Essay

Mimetic desire: why we want what others want, with data

Published · Updated 18 min read

Mimetic desire is the idea that people learn what to want by copying the desires of others, rather than by judging things purely on their own merits. The French thinker René Girard first set it out in Mensonge romantique et vérité romanesque (1961), published in English as Deceit, Desire, and the Novel (1965). He called it "triangular" desire at first and later "mimetic" desire.[1] The idea explains a lot of buying behaviour that a simple price-and-features model cannot: the busy restaurant that gets busier, the brand everyone suddenly wants, the expensive option that sells better than the cheap one.

This essay brings together four bodies of evidence. The first is an experiment with 14,341 people that showed how visible popularity reshapes a market.[2] The second is a brain-scan study of how price changes the experience of the same product.[3] The third is a crawl of about 11,000 shopping websites that counted fake popularity and fake scarcity signals.[4] The fourth is our own study of how unequally public attention is shared among the brands in nine markets, and how attention moves when a famous "model" points at someone. We close with a framework, the Honest Desire Ladder, and a step-by-step method.

Idea in brief

The problem. Buyers rarely judge an offer alone. They look at who else wants it. When people in an experiment could see what others had downloaded, popular songs became more popular, and which songs won became harder to predict.[2]

Why it happens. Visible choices act as a model for the next buyer. Price acts as a signal too: people rated the same wine as more pleasant when told it cost more.[3] Because these signals work, some sellers fake them. One crawl found 1,818 dark pattern instances on 1,254 shopping sites, led by low-stock messages, countdown timers and activity messages.[4]

The gap (our data). Attention in markets is very unequal. Across nine Wikipedia categories of brands and manufacturers (1,554 entries), the top 10% drew a median 64.0% of 2025 pageviews. Industrial categories were as concentrated as consumer ones.

The answer. Make real demand visible, choose the right model for each buyer, use only real limits, and let price tell the truth about quality. The Honest Desire Ladder puts these in order.

Key findings

The essay in 3 minutes
  1. Finding 114,341

    Visible popularity makes winners bigger and less predictable.

    In an artificial music market of 14,341 participants and 48 unknown songs, showing download counts increased both the inequality and the unpredictability of success.[2]

  2. Finding 2

    Price changes the experience itself.

    In a brain-scan study of 20 people, the same wine was rated as more pleasant when labelled at a higher price, and activity in a brain area linked to pleasure rose with it. In a blind tasting 8 weeks later, the differences disappeared.[3]

  3. Finding 3632

    Fake scarcity and fake popularity are common.

    A crawl of about 53,000 product pages on about 11,000 shopping sites found 632 low-stock messages, 393 countdown timers and 313 activity messages among 1,818 dark pattern instances.[4]

  4. Finding 486.8%

    Most verified deception is fake urgency, popularity or scarcity.

    Of 234 instances the researchers confirmed as deceptive, 203 (86.8%) were countdown timers, activity notifications or low-stock messages.[4]

  5. Finding 522.5%

    Attention is very unequal.

    In our study, the top 1% of articles in each category drew a median 22.5% of 2025 pageviews, and the bottom half drew a median 3.4%.

  6. Finding 64.0×

    Attention follows the model.

    When Taylor Swift attended a Kansas City Chiefs game at Travis Kelce's invitation on 24 September 2023, median daily views of Kelce's Wikipedia article over the next 28 days ran 4.0 times the 28 days before. Views of Chiefs quarterback Patrick Mahomes rose 1.3 times.

Chapter 1 · 1 min

What mimetic desire is

Girard built the idea from novels. In his analysis of Cervantes, Stendhal, Flaubert, Dostoevsky and Proust, desire is rarely a straight line from a person to an object.[1] It runs through a third party, the model, whose desire the person copies. Don Quixote does not choose his ideals; he copies them from the heroes of the romances he reads.

Girard split models into two kinds. External models are far away in status, time or place, so you copy them without competing with them. Internal models are close: a neighbour, a colleague, a rival firm. You copy them and compete with them, and that is where envy and rivalry start.

