---
title: "The Decoy Effect, and Four More Pricing Claims Retested"
description: "The decoy effect produced nothing in 19 studies using real products. Charm pricing shrank as samples grew. What pricing psychology actually supports."
url: https://fifteenthmeridian.com/blog/pricing-psychology-evidence-2026
---
[Back to Blog](https://fifteenthmeridian.com/blog)

Pricing · Evidence

September 18, 2026

16 min read

By [Bradyn Gardiyash](https://fifteenthmeridian.com/authors/bradyn-gardiyash)

# The Decoy Effect, and Four More Pricing Claims Retested

**The short answer.** The decoy effect, the most repeated pricing trick in marketing, did not appear in the predicted direction in any of nineteen studies that used real products rather than numbers on a page, though later work using photographs did find a small one. Charm pricing is real, and its measured lift shrank steadily as the experiments got bigger, down to a figure the largest study could not separate from zero. The "Sale" sign beat it. None of these effects are fake. All of them are narrower, more conditional and more context-bound than the version that reaches a pricing meeting. The last section is about the one that can get you a call from the Competition Bureau.

This is the sixth post in our evidence series. Same standard as the others: every claim traced to its primary source, sample sizes attached, and the limits stated even when they cut against the story.

Two disclosures before anything else. Two of the papers here sit behind paywalls we did not get through, so where we cite Thomas and Morwitz (2005) or Thomas, Simon and Kadiyali (2010) we describe only what their abstracts and the surrounding literature establish, and we print no sample size, test statistic or effect size from either. And most of this evidence predates preregistration becoming normal practice, which matters more for the lab studies than the field ones.

## The decoy effect does not survive real products

The decoy, or attraction effect, is the one everyone knows. Offer a good option and a better option, add a third that is obviously worse than the better one, and choice shifts toward the option you wanted to sell. It is in every pricing deck and the opening chapter of a very famous behavioural economics book.

It was first documented by Huber, Payne and Puto (1982) in the Journal of Consumer Research, and it is real in the setting they used. What happened next is the interesting part.

Frederick, Lee and Baskin ran 38 studies (2014, Journal of Marketing Research 51(4), 487-507). Of the five conditions presenting options as abstract numeric matrices, four produced a significant attraction effect. Across nineteen separate studies using natural stimuli, things a person could actually see, taste or touch, not one did.

11 of 91

Attempts that produced a reliable attraction effect, across 23 product classes and 73 decoyed choice sets, far fewer than the studies were powered to produce (Yang and Lynn, Journal of Marketing Research, 2014)

Their own summary is that attraction effects appeared "when stimuli were represented numerically, but not otherwise." Two of their studies show the switch directly. In one (N=791), a decoy moved the target gamble's share from 21% to 37% when the win probability was printed as a number, and from 34% to 35% when the identical probability was shown as a picture. In another (N=240, televisions), a decoy raised the target's share from 33% to 57% when picture quality was a numeric rating, and moved it from 53% to 35% when the sets could be seen. How the attribute was represented significantly changed the result (interaction b=-1.71, p<.01). The apparent reversal on its own did not reach significance (chi-square 3.37, p=.07), so read it as the effect failing rather than as a reverse tactic you could use.

Yang and Lynn (2014, Journal of Marketing Research 51(4), 508-513) made 91 attempts across 23 product classes and 73 decoyed choice sets. Eleven produced a reliable effect, far fewer than the studies were powered to produce.

And the canonical subscription-pricing story, the one with the print-only option nobody is supposed to choose, traces back to a classroom demonstration with about 29 and 30 students. Frederick, Lee and Baskin tried to replicate it twice at much larger samples and did not reproduce it. They write that they "failed to replicate several of the results most frequently cited as evidence."

### What the original authors said about it

The strongest sentence in this literature belongs to the people who started it. Huber, Payne and Puto replied in the same 2014 issue (Journal of Marketing Research 51(4), 520-525) and were candid about what their paper had been: "our 1982 article was designed as a demonstration study," and "We did not set out to suggest a tool for marketing practice."

They also described how the original stimuli were arrived at, gradually identifying the choice sets that produced the effect. That is a normal way to run a demonstration and a poor way to establish a reliable commercial lever, and it explains the pattern that followed.

Being fair to the disagreement: this is contested rather than closed. Simonson published a commentary in the same issue defending a narrower version, the original authors replied, and the exchange is worth reading before anyone declares the matter finished. What is not defensible is presenting the decoy as a dependable way to move buyers between real products they can see.

