Pricing Myths: Charm Pricing, Tiers, and the Numbers Founders Repeat

Which pricing tactics have real experiments behind them, and which are consultant folklore wearing a decimal point

Research date: August 9, 2026
Scope: B2C retail and B2B/SaaS pricing tactics, US and international. Findings from one channel (mail-order retail, packaged goods, SaaS platforms) are labeled as such and never generalized into a universal law of pricing.
Evidence standard: peer-reviewed academic pricing and decision research with a controlled design first (Anderson and Simester on 9-endings; Huber/Payne/Puto and the attraction-effect replication literature; Thomas and Morwitz on the left-digit effect; Tversky and Kahneman on anchoring; Mela/Gupta/Lehmann on promotion); transparent large-sample platform and consultant datasets with a stated method second (OpenView, ProfitWell/Paddle, McKinsey); vendor guides with partial methods third; content-farm "increase revenue X% with this pricing trick" figures are not accepted as proof.

Executive summary

Most pricing advice a founder hears is a real laboratory finding that has been stripped of every condition that made it true, then restated as a law. The honest version is more specific and more useful: a few of these tactics have genuine controlled experiments behind them, the effects are almost always narrow and context-dependent, and the single most-repeated pricing statistic in the category, the McKinsey "1% price beats 1% volume" number, is an accounting identity that quietly assumes the one thing that never actually holds.

What holds up:

• Prices ending in 9 really do sell more, in field experiments, but the effect is narrow. Anderson and Simester ran three field experiments with a real mail-order retailer and found a $9 ending raised demand in all three. In their headline test, a women's clothing item priced at $39 outsold the same item at $34, a lower price. The catch they state themselves: the effect was stronger for new items, and it weakened or disappeared when a "Sale" cue was also present. (Anderson and Simester, Quantitative Marketing and Economics, 2003)
• The decoy effect is a real lab phenomenon that does not reliably survive contact with real products. Huber, Payne, and Puto documented asymmetric dominance in 1982, and Ariely's Economist demonstration made it famous. But "always use three tiers so the middle wins" is a double error: the effect makes a specific target option win, not "the middle," and large replication efforts found it fragile once stimuli look like actual products instead of two-number abstractions. (Huber, Payne, Puto, Journal of Consumer Research, 1982; Yang and Lynn, Journal of Marketing Research, 2014)
• "A 1% price increase raises profit more than a 1% volume increase" is arithmetically true and practically misleading. It traces to Marn and Rosiello's 1992 study of 2,463 companies, where a 1% price improvement lifted operating profit 11.1% versus 3.3% for a 1% volume gain. The 11.1% is real. It also assumes volume does not move when you raise price, which is the entire question every founder is actually asking. (Marn and Rosiello, Harvard Business Review, 1992)
• The left-digit effect is a genuine, replicated cognitive finding, and it is conditional. Thomas and Morwitz showed $2.99 is perceived as meaningfully cheaper than $3.00, but only when the leftmost digit changes. $3.59 versus $3.60 does not get the same effect, because the dollar digit is identical. The mechanism is real; the "always drop a penny" rule ignores the condition that switches it on. (Thomas and Morwitz, Journal of Consumer Research, 2005)
• "Value-based pricing beats cost-plus" is directionally sensible and largely undemonstrated by controlled evidence. It is consultant doctrine, repeated for forty years, and the specific profit figures attached to it circulate without an auditable method. The reasoning is sound. The proof is mostly selection and survivorship: firms that can price on value tend to be firms with differentiated products, which would earn more anyway.
• "Annual plans should give about two months free (roughly 17% off)" is a convention, not a finding. The 16.7% figure is popular because "12 months for the price of 10" is easy to say, not because any experiment identified it as the profit-maximizing discount. It is a Schelling point, and it is fine to treat it as one, as long as nobody calls it evidence.

The practical conclusion for a founder setting a price: the tactics with the strongest evidence (9-endings, the left-digit effect) move behavior at the margin in specific contexts and can backfire on premium positioning. The tactics sold as universal levers (three-tier decoys, "always discount annually," value-based pricing as a guaranteed margin lift) are either fragile, conventional, or unproven at the level of generality they are stated. The single number to stop repeating uncritically is the "1% price" statistic, not because it is false, but because it silently holds volume constant. Test the tactic on your own prices. The one thing every honest source in this category agrees on is that the effect depends on your product, your buyer, and your context, which is exactly what a portable statistic erases.

