sold in the center, weakly.

## §6 — Left-digit / charm pricing (with Correction 5)

**Thomas & Morwitz (2005, JCR 32(1)) — the mechanism.** $2.99 vs $3.00 changes perceived magnitude; **$3.59 vs $3.60 does not** (left digit unchanged: F<1). Study 1b: magnitude ratings 35.6 vs 55.8 (p<.01) for 2.99/3.00; nothing for 2.79/2.80 or 3.19/3.20. Effect dies when the comparison price is far away ($2 distance: F<1). Latency data: "4.99" judged *faster* than "5.00" against a $5.50 standard — it genuinely feels farther below. `SOLID` — primary PDF read. Small lab samples (27–154); perception, not sales — which is why A&S (Correction 5) and the next entry matter.

**Strulov-Shlain (2023, Review of Economic Studies 90(5): 2612–2645) — the modern estimate.** NielsenIQ scanner data: **25 US chains, ~3,500 products, 78M observations, $4.3B in sales, 2006–2019.** In-sample: 41% of prices end .99; 87% end in 9.
- Bias parameter θ ≈ **0.2**: consumers respond to a 1-cent increase that crosses a dollar ($4.99→$5.00) **"as if it were more than a twenty-cent increase."** A within-dollar cent is perceived as ~0.8¢.
- **The twist: firms are the irrational ones.** Given θ=0.2, essentially *all* prices should end in 99 — firms use it on only 30–40%, pricing as if θ were 0.01–0.03. All 25 chains underestimate the bias.
- Retailers **forgo an estimated 1–4% of potential gross profits**. `SOLID` — published figures verified.
- Trap: "consumers see $4.99 as $4" is too strong — the model is a weighted blend (~$4.60–4.80 perceived).

## §7 — Price presentation micro-effects

| Effect | Finding | Source | Confidence |
|---|---|---|---|
| **Precise prices seem smaller** | $395,425 judged *smaller* than $395,000; in **>27,000 real-estate transactions** (S. Florida + Long Island), round-listed homes sold ≈**0.73% lower** than precise-listed comparables (~$3,600 on $500k) | Thomas, Simon & Kadiyali (2010), *Marketing Science* 29(1): 175–190 | `SOLID` (0.73% via reputable secondary) |
| **Precise anchors constrain adjustment** | Wholesale-cost guesses adjust much farther from $5,000 than from $4,988 or $5,012 — precise anchors evoke a finer mental scale | Janiszewski & Uy (2008), *Psych Science* 19(2) | `SOLID` concept; per-study means paywalled — don't quote exact deltas |
| Precise first offers anchor negotiations harder | $4,983 offers drew more conciliatory counteroffers than $5,000 | Mason et al. (2013), JESP 49 | `SOLID`, lab; can backfire vs experts (Loschelder) |
| **Font size** | Sale price in *smaller* font than regular price → lower perceived magnitude, higher purchase likelihood | Coulter & Coulter (2005), JCP 15(1), 3 experiments | `SHAKY` — one team, lab-only, later boundary/reversal papers (J. Retailing 2022). Not a law. |
| **Syllables / commas** | "$1,499.00" vs "$1499" — longer verbal encoding → larger perceived magnitude, even reading silently. Strip commas and ".00" | Coulter, Choi & Monroe (2012), JCP 22(3) | `SOLID` citation, lab-only |
| Currency sign | See Correction 4 — overall test n.s. | Yang, Kimes & Sessarego (2009) | `CONTESTED`/`BLACKLISTED` as law |