Economists approached the same ground from the demand side. Thorstein Veblen's The Theory of the Leisure Class (1899) gave us "conspicuous consumption", the idea that people spend visibly to show their standing.[5] A Veblen good is one that people want more as its price rises, because the high price is part of what they are buying.

Terms used in this essay

TermMeaning
Mimetic desireWanting something because a model wants it. Girard first called it "triangular" desire.[1]
Model (mediator)The person or group whose desire is copied. External models are distant; internal models are near rivals.
Social proofUsing what others choose as evidence of what is good. See the Hive notes on Cialdini's Influence.
Veblen goodA good people want more as its price rises, because price signals status or quality.[5]
Dark patternA design practice that tricks or manipulates users into choices they would not otherwise have made.[6]
Gini coefficientA measure of inequality from 0 (everyone equal) to 1 (one item takes everything). Used in the music market study and in ours.[2]

Four ways others shape demand

Type of demandWhat drives itEveryday exampleB2B example
Ordinary demandLower price, more buyersGroceries on discountCommodity fasteners bought on price
BandwagonOthers are buying itThe restaurant with a queueThe software every peer distributor uses
SnobFewer others have itA limited runA members-only supplier programme
VeblenThe high price itselfA luxury watchThe premium consultant hired because they are expensive

Chapter 2 · 1 min

The evidence, part 1: visible choices reshape markets

The cleanest test of mimetic desire in a market is the MusicLab experiment by Matthew Salganik, Peter Sheridan Dodds and Duncan Watts, published in Science in 2006.[2] They recruited 14,341 participants, mostly from a teen-interest website, and asked them to listen to and download 48 songs by unknown bands.

Some participants made choices independently, seeing only band and song names. The rest were randomly assigned to one of eight separate "worlds" in which they could also see how many times each song had been downloaded by earlier participants in that world. Each world started from zero and evolved on its own. In a second experiment, the researchers made the social signal stronger by listing songs in order of popularity instead of in a random grid.

The results:

The authors summed it up this way: "The best songs rarely did poorly, and the worst rarely did well, but any other result was possible."[2]

Limits of this evidence

Chapter 3 · 1 min

The evidence, part 2: price changes the experience

In a study published in PNAS in 2008, Hilke Plassmann, John O'Doherty, Baba Shiv and Antonio Rangel scanned the brains of 20 people while they tasted wine.[3] The participants were told they were sampling five different Cabernet Sauvignons, identified by price. There were only three wines.

Figure 1. Five price tags, three wines

WinePrice shown at low tag ($)Price shown at high tag ($)
Wine 1545
Wine 21090
Wine 33535

Wines 1 and 2 were each tasted twice, once at a low price and once at a high price; wine 3 was shown at one price. Wine 2's real retail price was $90. Source: Plassmann et al., PNAS, 2008, Materials and Methods.

People reported significantly higher pleasantness for wines 1 and 2 at the high price, and pleasantness was correlated with the stated price (r = 0.59). Activity in the medial orbitofrontal cortex, an area widely thought to encode experienced pleasantness, rose with price too.[3] In a follow-up session 8 weeks later, with no prices shown, there were no reported differences among the wines.[3]

For a seller, the lesson is narrower than it first sounds. Price did not change the wine. It changed expectation, and expectation changed the experience. The Hive's entry on the Coke and Pepsi brain study makes a related point about another famous scan, and Alchemy gathers many cases where perceived value beats a purely logical comparison.

Limits of this evidence

If you want a second opinion on what your price and proof signal to a buyer who has never heard of you, the Brave New audit looks at what a new visitor sees first and whether it makes your offer feel like a safe choice.

Chapter 4 · 1 min

The evidence, part 3: when sellers fake the signals

Because popularity and scarcity signals work, some sellers fake them. In 2019, Arunesh Mathur and colleagues at Princeton crawled about 53,000 product pages from about 11,000 shopping websites.[4] They found 1,818 instances of dark patterns on 1,254 sites (about 11.1%), in 15 types. Because the crawler only explored product, cart and checkout pages and only text-based interfaces, the authors call this a lower-bound estimate.[4]

Figure 2. The most common dark patterns imitate demand and scarcity

Low-stock message (scarcity)
632
Countdown timer (urgency)
393
Activity message (social proof)
313
Confirmshaming (misdirection)
169
Limited-time message (urgency)
88
Pressured selling (misdirection)
67
High-demand message (scarcity)
47
Testimonials of unclear origin (social proof)
12

Instances found

Instances of each dark pattern type found on about 11,000 shopping websites (eight of 15 types shown). Source: Mathur et al., Proceedings of the ACM on Human-Computer Interaction (CSCW), 2019, Table 1.

