You priced something at ninety-nine rather than a hundred. Somebody told you it would sell better, and somebody else told you it makes you look cheap. There is a decent body of field evidence on the first claim and a recent meta-analytic null on the second.
Key Takeaway
Three field experiments in which price endings were experimentally manipulated found that a $9 ending increased demand in all three, that the increase was stronger for new items than for items sold in previous years, and that there is some evidence nine endings are less effective when retailers use "Sale" cues. The authors conclude nine endings may be more effective when customers have limited information, which helps explain why retailers do not use them on every item[1]. Separately, a source reports that a recent meta-analysis found an overall null effect for the quality impairment of nine-ending prices[5].
Our Grades For These Claims
Applying the scheme from the first article in this series.
That nine endings increase demand is Grade B, edging toward A. It rests on three field experiments with experimental manipulation of real prices at a real retailer, which is a strong design. It falls short of Grade A only because it is one research team at one retailer, and we did not locate independent replication.
That the effect is stronger for new items is Grade B, from the same experiments.
That nine endings are weaker alongside sale cues is Grade C, and the authors themselves describe it as some evidence rather than a firm finding[1].
That nine endings damage quality perceptions is Grade D. A source reports a recent meta-analysis finding an overall null, with even some opposite findings[5].
The practical translation: the demand effect is the better-supported claim, and the brand-damage objection is the weaker one, which is close to the reverse of how the two are usually weighted in commercial argument.
A Note On Method
Everything here is verified to August 2026.
We obtained the published abstract of the field experiment paper from six independent sources reproducing it identically, including the publisher, the authors' own institution, and an economics working paper repository[1][2][3][7].
We did not obtain the paper in full and therefore state no effect size, no sample size and no lift percentage. That is a significant limitation for an article about a commercial tactic, and we return to it.
The meta-analytic null on quality perceptions reaches us through third-party commentary attached to a publisher record, not from the meta-analysis itself, and we did not identify which meta-analysis is meant[5].
We did not obtain the left-digit paper, the price-image paper or the sale-signs paper, and cite each only by citation or as a secondary source describes it.
All arithmetic on breakeven volume is our own, clearly marked, and rests on assumptions we state.
This article reviews consumer research. It is not pricing, marketing or legal advice. Price presentation to consumers is governed by consumer protection and competition law, including rules on the advertising of sale prices, and this article does not address it.
Widespread And Unevidenced
Where the research started, and the opening line is worth quoting.
"Although the use of $9 price endings is widespread amongst US retailers there is little evidence of their effectiveness."[1]
Three observations, ours.
That sentence describes a practice adopted at enormous scale on essentially no evidence, which this series has now encountered in brainstorming, in the feedback sandwich, in the jam study's derived doctrine and in growth mindset training. The pattern is consistent enough to be worth naming as a default expectation.
It also means the study was testing a convention rather than a theory. Nobody was defending a hypothesis; everybody was already doing it.
And the practice was old. Nine endings long predate the research, which makes the study a case of evidence catching up with behaviour rather than the reverse.
Three Field Experiments
The design.
Anderson and Simester published Effects of $9 Price Endings on Retail Sales: Evidence from Field Experiments in Quantitative Marketing and Economics, 1(1), 93–110, in March 2003[1][3].
They present a series of three field-studies in which price endings were experimentally manipulated[1]. A related description of the authors' broader work refers to historical data from a women's clothing catalog, a field study in that catalog, survey responses to catalog stimuli, and grocery store data[4].
Three features make this strong evidence, and this assessment is ours.
Prices were experimentally manipulated, not observed. That is the difference between an experiment and a correlation, and it is what allows a causal reading.
It was a real retailer selling to real customers with real money, which removes the laboratory objection this series has raised repeatedly.
And there were three of them, which is internal replication before anybody else got involved.
Demand Rose Every Time
The headline finding.
"First, use of a $9 price ending increased demand in all three experiments."[1]
Two observations, ours.
All three is the operative phrase. Not on average, not in two of three. The direction was consistent across every experiment they ran.
