A price can change because the market changed. It can also change because the seller learned something about the person looking at it. Those are different events. If we want to judge the second, we have to stop treating every example of the first as proof.
That is where this research brief begins. My concern is straightforward: a business should not gain a hidden advantage simply because it can study a customer more closely than the customer can study the deal. But concern is a reason to investigate. It cannot do the investigation for us.
This is a six-chapter research primer for a future book. It reviews selected primary studies, explains their disagreements and develops a small, reproducible economic illustration. It is a targeted review, not a systematic census of all research or a live audit of retailers. Evidence was checked on October 9, 2026. The short shopping examples and poem speakers are invented; they are not reports of my experiences. Model amounts are hypothetical units, not measured prices. Research findings have separate sources.
Start with the complete offer
Imagine two people comparing a hotel. One sees 140 and the other sees 160. A screenshot puts the difference beyond dispute. It does not explain it.
Are they looking at the same dates, room, occupancy and cancellation terms? Is breakfast included? Are taxes shown? Is one signed into a membership scheme? Did one search just after the last room in a cheaper category was sold? Are the amounts in the same currency? A reliable comparison needs answers to those questions before it can isolate a price difference associated with the customer.
The object of study is the complete offer. A room name alone is insufficient. So is a product photograph. A price belongs to a seller, an item, a quantity, a time, a delivery arrangement and a set of terms. Some of those features are easy to record. Others may be hidden until checkout. When researchers cannot match them, they should mark the comparison uncertain instead of quietly counting it as discrimination.
This is already a practical improvement. If you are asking a company to explain a difference, give it a comparison it can investigate. Two complete records are more useful than two disconnected numbers.
Here are the working distinctions used throughout this brief:
- Dynamic pricing: the price changes with time or market conditions, such as demand or available capacity. It does not, by definition alone, require personal data
- Group pricing: a known category receives a different price, perhaps through membership, a published concession or a particular sales channel
- Personalized pricing: information about a buyer helps determine the price offered to that buyer for the same product and terms
- Personalized promotion: information about a buyer helps determine a coupon, discount or bundle. The final price can differ even when the headline price is common
- Steering: the products shown, or their order, change. Two people can be directed toward different spending without seeing different prices for the same item
These are analytical categories, not a declaration of what is lawful in every jurisdiction. A practice can combine several of them. A shop could change its base price each day, offer a membership discount and rank products differently for each visitor. Researchers need to separate those operations rather than attach one dramatic label to the whole screen.
What a price measures, and what it leaves out
A seller would like to know the highest amount a buyer will accept. Economists call that willingness to pay. It is useful in a model, but it is not a measure of moral worth, need or comfortable affordability.
Consider two invented buyers who will each pay 30 for the same item. One has abundant spare money. The other needs it urgently and will cut spending somewhere else. Their maximum accepted price is identical. Their circumstances are not. A pricing rule that learns only the number could treat them alike while missing the difference that matters to a humane policy.
That example does not prove that an actual seller detects emergencies or deliberately exploits them. It shows why a willingness-to-pay estimate cannot settle a fairness question by itself. A model has to leave things out. We should know what they are before using its results to defend an institution.
Adam Smith called attention to the “higgling and bargaining of the market” in The Wealth of Nations. He was discussing how exchange values get compared, not predicting digital profiling. The phrase is useful here because it makes the negotiation visible. A posted number can still be the outcome of a bargaining process, even when only one side has a chance to propose it. In a digital setting, our question is which information enters that process and who can challenge the result. Smith, Book I, Chapter V
I do not think everyone should have to become a pricing researcher to buy ordinary things. The need for careful research is partly evidence of a public responsibility: institutions should make routine exchange understandable enough that people can get on with their lives.
A claim needs a matching test
There are several questions hidden inside the sentence “They use our data to charge us more.” Each needs its own evidence.
First, does a system collect the information? Second, does it use that information in a pricing decision? Third, does this change the offer compared with an appropriate alternative? Fourth, does it make people worse off? Fifth, who bears the loss?
A privacy policy might help answer the first question. A technical document could help answer the second. A controlled comparison might address the third. The fourth also needs a credible account of what people would buy under another pricing rule. The fifth requires looking beyond an average.
This chain prevents two common mistakes. One is to turn an impressive technical capability into a claim about its widespread use. The other is to dismiss a documented capability because nobody has yet measured its full social effect. We can investigate and set safeguards before possessing a national cost estimate. We should still be honest that the estimate is missing.
The Federal Trade Commission’s January 2025 interim work is important for precisely this reason. It describes an intermediary industry capable of using detailed information to influence prices, discounts and displays. It is not a representative survey establishing how often every retailer uses each practice. Its examples also include hypothetical scenarios used to protect confidential information. The distinction determines what we can responsibly say about particular businesses. FTC initial findings
The position developed in these chapters is that people deserve offers they can compare, reasons they can question and evidence that claimed benefits reach them. That position can survive a finding that some personalization lowers prices. In fact, it becomes stronger when we can distinguish a real benefit from a convenient story about one.