Claim-by-claim verification

Provenance flags

• The "1% price beats 1% volume" number holds volume constant. The 11.1% figure is real and traces to Marn and Rosiello (1992) on 2,463 companies. It is computed by moving one lever 1% with everything else, including unit volume, frozen. In practice, a price increase usually loses some volume, so whether price or volume "wins" for a specific business depends on that business's demand elasticity, which the statistic silently sets to zero. Quote the number with its assumption or not at all. (Marn and Rosiello, 1992)
• The Economist decoy story is a hypothetical-choice demonstration, not a live pricing test. Ariely's 16/0/84 then 68/32 split came from MIT students choosing on paper, not from a magazine's real revenue. It is a vivid teaching case, and it is not the same evidence as a field experiment. Cite it as a demonstration, and pair it with the replication debate. (Ariely, Predictably Irrational, 2008)
• "Value-based pricing earns 31% more" and "94% of promotions fail" have no auditable origin in this review. They surface in pricing-consultant and retail-analytics content without a linked sample. Do not publish either as a fact.

What the primary sources actually say

Charm pricing: a real field effect with named limits

The strongest evidence for "prices ending in 9 sell more" is not a blog and not a lab. Anderson and Simester ran three field experiments with an actual mail-order retailer, manipulating price endings on real merchandise sent to real customers. Across all three, a $9 ending raised demand. In the most-cited test, a women's clothing item generated more sales at $39 than at $34, a genuinely lower price, which is the finding that makes the tactic interesting rather than trivial. (Anderson and Simester, 2003)

The authors are careful about the boundaries, and the popular retelling drops all of them. The effect was stronger for new items than for items the retailer had sold in prior years, consistent with a 9-ending working as a signal of value that regular buyers already have their own reference price for. And the effect weakened when the retailer also used a "Sale" cue, because two low-price signals compete rather than stack. The honest statement is "a 9-ending can raise demand, most reliably on new items and absent other discount signals," not "always end in 9." On premium and luxury positioning, where a round number signals quality, the same tactic can work against you. That boundary is not in the study, and it is worth stating as inference rather than as a finding.

The left-digit effect: the cognition under the tactic, and its switch

Thomas and Morwitz supplied the mechanism behind 9-endings, and in doing so specified exactly when it fails. Their 2005 work showed that a nine-ending price is perceived as smaller than a price one cent higher only when the leftmost digit changes. $2.99 reads as meaningfully less than $3.00 because the dollar digit drops from 3 to 2. $3.59 versus $3.60 gets little of the effect, because the left digit is 3 either way, so the drop is happening in a place the mind discounts. They also found the effect depends on the numeric distance between the target price and a competing price. (Thomas and Morwitz, 2005)

This is why "just drop a penny" is an incomplete rule. Pricing at $4.99 exploits the effect; pricing at $4.59 mostly does not, because there is no left-digit rollover. The tactic is real and it is mechanical, which means it is also predictable enough to know when it does nothing.

The decoy effect: real in the lab, fragile in the wild, and misdescribed

The decoy or attraction effect has the most interesting evidentiary arc in this report. Huber, Payne, and Puto established it in 1982: adding an option that is clearly worse than one existing option but not worse than another (an asymmetrically dominated alternative) can raise the share of the option that dominates it. This violates a basic axiom of rational choice, which is why it became famous. Ariely's Economist demonstration is the popular version: offered web-only at $59, print-only at $125, and print+web at $125, students chose print+web 84% of the time, but with the useless print-only decoy removed, print+web fell to 32%. (Huber, Payne, Puto, 1982)

Two things get lost in the "always use three tiers" retelling. First, the effect elevates a specific target, the option the decoy is dominated by, not "the middle option" by virtue of being in the middle. Putting three prices on a page does not summon it; a decoy has to be built to be worse than the option you want chosen and only that option. Second, the effect is fragile. Frederick, Lee, and Baskin (2014) reported that it largely disappeared when stimuli were realistic products rather than two-number abstractions, and Yang and Lynn (2014), across a large batch of replication attempts, found it hard to reproduce with verbal or pictorial product descriptions. Huber, Payne, and Puto published a rejoinder the same year defending the effect under proper conditions. The defensible read is that asymmetric dominance is a real cognitive effect that shows up cleanly with simple numeric tradeoffs and often does not survive the messiness of a real catalog or pricing page. It is a tool that sometimes works, not a law that always does. (Yang and Lynn, 2014)