And we must be explicit about what we cannot tell you. We did not obtain the magnitude. This article states no lift percentage because we do not have one, and anybody who needs to know how large the effect was should obtain the paper. A tactic's worth depends entirely on its size relative to the margin it costs, which is the subject of the arithmetic later in this article.
Stronger On New Items
The first moderator, and it is the more interesting of the two.
"Second, the increase in demand was stronger for new items than for items that the retailer had sold in previous years."[1]
Three consequences, ours.
For an item a customer has bought before, they have a prior sense of what it should cost. The price ending has to compete with that knowledge.
For a new item they do not, so the price itself carries more of the informational load.
And it means the tactic is not uniformly available. A business selling the same things to the same customers year after year is in the weaker case; a business launching new lines is in the stronger one.
Weaker Alongside Sale Signs
The second moderator, reported by the authors with appropriate hedging.
"There is also some evidence that $9 price endings are less effective when retailers use 'Sale' cues."[1]
Note some evidence. The authors did not present this at the same confidence as the first two findings, and neither do we.
Two observations, ours.
If it holds, the two devices are substitutes rather than complements. A sale sign and a nine ending are both telling the customer the same thing, and the second adds little once the first has been said.
Which would mean the common practice of putting a nine ending on a sale price is paying for something you have already bought. You have taken the margin hit twice and collected the signal once.
We flag firmly that this is our inference from a finding the authors themselves hedge, and it is Grade C accordingly.
The Authors' Own Explanation
How the researchers tie the results together, and it is the most useful sentence in the abstract.
"Together, these results suggest that $9-endings may be more effective when customers have limited information, which may in turn help to explain why retailers do not use $9 price endings on every item."[1]
Three observations, ours.
The explanation unifies both moderators. New items mean less information. Sale cues supply information. Both point the same way.
It implies the mechanism is inferential rather than perceptual. The customer is not misreading $99 as being much less than $100; they are treating the ending as a signal about what kind of price this is, and signals matter most where nothing else is available.
And it makes a testable prediction about your own business: nine endings should do most for products your customers cannot independently value, and least for products they buy repeatedly and know the price of.
The Detail That Confirms It
A small observation in the abstract that we think is the strongest part of the paper. This section is our own analysis.
The authors note their explanation helps to explain why retailers do not use $9 price endings on every item[1].
Consider what that observation does.
If nine endings simply increased demand everywhere, the rational strategy would be to apply them universally, and every retailer would. They do not, and one source records that analysis of rightmost digits confirms an overrepresentation of the digits 0, 5 and 9 rather than dominance by 9 alone[4].
Two consequences.
The observed behaviour of retailers is itself evidence for a conditional effect. Practitioners with vastly more pricing data than any researcher have collectively declined to use the tactic universally, which is consistent with it having conditions.
And it is a nice example of a theory explaining something outside the data it was built on, which is a stronger position than merely fitting the experiments.
The Claim About Cheapening Your Brand
The standard objection, and where it comes from.
A source describing this literature summarises the position: because nine-ending prices are frequently used in low-price segments, consumers may have learned to implicitly associate them with a better price value, but impaired product quality compared to round, zero-ending prices, an idea it labels image effects[5].
A separate paper on the topic is cited as Image Communicated by the Use of 99 Endings in Advertised Prices[6]. We did not obtain it.
Two observations, ours.
The reasoning is coherent and widely believed. It is the reason premium brands typically price at round numbers, and the advice is given confidently by a great many marketing professionals.
And it is a claim about a different outcome from the field experiments. Those measured demand. This concerns perceived quality, which is a separate construct that may or may not translate into purchasing.
And The Meta-Analytic Null
The finding that undercuts it, reported with a clear statement of our sourcing limits.
The same source continues: "Despite extant research documenting such image effects, a recent meta-analysis reports an overall null effect for the quality impairment of 9-ending prices, with even some opposite findings."[5]
Four things about how we are treating this, all ours.
It comes from third-party commentary attached to a publisher record, not from the meta-analysis itself.
We did not identify which meta-analysis is meant, obtain it, or verify the characterisation.