## §8 — Partitioned & drip pricing: the strongest field evidence in the whole space

**⭐⭐ Blake, Moshary, Sweeney & Tadelis (2021), *Marketing Science* 40(4) — the StubHub experiment.** Aug 19–31, 2015; **several million desktop users**, ~50/50 randomized. Control saw all-in prices while browsing; treatment saw base prices with **~15% buyer fees revealed only at checkout.** All numbers verified against the NBER WP (25186) tables:
- Drip-fee users **spent +20.64%** (SE 1.38) — "about 21% more"
- **+14.1%** more likely to purchase at least once; +13.24% more transactions
- Conditional on buying, paid **+5.42% more per order** — they bought **better tickets**, not just more ("quality upgrade effect"; ≥28% of the revenue gain)
- The funnel tells the story: drip users clicked listings ~19% more, then were **~45% more likely to bail at the final checkout stage** when fees appeared — and *still* spent 21% more overall
- **Experienced users (10+ prior visits) still spent ~15% more.** Experience doesn't immunize.
- StubHub switched the whole platform to back-end fees on Sept 1, 2015. Sellers responded by listing higher-quality inventory (row-A listings +15%). `SOLID`
- Framing trap: this is a *revenue* result, not a welfare-positive one — the paper is about how hiding fees distorts choice. It's simultaneously the best evidence that drip pricing "works" and the reason it's now illegal in key categories (below).

**Morwitz, Greenleaf & Johnson (1998, JMR 35(4)) — the original partitioned-pricing paper.** Study 1: incentivized auction for a jar of pennies; a 15% buyer's premium group paid more in total — bidders anchored on the bid and under-adjusted. Study 2: telephone survey, catalog offer. Verified via Morwitz's own 2016 JCP review: **12.2–35.6% of consumers ignored the surcharge completely** when recalling total price; in one condition only **21.9% actually calculated** the total (54.8% used a heuristic, 23% ignored). `SOLID` design/direction; exact cell means paywalled. Trap: it was a lab auction + phone survey, not a "phone auction."

**Hossain & Morgan (2006, BE J. Econ Analysis & Policy) — eBay shipping.** 80 real auctions (CDs, Xbox games). Low opening bid + $3.99 shipping beat $4 opening + free shipping: CDs **+~21% revenue, 9/10 matched pairs**; pooled 16/20 pairs (p=.005). High-reserve Xbox: hidden-fee variant won **10/10 pairs (+11%)**. Boundary: the effect vanished when the total reserve hit ~53% of retail — shrouding fails when the fee is too large relative to value. Follow-up: Brown, Hossain & Morgan (2010, **QJE** 125(2)): raising shipping boosts revenue *especially when hidden*; disappears when shipping is displayed on the search page. `SOLID`
- Nuance for fairness: Bertini & Wathieu (Mktg Sci 2008) — partitioning can raise *or* suppress demand depending on which component gets attention. Not uniformly demand-increasing.

**Regulation (dates verified):** FTC "Junk Fees Rule" (16 CFR 464): announced Dec 17, 2024 (4–1 vote), **effective May 12, 2025** — live-event tickets and short-term lodging must display total price including mandatory fees upfront. California SB 478 economy-wide since **July 1, 2024**. ⚠️ The DOT airline ancillary-fee rule was **vacated by the Fifth Circuit** — do NOT cite it as in force. `SOLID`

## §9 — Temporal reframing: per-day, monthly, annual

(Coordinate with the paywall brief — covered here from the display angle only.)

**Gourville (1998, JCR 24(4)) — pennies-a-day.** Charity payroll deduction: **"85 cents per day" → 52% agreed; "$300 per year" → 30%** — and note the sleight: $0.85/day ≈ $310/yr, *more* money, and it still won. Mechanism: per-day framing retrieves trivial comparisons (coffee); aggregate framing retrieves big ones. `SOLID` (headline verified via multiple academic citations; original paywalled). **Boundary (Gourville 2003, Marketing Letters):** backfires when the daily number stops being trivial — people prefer monthly framing for rent at "$25/day" or taxes at "$58/day."
- ⚠️ Trap: the "$1/day vs $350/year" version is the **Atlas & Bartels (2018)** stimulus, not Gourville's.

**Atlas & Bartels (2018, JCR 45(2))** — 8 experiments + field test: periodic pricing raises intentions partly by increasing **perceived benefits** (not just shrinking perceived cost), and extends beyond trivial amounts in some contexts. `SOLID`.