The 1% price statistic: an identity, not a discovery

This is the most quoted number in pricing, and it is quoted with its assumption removed. Marn and Rosiello, working from the average economics of 2,463 companies in the Compustat aggregate, showed that improving price by 1% would raise operating profit by 11.1% for the average company, compared with 7.8% for a 1% cut in variable cost, 3.3% for a 1% gain in volume, and 2.3% for a 1% cut in fixed cost. Price ranks first because it flows straight to the bottom line with no cost attached. (Marn and Rosiello, 1992)

The result is arithmetic, and it is correct arithmetic. It also assumes each lever moves in isolation with everything else held constant. The 1% price improvement is computed as if you raise price 1% and sell exactly as many units as before. In reality, raising price usually sells fewer units, and how many fewer is the elasticity of your specific demand curve, which the statistic sets to zero. So "a 1% price increase beats a 1% volume increase" is true for a business facing no volume loss and progressively less true as demand gets more elastic. The number is a legitimate argument for taking pricing seriously. It is not a promise that any price increase pays, and using it that way inverts what the authors actually showed.

Anchoring: sturdy mechanism, under-tested application

Anchoring is one of the sturdiest findings in the judgment literature. Tversky and Kahneman (1974) demonstrated that people estimate unknown quantities by starting from whatever number is in front of them and adjusting, and that the adjustment is typically insufficient, so the final estimate stays biased toward the anchor. Their rigged wheel-of-fortune experiment moved estimates of an unrelated quantity (the share of African countries in the UN) simply by showing a high or low number first. (Tversky and Kahneman, 1974)

The pricing-page application, a deliberately expensive top tier that makes the tier below it feel reasonable, is a plausible extension of a very real mechanism. What is missing is controlled evidence of the size of that effect on actual purchase decisions, as opposed to on abstract number estimates. The lift figures attached to "add a high anchor tier" come from vendor case studies and A/B tests without disclosed methods, not from the anchoring literature itself. State that anchoring is real and that its specific dollar effect on your pricing page is a hypothesis to test, not a settled multiplier.

Value-based pricing: doctrine ahead of evidence

"Charge on value, not cost" is the most durable idea in pricing strategy, associated with Nagle's textbook and with Hinterhuber's work, and it is almost certainly directionally right. Cost-plus pricing ignores what a buyer will actually pay and leaves money on the table when your product is differentiated. The problem is the proof. The margin figures repeated for value-based pricing tend to be either survey associations or unsourced round numbers, and they run into an obvious confound: the firms capable of pricing on value are usually the firms with differentiated, hard-to-substitute products, and that differentiation is what earns the higher margin. The pricing method and the margin are both downstream of the same product advantage. (Hinterhuber, 2008)

That does not make value-based pricing wrong. It makes the causal claim, "switch to value-based pricing and your margin will rise by a specific amount," unproven. The defensible version is that pricing to value rather than cost is sound reasoning that removes an obvious leak, with the size of the gain specific to how differentiated your offer actually is.

Discounting: the training effect is real for consumer goods

The claim that discounting trains customers to wait has a strong academic anchor for at least one context. Mela, Gupta, and Lehmann analyzed 8.25 years of household panel data for a frequently purchased packaged good and found that consumers became more price- and promotion-sensitive over time, a shift they linked to manufacturers advertising less and retailers promoting more. In other words, sustained discounting reshaped how buyers responded to price, lowering their willingness to buy without a deal. (Mela, Gupta, Lehmann, 1997)

That is real evidence for the mechanism in consumer packaged goods. The broader claims stacked on top of it, that discounting "erodes the brand" universally and that some round percentage of promotions "fail," are directionally consistent with the reference-price literature but are either channel-specific or unsourced at the stated magnitude. The load-bearing, citable finding is narrower and stronger than the slogan: heavy promotion can raise long-run price sensitivity, measured over eight years in one category.