The phrase with even some opposite findings suggests heterogeneity of the kind this series has now encountered in choice overload, in conflict research and in nudging, where a null average conceals populations moving in different directions.
And on that sourcing we grade the quality-impairment claim D, meaning reportable as a contested idea and not usable as a basis for a pricing decision in either direction.
Our own position: we cannot tell you that nine endings are safe for a premium brand, and neither can anyone else on the evidence we located. What we can say is that the confident version of the objection is less well supported than its confident delivery implies.
What A Nine Ending Costs
The arithmetic nobody does. Everything in this section and the next two is our own calculation.
A nine ending is not free. Moving from $100 to $99 is a one percent price cut, and it comes straight out of margin.
So the tactic pays only if the demand lift exceeds the volume needed to replace the lost margin. That breakeven is computable.
The formula is simply the old unit margin divided by the new unit margin, minus one.
Two assumptions we are making, stated so you can vary them. Unit cost is unchanged by volume, and fixed costs are ignored. Both are simplifications, and in a real business the second in particular can push the answer either way.
The Breakeven Table
What the demand lift has to clear, on a $100 item cut to $99, at various gross margins. Our own arithmetic.
At 70 percent margin: unit margin falls from $70.00 to $69.00, requiring a 1.45 percent volume lift.
At 50 percent: from $50.00 to $49.00, requiring 2.04 percent.
At 35 percent: from $35.00 to $34.00, requiring 2.94 percent.
At 25 percent: from $25.00 to $24.00, requiring 4.17 percent.
At 15 percent: from $15.00 to $14.00, requiring 7.14 percent.
At 10 percent: from $10.00 to $9.00, requiring 11.11 percent.
Two observations, ours.
These are the hurdles the field experiments' effect size would have to clear, and we do not know that effect size, which is the honest limit of this article.
But the table is usable regardless, because you can test it yourself and compare against your own row.
The Margin Asymmetry
What that table implies, and it is counterintuitive. Our own analysis.
At 70 percent margin, a one percent price cut needs only a 1.45 percent volume lift to pay for itself.
At 10 percent margin, the same cut needs an 11.11 percent lift, roughly eight times as much.
Three consequences.
The tactic is far cheaper for high-margin businesses, because a fixed price cut consumes a much smaller share of a fat margin.
Which is close to the opposite of who typically uses it. Nine endings are associated with discount retail, which is the low-margin end, where the hurdle is highest.
And it means the same tactic can be obviously worthwhile for a professional services firm and marginal for a grocer, on identical demand effects, purely because of the margin structure underneath.
We are not claiming discount retailers are wrong. They have vastly more pricing data than we do, and the arithmetic above ignores everything except gross margin.
And The Price Point Asymmetry
A second asymmetry, also ours.
The cut is a fixed dollar amount but a variable percentage.
$20 to $19 is a 5.00 percent cut.
$50 to $49 is 2.00 percent.
$100 to $99 is 1.00 percent.
$500 to $499 is 0.20 percent.
$2,000 to $1,999 is 0.05 percent.
Two consequences.
A nine ending is nearly free at high price points and expensive at low ones. At two thousand dollars the tactic costs five hundredths of a percent of revenue.
Combine both asymmetries and the case is strongest for a high-margin business selling at a high price point, which describes professional services considerably better than it describes retail.
The Professional Services Case
Applying it to this publication's usual reader. This section is our own reasoning and is not tested by anything cited.
Score a professional services firm against the conditions.
High margin, so the breakeven hurdle is low.
High price point, so the percentage cost is trivial.
Customers frequently have limited information, which the authors identify as the condition under which the effect is strongest. A client rarely knows what a corporate reorganisation or an audit should cost.
Against that, three reasons for caution, also ours.
Professional fees are often negotiated or scoped rather than listed, so there may be no price ending to manipulate.
The quality-signalling objection, though poorly evidenced, is a real commercial consideration in a market where perceived expertise is the product.
And nothing in the cited research studied professional services. The field experiments concern catalogue retail and grocery. This whole section is an extrapolation and should be treated as one.