The three most common types all borrow the language of mimetic desire and scarcity: stock about to run out, an offer about to end, and other shoppers buying right now. Not every instance is a lie. A low-stock message can be true. So the researchers went further and tested which ones were deceptive.

Figure 3. Where the researchers confirmed deception

Countdown timer (reset or kept the offer after expiry)
157
Activity notification (falsely generated)
29
Low-stock message (scheduled or random stock counts)
17
All other deceptive types
31

Deceptive instances (count), of 234

Dark pattern instances the researchers verified as deceptive, out of 234 across 183 websites. Source: Mathur et al., CSCW 2019, Sections 1 and 5.

Of 393 countdown timers, 157 (39.9%) either reset after they ran out or the "expiring" offer stayed valid. Twenty-nine activity notifications were generated at random or hard-coded. Seventeen low-stock messages counted down on a schedule or showed random stock levels.[4]

The US Federal Trade Commission's 2022 staff report, Bringing Dark Patterns to Light, lists these practices by name: false low stock messages, false high demand messages and baseless countdown timers, with examples such as a message claiming only one item is left in stock, or an "Offer ends in 00:59:48" clock.[6] The report also cites a workshop study in which dark patterns doubled the share of consumers who signed up for a dubious identity theft protection service.[6]

Tip: If a customer would feel tricked on learning how a message was generated, do not show it. Regulators have named these practices, and the research shows they can be detected from the outside.

Chapter 5 · 3 min

Original research: how unequal is attention in a market?

The MusicLab study showed that visible popularity makes outcomes more unequal. We wanted to see what that inequality looks like in real markets today, including the industrial ones we work in most. Wikipedia pageviews give a free, public measure of how much attention each brand or company draws.

On 3 November 2026 we took every article listed directly in nine English Wikipedia categories: six consumer categories (luxury brands, watch brands, shoe brands, toy brands, cosmetics brands, car brands) and three industrial ones (machine tool builders, industrial machine manufacturers, pump manufacturers). For each of the 1,554 category entries (1,485 distinct articles, since a few appear in more than one category) we summed human pageviews for 2025, then measured how the attention was shared. The full method is at the end of this essay.

Finding 1: a few names take most of the attention

Figure 4. Share of 2025 attention captured by the top 1% and top 10% of articles

Top 1%Top 10%
Toy brands
46.9%
86.4%
Industrial machine manufacturers
29.7%
81.9%
Watch brands
22.1%
65.2%
Shoe brands
22.5%
64.4%
Luxury brands
25.7%
64%
Machine tool builders
24.8%
63.2%
Pump manufacturers
13.9%
63.1%
Cosmetics brands
17.1%
58%
Car brands
10.8%
47.4%

Share of category pageviews (%), 0 to 100

Share of all 2025 English Wikipedia pageviews in each category that went to its most-viewed 1% and 10% of articles. Source: Brave New Media analysis of Wikimedia pageviews data, 3 November 2026.

In the median category, the top 10% of articles drew 64.0% of all 2025 attention, the top 1% drew 22.5%, and the bottom half drew just 3.4%. Toy brands were the most concentrated: the top 10% took 86.4%, and one article, Labubu, drew 39.4% of the whole category's views on its own. That is the pattern the MusicLab worlds predicted: once a few names pull ahead, visible popularity keeps pulling attention their way. Car brands were the least concentrated, and even there the top 10% drew 47.4%.