SaaS conventions: usage-based pricing and the annual discount

Two SaaS-specific claims are more convention and trend than law. On usage-based pricing, OpenView's tracking showed a genuine rise, with a majority of SaaS firms adopting some usage-based element by 2023, but the same coverage stressed that usage-based pricing is expanding alongside seat-based models rather than replacing them, and that most companies run hybrids. The "usage-based is beating per-seat" narrative leans heavily on a handful of outliers whose success is partly why the story spread, which is survivorship. The trend is real; the takeover is not. (OpenView, 2023; TechCrunch, 2023)

On the annual discount, "two months free, about 17% off" is a convention with a communication rationale, not an experimental result. "12 months for the price of 10" is easy to say and caps how much annual pricing erodes list value, and survey data put most SaaS annual discounts in the 15-20% band. No study identifies 16.7% as the profit-maximizing figure. It is a coordination point the market settled on, which is a fine reason to use it and a bad reason to call it evidence.

Primary versus inferred: what can go behind your name

Peer-reviewed findings or transparent datasets, safe to state with method attached

• A $9 price ending raised demand in all three of Anderson and Simester's field experiments; a clothing item sold more at $39 than at $34; the effect was stronger for new items and weaker with a "Sale" cue. State that it is field-experiment evidence from a mail-order retailer. (Anderson and Simester, 2003)
• The left-digit effect operates only when the leftmost digit changes ($2.99 vs $3.00, not $3.59 vs $3.60) and depends on distance to a competing price. (Thomas and Morwitz, 2005)
• Asymmetric dominance (the decoy effect) is a real violation of rational-choice regularity in controlled settings, and it is difficult to replicate with realistic product stimuli. State both halves. (Huber, Payne, Puto, 1982; Yang and Lynn, 2014)
• For the average of 2,463 companies (1992), a 1% price improvement lifted operating profit 11.1%, versus 3.3% for a 1% volume gain, holding all else constant. Always state "holding volume constant." (Marn and Rosiello, 1992)
• Anchoring biases numeric estimates toward an initial value via insufficient adjustment. State that the pricing-page application is an untested extension. (Tversky and Kahneman, 1974)
• Over 8.25 years in one packaged-goods category, consumers grew more price- and promotion-sensitive as promotion rose and advertising fell. Label it CPG, one category. (Mela, Gupta, Lehmann, 1997)
• A majority of SaaS firms had adopted some usage-based element by 2023, but usage-based pricing is expanding alongside, not replacing, seat-based models. (OpenView, 2023)

Inference or myth: do not present as an established fact

• "Always end prices in 9" as a universal rule (it is conditional and can backfire on premium positioning).
• "Put three tiers on the page and the middle wins" (the decoy lifts a built target, not "the middle," and is fragile with real products).
• "A price increase always pays more than chasing volume" (the 1% statistic assumes zero volume loss).
• "Value-based pricing raises profit by a specific percentage" (unproven causal magnitude; likely selection).
• "16.7% is the right annual discount" (a convention, not an optimum).
• "Free trials beat freemium," or the reverse, as a universal (context- and funnel-dependent).
• "A high anchor tier lifts the tier below it by X%" (anchoring is real; the pricing-page magnitude is not established).
• "94% of promotions fail," "value-based pricing earns 31% more," and round "this trick lifts revenue X%" figures (unsourced as stated).
• "Usage-based pricing has beaten per-seat" (rising, not replacing).

Dated timeline

Steelman: why "these tactics are basically real and you should just use them" might be closer to right than we allow

The strongest case against this report's caution has three honest parts.

1. The best-evidenced tactics are the mechanical ones, and mechanical effects travel. The 9-ending and the left-digit effect are not soft attitude shifts. They are cognitive regularities with field-sales evidence and a specified mechanism. A founder who ends new-item consumer prices in 9, absent competing discount signals, is acting on some of the cleanest evidence in behavioral pricing. Dismissing that as "conditional" undersells how reliable the condition is.

1. Fragile in the lab is not the same as useless in practice. The decoy effect fails to replicate cleanly with pictures and words, but real pricing pages are closer to the abstract two-number tradeoffs where it does show up, and a well-constructed decoy costs almost nothing to test. A tool that works sometimes, for free, is worth reaching for even if it is not a law.

