Personalized pricing is the practice of setting a price for an individual consumer using data about that person, rather than posting one price that everyone in the market sees. It is the third and most disruptive lever in modern retail pricing, and unlike the first two it does not just change what you charge. It changes whether the phrase "our price" still refers to anything.
Every pricing research method in common use assumes a single posted price. Van Westendorp asks at what price a product becomes too expensive, which presumes a shared reference. Gabor-Granger builds a demand curve across a price ladder everyone could face. A price-image tracker asks whether shoppers think your store is cheap. Each of them quietly assumes a common denominator. Personalized pricing removes it.
What the regulator actually found
On 17 January 2025 the U.S. Federal Trade Commission published interim findings from its surveillance pricing study, based on documents obtained through Section 6(b) orders issued to eight companies in July 2024. The staff perspective drew on material provided by Mastercard, Accenture, PROS, Bloomreach, Revionics and McKinsey & Co.
The FTC's framing is worth quoting: staff found that consumer behaviours "ranging from mouse movements on a webpage to the type of products that consumers leave unpurchased in an online shopping cart can be tracked and used by retailers to tailor consumer pricing."
The specifics reported:
- Some respondents can set individualised prices and discounts from granular consumer data, with the example of a cosmetics company targeting promotions to specific skin types and skin tones
- Intermediaries examined can show higher priced products based on a consumer's search and purchase activity. The hypothetical the FTC gives is a consumer profiled as a new parent being shown higher priced baby thermometers on the first page of search results
- These intermediaries worked with at least 250 clients, from grocery stores to apparel retailers
Then-Chair Lina Khan framed the stakes as whether "firms are charging different people different prices for the same good or service." The report was released on a 3-2 Commission vote, with Commissioners Ferguson and Holyoak dissenting, and the FTC has been explicit that the study is ongoing and the examples published are hypothetical, because the underlying 6(b) material is confidential. Treat it as evidence about capability and market structure, not as a finding that any named company charged you more.
What the academic evidence says: it is rarer than the headlines
Now the counterweight, because the honest version of this article needs it.
The European Parliament study Personalised Pricing (PE 734.008, November 2022, by Rott, Strycharz and Alleweldt for the IMCO committee) summarises the empirical record. A study conducted for the European Commission across 8 Member States and 4 sectors - televisions, sports shoes, hotel rooms and airline tickets - ran a mystery shopping experiment on 160 websites and did not find personalised pricing. Price differences appeared in only 6% of situations with identical products, and the median difference was less than 1.6%. Separately, Vissers and colleagues ran 66 simulated user profiles searching 25 airlines twice a day in 2014 and found no evidence of individual price differences based on profile, although prices were highly dynamic.
So the measured incidence of true first-degree personalisation is low. Anyone selling you a panic about it is overselling.
The gap between capability and incidence is the whole story
Put the two bodies of evidence side by side and the picture is coherent, not contradictory.
| What it establishes | What it does not establish | |
|---|---|---|
| FTC 6(b) interim findings | The infrastructure exists, is commercially sold, and reaches at least 250 client retailers | That widespread individual price discrimination is occurring today |
| EC mystery shopping (160 sites) and Vissers (25 airlines) | Detectable first-degree personalisation was rare and small in the settings tested | That it is not happening in untested settings, or now |
Why the gap persists is explained by the same European Parliament study, and this is the most useful finding in the whole literature. Consumer attitudes are savage. A survey of Dutch consumers found more than 80% consider fully individualised pricing unacceptable and unfair to some extent. In the U.S., 91% of respondents had a negative attitude toward supermarkets personalising prices to individual consumers, and 64% responded negatively to individually personalised coupons. Notably, second-degree personalisation such as quantity discounts is broadly considered acceptable, because the shopper chooses it.
The study's conclusion is stark: there are only three ways to apply personalised pricing without deterring consumers - as a discount rather than a higher price, fully transparently, or secretly. Firms tend to abstain, or to stop immediately when discovered, as Amazon did in 2000 after customers found that deleting cookies lowered the price of a DVD.