Finding 2: industrial markets are just as unequal

Figure 5. Inequality of attention by category (Gini coefficient)

Toy brands
0.896
Industrial machine manufacturers
0.847
Watch brands
0.786
Pump manufacturers
0.770
Shoe brands
0.769
Luxury brands
0.769
Machine tool builders
0.722
Cosmetics brands
0.720
Car brands
0.697

Median across categories: 0.769

Gini coefficient, 0 to 1

Gini coefficient of 2025 pageviews across the articles in each category, from 0 (equal attention) to 1 (one article takes all). Source: Brave New Media analysis of Wikimedia pageviews data, 3 November 2026.

The median Gini coefficient across the nine categories was 0.769. The three industrial categories had a median of 0.770, against 0.769 for the six consumer ones. Industrial machine manufacturers (0.847) were more concentrated than luxury brands (0.769), and in that category the top 10% of articles drew 81.9% of attention. We expected buyers of machinery and pumps to spread their attention more evenly than shoppers for watches and shoes. They did not. In machine tools, the most-viewed article drew 60 times the views of the median article.

Finding 3: attention follows the model

Girard's theory predicts that desire flows through models. One of the most visible public tests of that in recent years came on 24 September 2023, when Taylor Swift attended a Kansas City Chiefs game at the invitation of tight end Travis Kelce, as the Associated Press reported that day.[7] We compared daily Wikipedia views for Kelce and three related articles in the 28 days before and the 28 days from that game.

Figure 6. Attention before and after the model pointed at someone

BeforeAfter
25,774
104,081
Travis Kelce
19,960
34,781
Jason Kelce
3,292
4,620
Kansas City Chiefs
12,620
16,136
Patrick Mahomes

Median daily pageviews

Median daily English Wikipedia pageviews, 27 August to 23 September 2023 against 24 September to 21 October 2023. Source: Brave New Media analysis of Wikimedia pageviews data, 3 November 2026.

Kelce's median daily attention rose four times, and on 25 September 2023 his article drew 747,470 views, 29 times his median for the four weeks before. The effect spilled over to people and things next to him: his brother Jason, a Philadelphia Eagles center, saw his article rise 1.7 times, and the team's article rose 1.4 times. Views of Chiefs quarterback Patrick Mahomes rose only 1.3 times. The extra attention went to the person the model chose, not to the team around him.

What our data does and does not show

Pageviews measure looking, not buying. Wikipedia categories are built by volunteers, so membership is uneven, and some category members are only loosely related (a few large companies sit in categories we would not choose). Views of redirect titles are not included in the concentration study. The Kelce case is one event, and the weeks before it already included rumours, so the 4.0 multiple understates the jump from his normal level. Read these as pictures of how attention is shared and how it moves, not as market shares.

Chapter 6 · 1 min

A framework: the Honest Desire Ladder

The research points to an order of work. Each rung depends on the one below it. Social proof does nothing if the product disappoints, and scarcity is worthless if nobody wants the product in the first place.

RungQuestion it answersEvidenceWhat to check
5. Honest price signalDoes your price tell the truth about quality?Price raised rated pleasantness of the same wine; the effect vanished without prices.[3]Price against peers, what each tier includes, whether delivery matches the promise
4. Real limitsAre your scarcity messages true?157 of 393 countdown timers were deceptive in one crawl.[4]Capacity, batch sizes, seasons, seat limits stated plainly
3. The right modelIs your proof from someone your buyer wants to resemble?Attention followed the model, not the biggest star (our Kelce data).Testimonials and logos matched to each buyer type and market
2. Visible demandCan a new buyer see that others chose you?Visible download counts made popular songs more popular.[2]Customer counts, named reviews, "most chosen" labels that are true
1. Real valueWould buyers choose this with no signals at all?The best songs rarely did poorly in any world.[2]Results, quality, delivery, repeat purchase

Most teams start at rung 4, adding urgency because it is fast. The research suggests starting at rung 1 and climbing. Signals multiply real value; they do not replace it.

Chapter 7 · 2 min

How to use mimetic desire honestly: a step-by-step method

Step 1: Count your real proof

List every customer number, review, result and logo you could publish, with its source and your permission to use it. If you cannot back it up, leave it out. Keep this as a one-page proof file.