1. The 1% statistic points at a real and common failure. Most founders under-price and under-invest in pricing. The assumption-laden statistic is directionally the right medicine even if the dose is overstated, because the modal error is leaving pricing power unused, not raising prices into an elastic wall.

The honest synthesis:

The pricing tactics with real experiments behind them are worth using, and they are worth using the way the experiments actually found them true: 9-endings on new consumer items, the left-digit effect where the dollar figure rolls over, a decoy only when you build it to lift a specific target, a high anchor as a hypothesis to A/B test, not a guaranteed lever. The failure mode is not using these tools. It is repeating the portable version of each one, the version with every condition sanded off, and then being surprised when the tactic that "always works" does nothing on your page. Test on your own prices. The evidence says the effect depends on your context, which is the one thing a statistic cannot carry for you.

The most authoritative sources

1. Anderson and Simester, "Effects of $9 Price Endings on Retail Sales" (Quantitative Marketing and Economics, 2003) - three field experiments with a real retailer; the primary evidence for charm pricing and its new-item and "Sale"-cue limits.
2. Thomas and Morwitz, "Penny Wise and Pound Foolish: The Left-Digit Effect in Price Cognition" (Journal of Consumer Research, 2005) - the cognitive mechanism under 9-endings and the left-digit-change condition.
3. Huber, Payne, Puto, "Adding Asymmetrically Dominated Alternatives" (Journal of Consumer Research, 1982) - the original decoy/attraction effect.
4. Yang and Lynn, "More Evidence Challenging the Robustness and Usefulness of the Attraction Effect" (Journal of Marketing Research, 2014) - the replication challenge that constrains the decoy claim; read alongside Frederick/Lee/Baskin (2014) and the Huber/Payne/Puto rejoinder.
5. Marn and Rosiello, "Managing Price, Gaining Profit" (Harvard Business Review, 1992) - origin of the 11.1% "1% price" statistic on 2,463 companies, with the all-else-constant assumption.
6. Tversky and Kahneman, "Judgment under Uncertainty: Heuristics and Biases" (Science, 1974) - the anchoring mechanism borrowed for pricing-page anchors.
7. Mela, Gupta, Lehmann, "The Long-Term Impact of Promotion and Advertising on Consumer Brand Choice" (Journal of Marketing Research, 1997) - the strongest evidence that discounting raises long-run price sensitivity, in one CPG category.
8. Ariely, Predictably Irrational (HarperCollins, 2008) - the Economist decoy demonstration (16/0/84, then 68/32); a hypothetical-choice teaching case, not a field test.
9. Hinterhuber, "Customer value-based pricing strategies: why companies resist" (Journal of Business Strategy, 2008) - representative of the value-based-pricing doctrine and the gap between its logic and its controlled evidence.
10. OpenView, State of Usage-Based Pricing (2023) and TechCrunch coverage, Feb 2023 - the usage-based trend, with the "rising, not replacing" qualification at the source.
11. ProfitWell/Paddle freemium benchmarks - transparent-ish platform data on freemium versus free-trial conversion, useful for direction with the funnel and denominator attached.

Open questions the primary sources do not answer

• Whether charm pricing (9-endings) helps or hurts in premium and B2B contexts, where a round number can signal quality. Anderson and Simester studied a mail-order retailer, not luxury or enterprise pricing.
• A controlled field test of the decoy effect on live SaaS or e-commerce pricing pages with real revenue, rather than paper choices or two-number lab tradeoffs.
• A causal, confound-controlled estimate of the profit lift from switching to value-based pricing, isolated from the product differentiation that usually accompanies it.
• The actual effect size of a high anchor tier on real purchase behavior on a pricing page, as opposed to on abstract number estimates.
• Whether the promotion-sensitivity "training" effect measured in packaged goods over 8 years generalizes to SaaS, services, and durable goods.
• A method-disclosed, apples-to-apples comparison of free-trial versus freemium on the same product and denominator, rather than benchmarks drawn from different funnels and populations.
• The profit-maximizing annual discount, tested rather than conventional, and how much it varies by category and buyer.

Verification pass: numbers to confirm before publish

Every figure below is either sourced above or explicitly flagged as not independently confirmed this session. Nothing here should be published as a verified fact without closing the gap noted.

