Your Mouse Movements Are Setting Your Price. The FTC Just Confirmed It.

You are looking at a pair of boots. You have looked at them twice before. You added them to your cart last Tuesday, then closed the tab when rent cleared. Now you are back, and the price is slightly different — not dramatically, not insultingly, just enough that you wonder if you remembered wrong. You probably assume you did. That assumption is worth money.

The FTC's 2025 market study on surveillance pricing^[2] confirmed what consumer advocates and a handful of academic economists had been circling for years: a growing layer of algorithmic intermediaries sits between retailers and customers, harvesting behavioral data at granular resolution — cursor position, hover duration, scroll patterns, cart activity, session frequency, device type, time of day — and using it to generate individualized price recommendations. Not broad demographic segments. Not zip-code tiers. Individual prices, updated dynamically, calibrated to what the data suggests each particular user will pay before walking away.

This practice has a technical name — personalized or first-degree price discrimination — and economists have theorized about it since at least the early days of e-commerce. The FTC study did something different: it documented the actual infrastructure. It named the intermediary firms. It described the data pipelines. It showed that this is not a fringe experiment being run by one or two aggressive retailers. It is a mature, scaled industry operating quietly inside the ordinary act of buying things online.

What makes this particular form of extraction so durable is how invisible it is. A price that is personalized to you does not announce itself. There is no asterisk, no signal that the number you are seeing differs from the one your neighbor sees. You are not being deceived in a way that feels like deception. You are just shopping. The manipulation is structural, not theatrical.

The Intermediary Layer Nobody Talks About

Most reporting on dynamic pricing focuses on the retailers themselves — airlines, hotels, ride-share apps. But the FTC study's more consequential finding concerns the intermediaries: third-party firms that retailers contract with to optimize pricing in real time. These companies sit between the merchant and the consumer, invisible to both in a practical sense. The retailer outsources the pricing logic; the intermediary supplies the model, the data infrastructure, and the customer-facing price. The retailer gets revenue optimization. The intermediary gets a cut and, crucially, gets the data — behavioral signals from millions of shopping sessions across multiple retail clients simultaneously.

This structure means the data harvested from your session on one retail site may inform the pricing model applied to you on another. The intermediary's value proposition is precisely this cross-retailer behavioral portrait. Your hesitation pattern on a home goods site, your cart abandonment behavior on an electronics retailer, your return-visit frequency on a clothing platform — these signals aggregate into something more predictive than any single retailer could build alone. The incentive to share data with a pricing intermediary is also an incentive to feed the intermediary's model of you, which that intermediary then deploys everywhere.

“The retailer outsources the pricing logic; the intermediary gets the data — behavioral signals from millions of sessions across multiple retail clients at once.”

This is the same structural logic that made third-party advertising networks so powerful and so hard to regulate. Research on third-party data broker ecosystems has long noted that once data leaves the first-party context where a user might plausibly have consented to its collection, it moves through a chain of secondary uses that no consent framework meaningfully covers. Surveillance pricing is that problem applied directly to the number on the checkout screen.

What the Cursor Actually Reveals

The behavioral signals the FTC study catalogued are not exotic. Cursor hesitation — the pause before clicking, the hover over a price — is a well-established proxy for purchase uncertainty. Research on mouse-tracking as a window into consumer decision-making has demonstrated that cursor movement patterns correlate meaningfully with deliberation, ambivalence, and price sensitivity^[3]. Marketers have known this for years. What the pricing intermediaries have done is industrialize it: convert that micro-signal into a real-time input to a pricing algorithm.

Cart abandonment is similarly legible. A cart abandoned at checkout signals something different from a cart abandoned while browsing. An item added and removed multiple times signals something different from an item added and left. Return visits within a short window signal urgency or growing intention. The algorithm does not need to know why you are doing these things. It only needs to know that users who do them in this pattern tend to convert at a higher price. You are not being read as a person. You are being read as a pattern that resembles other patterns that yielded a particular outcome.

Device type feeds into this too. Studies on device-based price discrimination in e-commerce have shown that shopping from a newer iPhone versus an older Android handset functions as an income proxy, and that some retailers and platforms have used this signal, among others, to vary prices or offers. This is not speculation — it has been documented in real retail environments. The FTC study gives it regulatory acknowledgment rather than just journalistic exposure.