That is the real structure of this market. The capability is widely sold, the observed incidence is low, and the reason it stays low is that the practice cannot survive disclosure. Which means the entire equilibrium rests on shoppers not finding out, and disclosure is exactly what regulators are now mandating.
Disclosure is arriving, and it is being written in your words or theirs
European Union. The Omnibus Directive inserted Article 6(1)(ea) into the Consumer Rights Directive, requiring traders to inform consumers where the price has been personalised on the basis of automated decision-making, so the consumer can decide whether to enter the contract on that basis.
New York. The Algorithmic Pricing Disclosure Act requires companies using algorithmic pricing based on personal data to display a clear and contemporaneous disclosure reading: "THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA", with civil penalties of up to $1,000 per violation. The National Retail Federation sued on 2 July 2025, arguing the requirement compelled speech in violation of the First Amendment. The court upheld the disclosure as plainly factual, and the law took effect on 10 November 2025.
Read that mandated sentence as a piece of copy. It is the single most hostile framing of the practice available, it is now law in a major market, and no brand tested it before it was written.
The pressure is not only online. In August 2024 Senators Warren and Casey wrote to Kroger's CEO over electronic shelf labels, warning about the potential to "abuse their power and surge grocery prices suddenly" and raising the plan to use cameras and facial recognition on shelf displays to infer shopper attributes. Kroger responded that it "does not and has never engaged in surge pricing". In July 2025 Senators Gallego, Warner and Blumenthal wrote to Delta over its AI pricing work with Fetcherr; Delta replied that "There is no fare product Delta has ever used, is testing or plans to use that targets customers with individualized prices based on personal data."
The pattern across all three is identical. An accusation of individualised pricing, an emphatic denial, and no independent way for anyone outside the company to verify either claim.
What this does to your pricing research
This is the part that should change how you work, and it applies whether or not you personalise anything.
1. Your reference price may no longer exist. Van Westendorp, Gabor-Granger and price-image tracking all assume respondents share a common price environment. If some of your respondents have been seeing algorithmically adjusted prices, you are averaging across people who faced different offers and reporting the mean as "the" acceptable price. The method does not fail loudly. It returns a confident number that describes nobody.
2. You cannot detect a competitor's personalisation from outside. Mystery shopping needs profiles, and 160 websites and 66 simulated profiles found almost nothing. Scanner and panel data record transactions, not the counterfactual price a different shopper would have been shown. The only instrument that reaches this is asking a large, diverse group of real shoppers what they were actually charged and what they believe about it.
3. The risk is asymmetric and the tail dominates. Measured incidence is low. Measured outrage is 80% to 91%. Any expected-value calculation on personalised pricing is driven almost entirely by the disclosure scenario, not the base case. That is precisely the scenario nobody researches.
So research the disclosure, not the algorithm. The practical programme:
- Test the mandated wording itself. Show shoppers the New York sentence in a realistic checkout and measure abandonment with a yes_no, confidence with a scale, and reasons in open_ended conversation. This is standard price display comprehension work applied to a sentence a legislature already wrote.
- Separate the acceptable forms from the unacceptable ones. Use ranking across quantity discounts, loyalty offers, targeted coupons, time-based dynamic pricing and fully individualised prices. The Dutch evidence says these are not one attitude but several, and brands routinely defend a benign practice using language that implies the malignant one.
- Establish what shoppers already believe about you. Use single_choice and multiple_choice to capture whether they think they have been charged a personal price, and by whom. Belief drives behaviour here regardless of truth, and the brand tracking instrument most companies run does not ask.
- If you use loyalty pricing, test whether shoppers experience it as a reward or a penalty for non-members. It is the same price gap read two ways, and the reading is the entire risk. Related: dark patterns testing.