Sourced and safe to publish with the stated attribution:

• Anderson and Simester: $9 ending raised demand in all three field experiments; $39 outsold $34; stronger for new items; weaker with "Sale" cue. Sourced to the 2003 paper's abstract and secondary summaries. The exact percentage demand lift (widely repeated as "about a third" for the $34-to-$39 case) comes from secondary retellings, because the source PDF is a scanned image and its precise in-text figures were not machine-extracted this session. Publish the direction and the $34/$39 result; do not attach a specific percentage without confirming it in the paper's tables.
• Marn and Rosiello: 11.1% (price), 7.8% (variable cost), 3.3% (volume), 2.3% (fixed cost), 2,463 companies, 1992. Confirmed across multiple independent summaries of the HBR article; confirm against the HBR original before quoting the four figures together.
• Ariely Economist split: 16/0/84 then 68/32 among 100 students each. Consistent across multiple summaries of Predictably Irrational; confirm against the book before quoting the exact counts.
• Thomas and Morwitz (left-digit change condition), Huber/Payne/Puto (1982), Tversky and Kahneman (1974), Mela/Gupta/Lehmann (8.25 years, one CPG category, rising price sensitivity): confirmed against journal abstracts and citations.

Could NOT independently source this session (do not publish as fact):

• "Value-based pricing firms earn 31% higher operating income" stays unsourced. The real, sourced figure is smaller. A second search could not trace the 31% to any auditable study, so it remains MYTH-grade. What is sourced: McKinsey estimates value-based pricing can improve return on sales by 5 to 10%, and Liozu and Hinterhuber's academic work finds a positive association between value-based pricing and firm performance, not a clean 31% causal lift. Publish the McKinsey 5-to-10% range with its name attached; do not publish the 31%.
• "94% of promotions fail to increase category value." Surfaced in retail-analytics content without a linked dataset. Not confirmed. Do not publish.
• The freemium and free-trial conversion percentages are now sourced to specific 2026 studies. ProfitWell's 2026 SaaS Monetization Index puts organic freemium-to-paid at about 2.6%, rising to about 5.1% with role-based feature gating. Opt-in free trials (no card) averaged about 8.9% and opt-out trials (card required) about 31.4% in 2026 benchmark data, and an OpenView and ChartMogul analysis of 3,200-plus SaaS companies put the overall average trial-to-paid near 24.8%. The direction the report stated holds; quote each figure with its own source, sample, and year rather than blending them into one portable number.
• The exact OpenView usage-based adoption percentages (variously reported as roughly a majority, ~46% hybrid, ~61% projected) shift by edition and definition. OpenView wound down as a firm, so confirm the specific figure against the specific report edition before quoting a number; publish the qualified trend ("rising, not replacing") rather than a precise share.
• The McKinsey restatement figures (the 1992 number is frequently re-presented in the 2003 "Power of Pricing" with a different company set and a lower percentage). The 2003 PDF was not machine-readable this session, so no restated figure is quoted here. Use the 1992 primary; do not cite a specific 2003 restated percentage without opening that source.
• Any specific "anchor tier lifts the lower tier by X%" pricing-page figure. No controlled source found. The anchoring mechanism is verified; the pricing-page magnitude is not.

The absence of a confirmable source is not evidence a number is false. It is evidence the number cannot responsibly be presented as verified. Where a figure resolved to a primary source, it is cited; where it did not, it is flagged here rather than laundered into the body.

Research notes and limitations

This report prioritizes peer-reviewed pricing and decision research for the psychological claims (9-endings, left-digit, decoy, anchoring, promotion) and reads the original studies or their journal abstracts directly. Its strongest sources are controlled experiments and long-panel analyses, which is a higher bar than the vendor benchmarks that dominate the SaaS claims (annual discounts, freemium versus free trial, usage-based pricing), where every figure is a platform's own book of business with its own denominator. Two named traps recur across the category: portability (a narrow, conditional lab effect restated as a universal rule) and the buried assumption (the "1% price" statistic holding volume constant). The report's central instruction to a founder is the one it applies to itself: read the condition and the assumption before the number, and test the tactic on your own prices rather than adopting someone else's optimum.