**⭐ Hershfield, Shu & Benartzi (2020, Marketing Science 39(6)) — the modern field result.** Fintech savings app (Acorns): framing the identical recurring deposit as **"$5/day" vs "$150/month" roughly quadrupled enrollment (~30% vs ~7%)** — and the daily frame *eliminated* the participation gap between lowest- and highest-income users. `SOLID`. This is your best per-day-framing number: recent, field, peer-reviewed.

**Flat-rate bias — Lambrecht & Skiera (2006, JMR 43(2)).** 10,882 DSL customers: **48.1% exhibited flat-rate bias vs 8.5% pay-per-use bias; over half of the flat-rate-biased paid ≥100% more** than their cheapest tariff — knowingly, happily (insurance + taxi-meter effects, confirmed), and it did NOT increase churn. Pay-per-use bias did. `SOLID`. The honest case for why unlimited plans print money and customers don't resent them.

**"$X/mo billed annually" specifically: no direct peer-reviewed test exists.** It's an untested extrapolation from PAD/periodic-pricing research. Label it as such.

## §10 — Price ordering: high-to-low vs low-to-high

**Suk, Lee & Lichtenstein (2012, JMR 49(5)) — the actual evidence for "anchor high first."** 8-week field experiment in a bar, 13 beers: descending menu (expensive first) → average paid **$6.02** vs ascending → **$5.78**. **+$0.24/beer, ~4%, significant.** Mechanism: high-to-low sets a high reference; trading down feels like a quality loss. `SOLID` (figures via reputable secondaries of the JMR paper; venue/direction confirmed in abstract). That's the entire real basis for "show your expensive plan first" — a 4% nudge in one bar plus labs. Not nothing; not a law.

## §11 — SaaS practitioner data: vendor, not research

Price Intelligently → ProfitWell (Patrick Campbell) → acquired by **Paddle, 2022 (~$200M)**. Major citability problem: old profitwell.com data posts now redirect to Paddle pages where the underlying data is often gone.

| Claim in circulation | Reality | Confidence |
|---|---|---|
| "Companies spend only 6–8 hours on pricing" | Campbell's stated claim is **~8 hours over the company's lifetime** — self-reported survey of PI's audience, methodology never published | `SHAKY` — attribute to Campbell, never "research shows" |
| "Value-based pricing lifts revenue 30–40%" | Traces to PI's own **service pitch** laundered into a finding | `BLACKLISTED` |
| Monetization beats acquisition **2–4×** per 1% improvement | ProfitWell panel analysis (~23.4k companies); original posts largely dead links | Vendor-grade `CONTESTED` |
| Annual plans reduce churn | ProfitWell panel (~2,500 companies), direction consistent; content now redirected | Vendor-grade `CONTESTED` |
| Freemium conversion benchmarks | Cite **OpenView (2–5%)** and **ChartMogul (3–5% good, 8–12% great)** instead of "ProfitWell says" | `CONTESTED` (vendor, but published) |
| "Revisit pricing quarterly → grow faster" | No primary publication located | `BLACKLISTED` — conference lore |

---

# PART 3 — HOOK CANDIDATES (verified, striking)

1. **The StubHub number:** "StubHub randomly hid its ~15% fees until checkout for half of several million visitors. Those people spent **21% more** — and bought *better seats*. Two weeks later StubHub switched the whole site. Ten years later, that pricing style is illegal for ticket sites." (§8, `SOLID`)
2. **The $39 dress:** "A catalog sold the same dress at $34, $39, and $44. The $39 version outsold the $34 version. Charging five dollars MORE sold more units." (Correction 5, `SOLID` — flag small pilot; back with the 60k–270k-catalog studies)
3. **The expert anchor:** "Real-estate agents toured a real house. A fake listing price moved their professional appraisals by **$7,379** — and their written reports 'flatly denied' using the listing price at all." (§1, `SOLID`)
4. **The penny that costs twenty cents:** "Across 78 million supermarket price observations, crossing from $4.99 to $5.00 hits demand like a **20-cent** increase — and the punchline is that *retailers* underprice this, leaving 1–4% of profit on the table." (§6, `SOLID`)
5. **The decoy scoreboard:** "In 27 studies where people could actually see or taste the products, the famous decoy effect worked **zero** times." (Correction 2, `SOLID`)
6. **$5/day:** "Reframing $150/month as $5/day quadrupled savings-plan enrollment — 7% to 30% — in a real fintech field experiment." (§9, `SOLID`)