Step 2: Choose the model for each buyer

For each main buyer type, decide who they most want to resemble. In industrial and B2B sales, where our supply chain and manufacturing experience sits, the strongest model is usually a peer one step away in the same chain: a mid-sized distributor trusts another distributor more than a global brand. When you enter a new market, do this again, because buyers there may not know your current customers. Matching the right proof to each market and buyer type is one of the jobs we build AI flows for.

Step 3: Make real demand visible

Show customer counts that are accurate and current, named reviews with specific results, and "most chosen" labels on the option most people really pick. The MusicLab data suggests even a weak signal shifts choice.[2] The Hive's notes on Influence cover the strengths and limits of social proof.

Step 4: Write down your honest limits

List your real capacity, batches, seasons and seats. "We take four new projects a quarter" or "this batch is 2,000 units" is honest scarcity. A timer that resets is not. Scarcity borrows force from loss aversion, whose strength is debated, so lean on it lightly.

Step 5: Check your price signal

Ask whether your price tells buyers the right story. If you charge much less than peers of similar quality, buyers who cannot judge quality before purchase may read the low price as risk. Test a higher tier that really includes more.

Step 6: Run the "found out" test

Read every claim on your site and ask: would a customer feel misled if they knew how this was produced? If yes, rewrite it or remove it.

A made-up but typical case

This case is made up, but it's typical. Say a 60-person pump manufacturer sells to water utilities and wants to enter a new region. Its website shows logos of two global engineering firms and a "Limited offer" banner that never changes.

It runs the method. It replaces the global logos with three named utilities of similar size, each with a measured result. It removes the banner and states its real lead time and quarterly production slots. It adds a premium service tier with remote monitoring, priced above its standard tier. In the new region, it repeats step 2 with local reference customers. Nothing is fake, and every signal now points a buyer toward a model they recognise.

Which moves to make first

MoveLadder rungEffortLikely impactWhen
Remove any fake timer, stock count or activity message4LowHigh (risk removed)This week
Build the proof file2LowMediumThis month
Match testimonials to each buyer type3MediumHighThis quarter
State real limits plainly4LowMediumThis month
Review price tiers against peers5MediumMedium to highNext pricing cycle
Repeat model matching for each new market3MediumHighEach market entry

For questions about how this applies to your situation, the Brave New Q&A answers the ones we hear most.

Chapter 8 · 1 min

What this means for leaders

For manufacturing and supply-chain companies, our data has an uncomfortable message: attention in industrial categories is as unequal as in consumer ones. A small number of names get most of the looking. Buyers who search a category see the leaders first, and their choices then confirm the leaders' position, as the MusicLab worlds did.

The research also shows a way in. In the music market, the winners differed from world to world, so early visible demand mattered a great deal.[2] And in the Kelce case, attention went to the person the model pointed at. For a challenger expanding into a new market, that means the first reference customers in that market are worth more than any campaign: they are the model the next buyers copy. The growth work is to win those first models, make their choices visible and repeat it market by market. That is the work we build AI flows for, so a small team can run proof, reference and outreach programmes in several markets at once.

If you want an AI flow that puts honest proof in front of more buyers in more markets, start with the Brave New audit.

Chapter 9 · 1 min

Common mistakes

Chapter 10 · 1 min

Methodology

Concentration study. On 3 November 2026 we listed every article (namespace 0) that is a direct member of nine English Wikipedia categories: Luxury brands, Watch brands, Shoe brands, Toy brands, Cosmetics brands, Car brands, Machine tool builders, Industrial machine manufacturers and Pump manufacturers. Subcategories were not included. For each article, a script requested monthly pageviews for January to December 2025 from the Wikimedia REST API (English Wikipedia, all access methods, agent type "user", which excludes known bots) and summed them. Articles with no recorded views counted as zero (35 of 1,554 entries). The script identified itself with the user agent "BraveNewResearch/1.0" and a link to brave-new.com; when the API asked it to slow down (HTTP 429), it waited and repeated the same request. For each category we computed the share of views going to the top 1% and top 10% of articles (rounded up to whole articles), the share going to the bottom half, and the Gini coefficient. Medians are across the nine categories.