Why this needs conversation, not a survey
A survey cannot do this work. Asked flatly whether they mind personalised pricing, people say they mind, and you learn nothing you could not have guessed. The useful signal is in the distinctions: which forms are acceptable, why a loyalty discount is fine and an inferred-income discount is not, what they think happened to them last time. Those are follow-up questions, and forms cannot ask follow-up questions.
The legacy stack handles this badly. Typeform and SurveyMonkey cannot probe. Qualtrics can field at scale but a sensitive topic like this needs a moderator who reacts to what was just said. UserTesting and dscout give you depth across a handful of sessions, at which point you have anecdotes about a question where the distribution is the finding. Dovetail organises the transcripts after somebody else has gathered them.
Koji is the AI-native platform built for this. AI-moderated voice or text interviews with several hundred shoppers, running in parallel, each one probing the specific distinction that respondent just drew - and no human moderator whose discomfort with the topic shapes the answer. Automatic thematic analysis turns several hundred conversations about fairness into a ranked set of themes with supporting quotes, and a one-click report goes to legal and pricing on the same day. Six structured question types sit alongside the conversation - open_ended, scale, single_choice, multiple_choice, ranking and yes_no - so you get the distribution and the reasoning from the same respondents.
Ten times faster than a traditional fielded study, no research expertise required, and from question to insight in hours rather than weeks.
Frequently Asked Questions
What is personalized pricing?
Personalized pricing sets a price for an individual consumer based on data about that person, such as location, browsing history, purchase history, device or inferred demographics. It differs from dynamic pricing, which varies prices over time for everyone, and from quantity discounts, which any shopper can choose to take.
Is personalized pricing actually widespread?
The evidence is mixed and should be reported as such. The FTC found the enabling infrastructure is commercially sold, with the intermediaries it examined working with at least 250 client retailers. Academic and European Commission testing found low detectable incidence: a mystery shopping study across 160 websites in 8 Member States observed price differences in only 6% of situations, with a median difference under 1.6%.
Is personalized pricing legal?
Generally yes, subject to disclosure and anti-discrimination law. In the EU, Article 6(1)(ea) of the Consumer Rights Directive requires traders to tell consumers when a price has been personalised through automated decision-making. In New York, the Algorithmic Pricing Disclosure Act requires the notice "THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA", with penalties up to $1,000 per violation; it took effect on 10 November 2025 after a First Amendment challenge failed.
How do consumers feel about personalized pricing?
Strongly negative when it is individualised. More than 80% of surveyed Dutch consumers considered fully individualised pricing unacceptable and unfair to some extent, and 91% of U.S. respondents had a negative attitude toward supermarkets personalising prices to individuals, with 64% negative on individually personalised coupons. Quantity discounts, which the shopper chooses, are broadly accepted.
How does personalized pricing affect pricing research methods?
It undermines the shared reference price that Van Westendorp, Gabor-Granger and price-image tracking assume. If respondents have faced different algorithmically set prices, the resulting acceptable-price range averages across incompatible experiences and produces a confident number that describes no actual shopper. Ask what people were charged and what they believe, not only what they would accept.
Can I detect whether a competitor is personalising prices?
Not reliably from outside. Mystery shopping requires simulated profiles and has historically found very little, and transaction data cannot show the price a different shopper would have been offered. The practical route is asking a large and diverse sample of real shoppers what they were shown and what they paid, which is a research problem rather than a data-purchasing problem.
Find out what your shoppers think they are being charged
If personalised pricing is rare, that is worth proving. If your shoppers already believe it is happening to them, that is worth knowing before a mandated disclosure appears at your checkout. Both are questions about belief and reaction, and both are answerable this week.
Koji runs AI-moderated interviews with several hundred of your category's shoppers, tests the exact disclosure wording regulators have already written, separates the pricing practices shoppers accept from the ones they punish, and delivers a report your pricing and legal teams can act on immediately.
The price you research has to be a price somebody actually saw.