# PART 4 — NOVEL-ANGLE CANDIDATES

- **"The famous effect is the fragile one."** Decoy (famous) barely survives contact with real products; compromise (obscure) survives a 142-observation meta-analysis and is what the market actually uses. Almost no creator gets this right — it's the video's spine.
- **"You won't feel it working."** Three-way sourced denial pattern: Northcraft's experts flatly denied using the anchor; 72% of Ariely's executives claimed no influence; T&K's subjects weren't fixed by accuracy payoffs. Introspection is not a defense.
- **"The irrational actor is the firm."** Strulov-Shlain's inversion — consumers' left-digit bias is real, and *retailers* systematically misprice around it. Fresh take vs. the tired "brains are broken" framing.
- **Drip pricing's quality-upgrade effect:** hiding fees doesn't just extract more money — it changes *what* people buy (better seats, +5.4%/order). Price display shapes product choice, not just conversion.
- **The regulation arc:** 1998 lab paper → 2015 platform-scale experiment → 2024 FTC rule. A rare complete story of psychology research becoming law.
- **Flat-rate bias as the honest ending:** half of subscribers overpay ≥100% on flat plans, don't churn, and are *happier* — the uncomfortable case that some "exploitation" is what customers actively want (insurance against the taxi-meter feeling).
- **The repulsion effect:** badly executed decoys backfire — a moldy orange next to your target makes people flee the category. Practical stakes for copying "add a decoy tier" advice blindly.

# PART 5 — THE BLACKLIST

Never say these on camera:

| Claim | Why |
|---|---|
| "The Economist tripled subscriptions with a decoy" | Nobody measured Economist conversions. Classroom demo, n=100 MBA students. |
| "The decoy effect boosts sales ~30%" | 1982 original: +9.2 share points, text stimuli tuned to produce the effect. |
| "The decoy effect was debunked" | Also wrong — it replicates with numeric attribute tables (4/5); it fails with perceptual stimuli (0/27). Format-dependent. |
| "Removing the dollar sign makes people spend 8% more" | Overall test n.s.; one contrast p<.05; one restaurant, 201 checks; confounded with rounding. |
| "Prices ending in 9 always win" | Conditional: new items yes, Sale-cue present barely, hedonic/luxury can reverse, $9.50 hurts, preregistered null exists. |
| "Research shows 3 tiers is optimal" | No primary source. Compromise effect ≠ tier-count prescription. |
| "Consumers read $4.99 as $4" | Model says perceived ≈$4.60–4.80; the 20¢ figure applies only at dollar crossings. |
| "People bid their SSN" without the replication asterisk | AER/AEJ replications: half the size to nothing on WTP. |
| "57–107% more" as the SSN quintile gap | That's above/below median. Quintiles = 2.2–3.5×. |
| "92% of amateurs vs 56% of experts denied the anchor" | Real figures: 56.2% vs 24.0% *mentioned* it (Exp 1). |
| The Wimbledon center-stage study | Doesn't exist. It's *The Weakest Link* (game show), 42.5%/17.5%. |
| Soup study without the Wansink disclosure | Paper not retracted, but lead author had ~13–18 retractions; disclose or drop. |
| "Gourville: $1/day vs $350/year" | That stimulus is Atlas & Bartels 2018. Gourville: $0.85/day vs $300/yr, 52% vs 30%. |
| "Value-based pricing lifts revenue 30–40% (research)" | It's Price Intelligently's service pitch. |
| "Companies spend 6–8 hours/year on pricing" | Campbell's claim is 8 hours over the company's *lifetime*, unpublished survey. |
| DOT airline fee rule as current law | Vacated by the Fifth Circuit. FTC rule (tickets/lodging) and CA SB 478 are the live ones. |
| "The middle option gets chosen 66% of the time" (any universal center %) | Center-stage flagship result is n=48 gum choice (50%). |
| "+17.5% compromise-effect average" | Secondary-source figure; not verified in Simonson 1989. Use the 43→60% or camera 50→57% instead. |