A Note On Sale Signs Themselves
A finding from the same researchers, worth recording briefly.
A description of the authors' related work states: "Sale signs increase demand. The apparent effectiveness of this simple strategy is surprising; sale signs are inexpensive to produce and stores generally make no commitment when using them."[4] The associated paper is cited as Are Sale Signs Less Effective When More Products Have Them?[2].
We did not obtain that paper and report only this description.
Two observations, ours.
The title's question implies the answer is yes, they dilute, which would mean sale signage has a budget rather than being free.
And that would fit the same informational account. A sale sign works by distinguishing an item from the others; if everything has one, nothing is distinguished. We flag that this is our inference from a title, which is thin ground, and we are not asserting it as a finding.
Why Zero, Five And Nine
A descriptive observation about actual pricing practice.
A source records that analysis of the rightmost digits of selling prices in a sample of retail price advertisements confirmed past findings indicating the overrepresentation of the digits 0, 5 and 9[4].
Two observations, ours.
Three digits dominate, not one. Zero and five are round-number signals and nine is the discount signal, which suggests retailers are choosing between at least two deliberate registers rather than defaulting to one.
And it supports the conditional reading throughout this article. If the practice were unconditionally superior, the distribution would be far more skewed toward nine than it apparently is.
What To Do
Compute your own breakeven before adopting it. Old unit margin divided by new unit margin, minus one, gives the volume lift the tactic must clear. It is one line of arithmetic.
Note that the hurdle is far lower at high margin. At 70 percent margin a one percent cut needs a 1.45 percent lift; at 10 percent it needs 11.11 percent.
And far lower at high price points. A dollar off two thousand is five hundredths of a percent.
Use it where the customer cannot independently value the item. The authors' own explanation is that nine endings work best where customers have limited information.
Expect less from it on items customers buy repeatedly. The effect was stronger for new items than for those the retailer had sold in previous years.
Do not stack it on a sale price. On the authors' hedged finding, the two devices appear to substitute rather than add, so you may be paying twice for one signal.
Do not accept the brand-damage objection as settled. A source reports a recent meta-analysis finding an overall null on quality impairment, with some opposite findings.
Do not accept the demand benefit as settled either. We could not obtain the effect size, which means neither we nor a consultant quoting this study can tell you whether it clears your breakeven.
Test it. Price endings are among the few interventions a business can vary cheaply by product, channel or period, and your own result beats any published one for your purposes.
Check the legal position on any sale claim. Advertising a price as reduced carries obligations this article does not address.
The Limits Of This Analysis
Several caveats matter. This article reviews consumer research and is not pricing, marketing or legal advice; price presentation to consumers is governed by consumer protection and competition law, including rules on advertising sale prices, which are not addressed here. Everything is verified to August 2026. We did not obtain the field experiment paper in full and therefore state no effect size, no sample size and no lift percentage, which is a material limitation in an article about a commercial tactic; anyone whose decision depends on the magnitude should obtain the paper. The evidence is one research team at what appears to be one catalogue retailer plus grocery data, and we located no independent replication. The meta-analytic null on quality impairment reaches us through third-party commentary attached to a publisher record; we did not identify which meta-analysis is meant, did not obtain it, and did not verify the characterisation. We did not obtain the left-digit paper, the price-image paper or the sale-signs paper, and our reading of the sale-signs title is an inference from a title rather than a finding. All breakeven arithmetic is our own, assumes unit cost is unchanged by volume, ignores fixed costs, and is illustrative. Nothing cited studied professional services, and that entire section is our own extrapolation from retail and grocery evidence. The margin and price-point asymmetries, the substitution reading of sale cues, and the interpretation of digit distribution are our own reasoning, not findings. Roughly twenty-three years of subsequent literature was not reviewed.
Frequently Asked Questions
Do nine endings actually increase sales?
When do they work best?
Will it make my brand look cheap?
What does it cost me?
Should a professional services firm use one?
What is the single best thing to do?