And the critics do not have the last word. Dumbalska and colleagues (2020, PNAS 117(40), 25169-25178) ran incentive-compatible property choices using photographs, N=233, and found a significant attraction effect, t(188)=4.74, p<0.001, worth roughly 3 to 5 percentage points of preference. A 2025 analysis of about 3.6 million UK wine transactions (Devine and colleagues, npj Science of Learning) detected one near 1%. So the effect is not confined to numbers on a page. It is just far smaller in the wild than the version that gets taught.

## Charm pricing is real, and it got smaller every time it was measured better

The evidence for prices ending in 9 is better than the decoy's, because it comes from randomized field experiments rather than lab choices. Anderson and Simester (2003, Quantitative Marketing and Economics 1(1), 93-110) ran three of them with two US mail-order women's clothing catalogues, average item price around $50. Worth noting before anyone applies it: these were whole-dollar endings, $39 against $44, not the $X.99 endings people usually mean.

The 9 ending increased demand in all three. Here is the part that never travels with the finding:

The same effect, measured three times

Study 1, 60,000 customers about 35%

Study 2, 62,500 customers about 15%

Study 3, 270,000 customers, items with no Sale cue about 7%

One paper, one team, one method, and the measured lift falls as the sample grows. That is the ordinary signature of an early estimate that was too generous, and the figure the marketing industry quotes is almost always the one from the smallest study. We are not going to invert the cherry-pick and tell you the real number is 7%, because in the specification without price that estimate is not statistically significant either. The honest version is the range and the direction of travel.

Two more conditions from the same paper. The lift concentrated on items new to the catalogue, where the significant results sit; on established items the estimates were smaller and not statistically significant, and for footwear the coefficient was negative and not significant. Not significant means the study could not tell, which is not the same as showing the effect is absent, so read that as "unproven outside new items" rather than "proven not to work."

And a "Sale" cue on the item had a larger coefficient than the 9 ending did. If you are choosing between telling someone a price ends in 9 and telling them it is on sale, the paper points at the sign.

The authors' own explanation is a signal rather than a quirk of arithmetic: a 9 ending tells a customer who has no reference point that this is the cheap end. They explicitly reject the popular left-digit account, partly because $9.50 endings did not behave like $9 endings. Their closing note is that what remains largely unexplained is "Why 9?"

Elsewhere the picture stays mixed. Schindler and Kibarian (1996, Journal of Retailing 72(2), 187-199) mailed 90,000 catalogues and found the 99-ending version produced 8.0% more sales volume than the 00-ending version, but the number of purchasers was not significantly different, and an 88 ending produced slightly less volume than 00. Ngobo, Legoherel and Gueguen (2010, Journal of Retailing and Consumer Services 17, 374-385) looked at more than 11,000 grocery SKUs and found nine-endings did not influence choice on their own, with the effect depending on the category.

## The left-digit effect is real, and the famous paper is not about sales

The mechanism people reach for to explain charm pricing is the left digit: $2.99 reads as "two something" rather than "almost three." The citation is Thomas and Morwitz (2005, Journal of Consumer Research 32(1), 54-64), five laboratory experiments.

Two things to keep straight. First, it is a magnitude-judgment study rather than a purchase study, so it is not the right citation for a claim about sales or conversions. Second, the paper's actual contribution is a boundary condition that gets dropped in the retelling: the effect shows up when the leftmost digit changes, and not when it does not. $2.99 against $3.00 produces it. $3.59 against $3.60 does not. That single rule disposes of most of the pricing advice built on this paper.

The strongest commercial evidence is elsewhere and more recent. Strulov-Shlain (2023, Review of Economic Studies 90(5), 2612-2645) analysed roughly 78 million store-product-week observations of retail scanner data and found demand drops by about 6 to 7% on average when the dollar digit changes, for example moving from $2.99 to $3.00. That is observational rather than experimental, so it describes what happened rather than isolating a cause.

The useful half of that paper is not the bias. It is that retailers underuse it. Consumers behave as though a penny over a 99 ending costs far more than a penny, while firms set prices as if the effect were almost nothing, and the paper estimates that mismatch costs a meaningful share of gross profit. The interesting finding is not that shoppers are irrational. It is that sellers are leaving money on the table by not taking the irrationality seriously enough.

One more honest note. A 2022 null result in PLOS ONE (Fenneman and colleagues) did not find the effect, but its design showed one price at a time with no reference price, which is the condition where later work (Sokolova, Seenivasan and Thomas, 2020, Journal of Marketing Research 57(4), 771-788) would predict the weakest result. We mention it because leaving out an inconvenient null is exactly the behaviour this series exists to criticize.