# APPENDIX — AT-A-GLANCE VERDICTS

| Tactic on the pricing page | Verdict | Best evidence |
|---|---|---|
| Show an expensive plan/anchor first | Real, small (~4% in the one field test) | Suk et al. 2012 bar study |
| Strikethrough "was/now" reference prices | Works, even when implausible | Urbany et al. 1988 |
| Decoy tier (asymmetric dominance) | Only in numeric comparison-table formats; fails/reverses with rich stimuli | FLB 2014; HPP 2014 |
| Compromise / middle tier | Robust — the workhorse effect | Neumann et al. 2016 meta (142 obs) |
| Center placement of featured plan | Weak tilt, thin literature | Valenzuela & Raghubir 2009 (n=48) |
| Charm pricing (.99 / $X9) | Field-verified but conditional; can backfire on premium/hedonic | A&S 2003; Strulov-Shlain 2023; Wadhwa & Zhang 2015 |
| Precise (non-round) prices | Seem smaller; 27k-transaction field support | Thomas, Simon & Kadiyali 2010 |
| Small price font | One lab team; not a law | Coulter & Coulter 2005 |
| Drop "$", commas, ".00" | "$"-removal overclaimed (n.s. overall); comma/cents-stripping has lab support | Yang et al. 2009; Coulter et al. 2012 |
| Drip/partitioned fees | The strongest effect in the space (+21% spend) — and now regulated | Blake et al. 2021; FTC 2025 |
| Per-day reframing ("$5/day") | Strong, field-replicated; backfires for large daily amounts | Hershfield et al. 2020; Gourville 2003 |
| "$X/mo billed annually" | No direct peer-reviewed test — practitioner extrapolation | — |
| Exactly 3 tiers | Myth as a research claim | — |

---

# PART 6 — PRIMARY SOURCES

**Anchoring & coherent arbitrariness**
Tversky & Kahneman (1974) — https://www.cs.tufts.edu/comp/150AIH/pdf/TverskyKa74.pdf
Northcraft & Neale (1987) — https://www.smallprojectsbureau.com/wp-content/uploads/2020/01/northcraft_neale.pdf
Klein et al., Many Labs 1 (2014) — https://stanford.edu/~knutson/jdm/klein14.pdf
Critcher & Gilovich (2008) — http://static1.1.sqspcdn.com/static/f/409296/3798405/1249679424930/Critcher_Anchoring.pdf
Wansink, Kent & Hoch (1998) — https://journals.sagepub.com/doi/10.1177/002224379803500108
Ariely, Loewenstein & Prelec (2003) — https://web.mit.edu/ariely/www/MIT/Chapters/CA.pdf (QJE: https://academic.oup.com/qje/article/118/1/73/1917051)
Fudenberg, Levine & Maniadis (2012) — https://www.aeaweb.org/articles?id=10.1257/mic.4.2.131
Maniadis, Tufano & List (2014) — https://www.aeaweb.org/articles?id=10.1257/aer.104.1.277
Data Colada #7 (counter-critique) — https://datacolada.org/7
Urbany, Bearden & Weilbaker (1988) — https://academic.oup.com/jcr/article-abstract/15/1/95/1840979