References
- Anderson, E. T., & Simester, D. I. (2003). Effects of $9 Price Endings on Retail Sales: Evidence from Field Experiments. Quantitative Marketing and Economics, 1(1), 93–110. DOI 10.1023/A:1023581927405, published abstract, on the use of $9 price endings being widespread amongst US retailers while there is little evidence of their effectiveness; on the authors presenting a series of three field-studies in which price endings were experimentally manipulated; on the first conclusion that use of a $9 price ending increased demand in all three experiments; on the second conclusion that the increase in demand was stronger for new items than for items the retailer had sold in previous years; on there being some evidence that $9 price endings are less effective when retailers use "Sale" cues; and on the results together suggesting $9-endings may be more effective when customers have limited information, which may help explain why retailers do not use them on every item. Note: we obtained the published abstract only. No effect size, sample size or lift percentage is stated anywhere in this article because we do not have one. link.springer.com
- Kellogg School of Management, Northwestern University, research record for Anderson and Simester (2003), reproducing the abstract and confirming the citation; and the same publisher record listing Anderson, E. T., and Simester, D. I. (2001), Are Sale Signs Less Effective When More Products Have Them?, Marketing Science, 20(2), 121–142. Note: the lead author's own institution; we did not obtain the 2001 sale signs paper. kellogg.northwestern.edu
- Ovid publisher record for Anderson and Simester (2003), Quantitative Marketing & Economics, 1(1), 93–110, March 2003, reproducing the abstract identically. Note: a publisher record, used to corroborate the abstract text and citation independently of reference 1. ovid.com
- Semantic Scholar record for Anderson and Simester (2003) with associated citation summaries, on the authors having analysed data from a variety of sources including historical data from a women's clothing catalog, a field study in that catalog, survey responses to catalog stimuli, and grocery store data; on sale signs increasing demand, with the apparent effectiveness of this simple strategy being surprising given that sale signs are inexpensive to produce and stores generally make no commitment when using them; and on analysis of the rightmost digits of selling prices in a sample of retail price advertisements confirming past findings indicating the overrepresentation of the digits 0, 5 and 9, attributed to Schindler and Kirby. Note: a bibliographic record with third-party summaries of related papers, none of which we obtained. semanticscholar.org
- ResearchGate record for Anderson and Simester (2003) with associated third-party commentary by Petrowsky and Loschelder, on 9-ending prices being frequently used in low-price segments such that consumers may have learned to implicitly associate them with better price value but impaired product quality compared to round, 0-ending prices, described as image effects; and on a recent meta-analysis reporting an overall null effect for the quality impairment of 9-ending prices, with even some opposite findings. Note: third-party commentary attached to a publisher record, not peer-reviewed in itself. We did not identify which meta-analysis is meant, did not obtain it, and did not verify the characterisation. This is the sole basis for the null we report on quality impairment. researchgate.net
- Reference list identifying Schindler, R. M., & Kibarian, T. M. (2001), Image Communicated by the Use of 99 Endings in Advertised Prices, Journal of Advertising, 30(4), 95–99; and Thomas, M., & Morwitz, V. G. (2005), Penny Wise and Pound Foolish: The Left-Digit Effect in Price Cognition, Journal of Consumer Research, 32(1), 54–64. Note: a commercial pricing-tool website's reference list, used solely to record two citations. We obtained neither paper and report no findings from either. testdrivepricing.com
- EconPapers bibliographic record for Anderson and Simester (2003), Quantitative Marketing and Economics, 1(1), 93–110, confirming authorship, journal, volume, issue and pages. Note: an economics repository record, used as a further independent confirmation of the citation. econpapers.repec.org
This article reviews consumer research and is not pricing, marketing or legal advice. Advertising a price as reduced carries obligations under consumer protection and competition law not addressed here. The field experiment paper was not obtained in full; no effect size, sample size or lift percentage is stated. The evidence is one research team and no independent replication was located. The meta-analytic null on quality perceptions reaches this article through third-party commentary and the underlying meta-analysis was neither identified nor obtained. All breakeven arithmetic is the authors' own and rests on stated simplifying assumptions. Nothing cited studied professional services.