## Precise prices work, until they do not

Precise prices, $395,425 rather than $395,000, are judged as smaller magnitudes than round numbers, and buyers have been shown to pay more against a precise asking price. The source is Thomas, Simon and Kadiyali (2010, Marketing Science 29(1), 175-190), five studies including a nationally representative homeowner sample and an analysis of real-estate transactions.

It replicates, and it is small. A preregistered pair of experiments (Leib, Kee, Loschelder and Roskes, 2022, Journal of Experimental Social Psychology, N=939) reproduced it on a house counteroffer task, with the gap between asking price and buyer counteroffer around $31,300 against a round asking price and around $25,600 against a precise one.

Then it reverses, twice, in ways that matter more than the effect itself. Loschelder and colleagues (2016, Psychological Science 27(12), 1573-1587) tested 1,320 experts and amateurs across real estate, jewellery and cars and found too much precision backfires with experts. Leib and colleagues (2021, Management Science 67(2), 1048-1055, preregistered, n=1,809) found precise prices raise counteroffers in a buyer's market and lower them in a seller's market.

An effect that flips sign depending on who is reading it and which way the market is running is not a rule you apply. It is a variable you test.

## The struck-through price, and the part with legal exposure

This section is about Canada, and it is the only one here where getting it wrong costs more than a bad quarter. It is general information rather than legal advice, and the authority is the Competition Bureau's own guidance on ordinary price claims, which is worth reading in full rather than through anyone's paraphrase.

The research first. Compeau and Grewal's review of 38 studies (1998, Journal of Public Policy and Marketing 17(2), 257-273) found that an advertised reference price raises perceived value and perceived savings, and a meta-analysis of 345 observations across 20 articles (Krishna, Briesch, Lehmann and Yuan, 2002, Journal of Retailing 78(2), 101-118) is consistent with that. What no source we found produces is a percentage lift in actual sales from a struck-through price. The measured outcomes are perception, not revenue.

Urbany, Bearden and Weilbaker (1988, Journal of Consumer Research 15(1), 95-110) found that an exaggerated reference price moved perceptions much as a plausible one did. We are reporting that as the reason the rules exist, not as a tactic, because in Canada acting on it is reviewable conduct.

In Canada an ordinary price claim is regulated under the Competition Act, and the tests turn on how much you actually sold at that price and how long you genuinely offered it. Both can be satisfied by sales or offers made after the claim as well as before, which surprises people who assume only past sales count. We are deliberately not paraphrasing the thresholds here, because the wording matters and the version that circulates on marketing blogs is not reliably the version in the legislation. Read the Competition Bureau's own guidance on ordinary price claims, and if real money rides on it, ask a lawyer rather than an agency.

Enforcement is not theoretical. As announced by the Competition Bureau and covered in the press, the Competition Tribunal found against Sears Canada on tire pricing in January 2005, with a $100,000 penalty plus $387,000 in costs, and Amazon.com.ca reached a $1,000,000 settlement plus $100,000 in costs in January 2017 over comparison prices. Provincial consumer protection legislation in Alberta and Ontario also addresses misleading price representations. If you sell into the United States as well, the rules there are different and are not covered here.

The practical framing is not "how do I structure a discount." It is that your reference price is a factual claim about your own sales history, and the standard your price data has to meet is written down.

## What we are not saying

**None of these effects are fake.** Anderson and Simester found a lift in all three field experiments. Reference prices genuinely move perceived value. The left-digit distortion shows up at enormous scale in scanner data. "Debunked" and "myth" are not available words here, and we are not using them.

**Small effects are still money.** A couple of points on a grocery category, or a few points of gross profit given away by mis-set price endings, are real revenue at scale. The target of this post is the claim that these are large, reliable levers you can pull on any product, not the claim that they exist.

**The decoy argument is live.** A peer-reviewed commentary defending a narrower version was published alongside the critiques, and the original authors replied. We think the weight of evidence favours the critics for practical purposes. We do not think the question is closed.

**We cannot run a file-drawer argument.** One of the critical papers looked for publication bias in the attraction-effect literature and did not find a strong signal, so we are not claiming the positive results are simply the surviving half of a bigger pile.

**Almost none of this is about the work we do.** The evidence base here is catalogues, groceries, gambles, televisions and residential real estate. There is no published field evidence in this set for professional services, software pricing, B2B, or an agency retainer, which is most of what we and our clients actually sell. We are not going to pretend a catalogue result transfers. For that context the honest tool is a test on your own numbers, and a clear sense of what you would have to see to believe it.

## What this changes in practice

Less than the original claims promised, and more than nothing.