**Decoy, compromise, center-stage**
Huber, Payne & Puto (1982) — DOI 10.1086/208899
Frederick, Lee & Baskin (2014) — https://web2-bschool.nus.edu.sg/wp-content/uploads/media_rp/publications/tF83O1430805722.pdf
Huber, Payne & Puto (2014) — https://people.duke.edu/~jch8/bio/Papers/HuberPaynePutoJMR%202014.pdf
Yang & Lynn (2014) — DOI 10.1509/jmr.14.0020
Simonson (1989) — DOI 10.1086/209205
Simonson & Tversky (1992) — https://cognition.aau.at/bg/BA/Simon%20&%20tversky,%201992.pdf
Neumann, Böckenholt & Sinha (2016) — DOI 10.1016/j.jcps.2015.05.005
Valenzuela & Raghubir (2009) — DOI 10.1016/j.jcps.2009.02.011
Raghubir & Valenzuela (2006) — OBHDP 99(1), 66–80
Rodway, Schepman & Lambert (2012) — DOI 10.1002/acp.1812
Mohammed, HBR good-better-best (2018) — https://hbr.org/2018/09/the-good-better-best-approach-to-pricing

**Left-digit, precision, presentation**
Thomas & Morwitz (2005) — archived: web.archive.org copy of forum.johnson.cornell.edu/faculty/mthomas/LeftDigitEffect.pdf
Anderson & Simester (2003) — https://www.kellogg.northwestern.edu/faculty/anderson_e/htm/personalpage_files/Papers/Effects_of_9_Price_Endings_on_Retail_Sales.pdf
Strulov-Shlain (2023) — https://academic.oup.com/restud/article-abstract/90/5/2612/6931812 (PDF: https://gwern.net/doc/economics/2023-strulovshlain.pdf)
Wadhwa & Zhang (2015) — https://www.smallprojectsbureau.com/wp-content/uploads/2020/01/wadhwa-zhang-2015.pdf
Stiving (2000) — Management Science 46(12), 1617–1629
Escher et al. (2026, preregistered null) — https://www.frontiersin.org/journals/behavioral-economics/articles/10.3389/frbhe.2026.1828446/full
Janiszewski & Uy (2008) — https://journals.sagepub.com/doi/10.1111/j.1467-9280.2008.02057.x
Thomas, Simon & Kadiyali (2010) — https://pubsonline.informs.org/doi/10.1287/mksc.1090.0512
Mason et al. (2013) — https://www.sciencedirect.com/science/article/abs/pii/S0022103113000401
Coulter & Coulter (2005) — https://myscp.onlinelibrary.wiley.com/doi/abs/10.1207/s15327663jcp1501_9
Coulter, Choi & Monroe (2012) — https://myscp.onlinelibrary.wiley.com/doi/abs/10.1016/j.jcps.2011.11.005
Yang, Kimes & Sessarego (2009) — https://ecommons.cornell.edu/bitstreams/d9504484-4912-4291-a65c-f5b44461302b/download

**Partitioned/drip pricing, framing, ordering**
Morwitz, Greenleaf & Johnson (1998) — DOI 10.1177/002224379803500404 (2016 review w/ figures: Morwitz et al., JCP 26(1), Wharton-hosted PDF)
Blake, Moshary, Sweeney & Tadelis (2021) — DOI 10.1287/mksc.2020.1261 (WP: https://www.nber.org/papers/w25186)
Hossain & Morgan (2006) — https://faculty.haas.berkeley.edu/rjmorgan/ebay.pdf
Brown, Hossain & Morgan (2010) — QJE 125(2), 859–876
FTC Junk Fees Rule — https://www.federalregister.gov/documents/2025/01/10/2024-30293/trade-regulation-rule-on-unfair-or-deceptive-fees
California SB 478 — https://oag.ca.gov/hiddenfees
Gourville (1998) — JCR 24(4), 395–408; Gourville (2003), Marketing Letters 14(2)
Atlas & Bartels (2018) — JCR 45(2), 350–367
Hershfield, Shu & Benartzi (2020) — Marketing Science 39(6)
Lambrecht & Skiera (2006) — JMR 43(2), 212–223 (authors' PDF via uni-frankfurt.de)
Suk, Lee & Lichtenstein (2012) — JMR 49(5), 708–717

**SaaS practitioner (vendor-grade)**
OpenView Product Benchmarks; ChartMogul SaaS Conversion Report; Paddle/ProfitWell (note: pre-2022 data posts largely dead-linked)