Model effect study. Daily pageviews (same API settings) for Travis Kelce, Jason Kelce, Kansas City Chiefs and Patrick Mahomes from 27 August to 21 October 2023. We compared the median of the 28 days before 24 September 2023 with the median of the 28 days starting that day.

Limits. Pageviews measure interest, not purchases. Category membership is set by Wikipedia editors and is incomplete; a few members are large companies only loosely tied to the category. Views arriving through redirect titles are not counted. One event cannot show a general rule.

The data behind every figure in this essay is published alongside it as CSV files, with the scripts that produced it.

Chapter 11 · 1 min

Frequently asked questions

What is mimetic desire in simple terms?

Mimetic desire is the idea that people learn what to want by watching others. Instead of choosing purely on their own, they copy the desires of a model: someone they admire or compare themselves with. René Girard developed the idea in his 1961 book, published in English in 1965 as Deceit, Desire, and the Novel.

What is an example of mimetic desire in marketing?

Showing how many people chose a product is the simplest example. In the MusicLab experiment with 14,341 participants, letting people see download counts made popular songs more popular and made the final rankings less predictable. Testimonials, customer counts and "most popular" labels all work on the same principle.

What is a Veblen good?

A Veblen good is something people want more as its price rises, because the price itself signals status or quality. The name comes from Thorstein Veblen's 1899 book The Theory of the Leisure Class, which introduced conspicuous consumption. Research also shows price can change experience: people rated the same wine as more pleasant at a higher price.

Is scarcity marketing ethical?

Scarcity marketing is ethical when the limit is real, such as a fixed batch, set capacity or a seasonal product. It becomes deceptive when the scarcity is fake. One crawl of about 11,000 shopping sites found 157 countdown timers that reset or kept the offer after expiry, and the US FTC names false low-stock messages and baseless countdown timers as dark patterns.

How can a B2B company use mimetic desire?

Show proof from customers your next buyer wants to resemble, usually a peer of similar size in the same industry. Make real demand visible with named results and accurate numbers. When entering a new market, win and publish local reference customers first. Brave New Media builds AI flows that match proof to buyers across markets.

Your next step

This week, open your homepage and list every claim of popularity, limit or quality on it, with its source next to each one. Anything you cannot source, cut or fix. Anything you can source but are not showing, add. Then ask who each testimonial is a model for, and whether that is the buyer you want next.

If you want a team to do this across every market you sell to, start with the Brave New audit. For a weekly read on how money, risk and desire show up in business, subscribe to The Brave New Business Journal.

Brave New Media is a marketing agency that builds AI flows to help companies scale their ideas, grow their customer base and expand into new markets, with experience across supply chain and manufacturing.

See also

References

1. Imitatio. "Girard's works": Mensonge romantique et vérité romanesque (Grasset, 1961); Deceit, Desire, and the Novel, trans. Y. Freccero (Johns Hopkins University Press, 1965). imitatio.org

2. Salganik, M. J., Dodds, P. S. and Watts, D. J. "Experimental Study of Inequality and Unpredictability in an Artificial Cultural Market." Science 311(5762), 2006, pp. 854-856. pdodds.w3.uvm.edu

3. Plassmann, H., O'Doherty, J., Shiv, B. and Rangel, A. "Marketing actions can modulate neural representations of experienced pleasantness." PNAS 105(3), 2008, pp. 1050-1054. pmc.ncbi.nlm.nih.gov

4. Mathur, A., Acar, G., Friedman, M. J., Lucherini, E., Mayer, J., Chetty, M. and Narayanan, A. "Dark Patterns at Scale: Findings from a Crawl of 11K Shopping Websites." Proceedings of the ACM on Human-Computer Interaction 3 (CSCW), 2019. arxiv.org

5. Veblen, T. The Theory of the Leisure Class. 1899. gutenberg.org

6. US Federal Trade Commission. Bringing Dark Patterns to Light. Staff report, September 2022. ftc.gov

7. Skretta, D. "Taylor Swift turns out to see Travis Kelce, Kansas City Chiefs play Chicago Bears." Associated Press, 24 September 2023. ksat.com

8. Brave New Media. Attention concentration in nine Wikipedia categories (2025 pageviews) and model effect study, 3 November 2026. Data and scripts published with this essay.