**Do not build a pricing tier around a decoy.** The evidence that it works is concentrated in abstract numeric comparisons, and where it does show up with real products it is small. Somewhere between nothing and a couple of points is the honest expectation, which is not enough to design a price list around.

**Treat a 9 ending as a cheapness signal, not a hack.** It says "this is the affordable end" to someone who has no reference price. That means it is doing its most plausible work on something new, and it may be doing nothing on an item your customer already knows the price of. It also means a 9 ending on a premium product is arguing against your own positioning.

**If the message is that it is on sale, say it is on sale.** In the one experiment that put them side by side, the sale cue carried more weight than the price ending.

**Do not change the cents and expect a result.** The left-digit boundary is the dollar digit. Moving $3.59 to $3.55 is a discount, not a psychological price.

**Make your reference prices survive an audit.** Not because the tactic fails, but because in Canada the "was" price is a claim about your sales records, and the test is public.

## How we checked this

Every citation in this post was traced to its journal, volume, issue and page range before it was written, and two of them are flagged above as abstract-level reads. A few widely circulated figures did not survive that process and are deliberately absent: a popular meta-analytic number for the decoy effect comes from an unpublished dissertation, and several precise-sounding statistics attached to the two paywalled papers could not be traced to them, including one frequently repeated example price that the source of record says is not in the paper at all.

We would rather publish a post with fewer numbers in it than one with numbers we cannot point at. If you find something here that does not hold, tell us and we will correct it in public, the way the previous post in this series corrected two unsourced claims of our own.

## Frequently asked questions

### Does the decoy effect actually work?

The decoy effect works reliably in a narrow setting and mostly does not outside it. Across 38 studies by Frederick, Lee and Baskin (Journal of Marketing Research, 2014), attraction effects appeared when options were presented as abstract numbers, while nineteen studies using real products a person could see produced none, and in one television study the effect reversed. The widely told subscription-pricing example traces to a classroom demonstration with about 30 students and did not replicate at larger samples.

### Do prices ending in 9 increase sales?

Prices ending in 9 did increase demand in all three randomized field experiments by Anderson and Simester (2003), but the measured lift fell from about 35% in the smallest study to about 7% in the largest, and it concentrated on items customers had not seen before. In the same research, a "Sale" cue carried more weight than the price ending. The experiments used whole-dollar endings such as $39 against $44, in mail-order clothing catalogues, so treat the finding as conditional rather than a general rule.

### Is the left-digit effect real?

The left-digit effect is real and has a boundary that matters: it applies when the leftmost digit changes, so $2.99 against $3.00 produces it while $3.59 against $3.60 does not. The original paper by Thomas and Morwitz (2005) measured how large a price seemed rather than whether anyone bought, so it should not be cited as evidence of a sales lift. Retail scanner analysis by Strulov-Shlain (2023), covering roughly 78 million store-product-week observations, found demand drops about 6 to 7% on average when the dollar digit changes.

### Should I use precise prices instead of round numbers?

Precise prices are judged as smaller magnitudes than round ones, and the effect replicates, but it is small and it reverses in two documented situations: experts respond badly to too much precision, and precise prices push counteroffers the other way in a seller's market compared with a buyer's market. Because the direction depends on your audience and your market conditions, precision is worth testing on your own pricing rather than adopting as a rule.

### What are the rules for showing a struck-through "was" price in Canada?

In Canada a struck-through price is a regulated claim under the Competition Act rather than a design choice, and it rests on your actual sales history: how much of that product genuinely sold at the higher price, and how long it was genuinely offered there. Both can count sales and offers made after the claim as well as before. The thresholds and their exact wording are set out in the Competition Bureau's guidance on ordinary price claims, which is the document to read rather than any marketing summary of it, including this one. This is general information, not legal advice.

## Related reading

- [Behavioural science and the replication crisis](https://fifteenthmeridian.com/blog/behavioural-science-replication-crisis-2026), where anchoring survived and several famous effects did not.
- [Colour psychology, fact-checked](https://fifteenthmeridian.com/blog/colour-psychology-evidence-2026), another set of claims whose famous numbers had no study behind them.
- [Five marketing myths peer-reviewed research says are wrong](https://fifteenthmeridian.com/blog/marketing-myths-peer-reviewed-2026), the first post in this series.

Evidence series

## We would rather test it than quote it.

There is no published field evidence for most of what gets sold as pricing psychology in professional services. If you want to know whether a price change moved anything, the answer comes from your own numbers, and we will tell you what it would take to find out.

[Talk to us](https://fifteenthmeridian.com/contact)