“You are not being read as a person. You are being read as a pattern that resembles other patterns that yielded a particular outcome.”

Why This Is Different from Coupons

A common defense of personalized pricing is that it is simply dynamic pricing with better data, no different in principle from loyalty discounts, coupons, or negotiated rates. This framing is worth taking seriously, because it contains a grain of truth and a significant distortion. Coupons and loyalty programs are legible. You know you have the coupon. You know other people might not. The discount is explicit, and you had to do something to get it — clip it, sign up for a program, use a code someone shared. The pricing asymmetry is visible and, in a loose sense, voluntary.

Surveillance pricing inverts this. The asymmetry is invisible and unearned in either direction. You do not know you are paying more than someone else. You did not fail to take an action that would have gotten you the lower price. You simply moved your cursor in a way that registered uncertainty, or you came back to the site one time too many, and the algorithm decided you would pay. The mechanism is opaque, the signal is involuntary, and there is no opt-out pathway because there is no disclosure. You cannot clip a coupon against a price you do not know exists.

There is also a distributional concern that the coupon analogy obscures. Research on price discrimination and consumer welfare suggests that first-degree personalized pricing does not simply extract more from the wealthy while offering deals to the price-sensitive. In practice, the signals used as willingness-to-pay proxies — device type, location, browsing frequency, cart behavior — often correlate with economic precarity rather than affluence. Someone who returns to a product page multiple times may be doing so precisely because they are carefully managing a limited budget, not because they are desperate to pay a premium. The algorithm reads the behavior, not the underlying reason for it.

The Consent Problem That Isn't Going Away

Every major retail site you visit asks you to accept cookies. The prompt is almost universally designed to make acceptance the path of least friction — a single large button versus a collapsed menu of granular options that takes four clicks to configure. Studies on cookie consent interface design and dark patterns have consistently found that the vast majority of users accept all cookies by default, and that acceptance rates drop sharply when interfaces are made genuinely neutral^[1]. The consent architecture is not designed to inform. It is designed to collect.

Even setting aside the design manipulation, the consent frameworks in place were not written with surveillance pricing in mind. When a user accepts cookies for "analytics and performance" or "personalized content," they are not agreeing, in any meaningful sense, to have their cursor hesitation converted into a pricing signal that costs them money. The legal text may technically permit it. The practical understanding does not. This is the gap the FTC study implicitly identifies, and it is the gap that makes the current moment feel less like a policy question and more like a basic legitimacy problem. As covered in our earlier piece on what platforms learn from chatbot conversations, the data you generate while interacting with a service rarely stays inside the box the interface suggests.

What Regulation Can and Cannot Fix

The FTC study stops short of proposing specific remedies, which is partly procedural and partly honest. Regulating personalized pricing is genuinely hard. Prohibiting price discrimination outright would have significant downstream effects on legitimate dynamic pricing — the kind that shifts airline seat prices based on aggregate demand, not individual behavioral surveillance. Requiring disclosure that a given price is personalized raises its own questions: what exactly triggers the disclosure, and how do you audit the algorithm to know when it is happening? The intermediary structure adds another layer of evasion potential, since the retailer can claim it does not control the pricing model it contracted for.

What cleaner regulation could plausibly do is restrict the data inputs. If cursor behavior, cart activity, and session frequency cannot be legally used as pricing inputs — as opposed to product recommendation inputs or site optimization inputs — the personalized pricing model loses much of its signal. This is harder to enforce than it sounds, because the same data is often used for multiple purposes simultaneously and disentangling those uses requires the kind of algorithmic transparency that most firms resist. The pattern here mirrors what the FTC's study on government surveillance outsourcing described: the infrastructure for data-driven targeting is general-purpose, and the same pipes that serve one function can serve another with minimal reconfiguration.

“If cursor behavior and cart activity cannot be used as pricing inputs, the personalized pricing model loses much of its signal — but disentangling those uses requires transparency that most firms resist.”

There is also a competitive dynamics problem. A retailer that does not use surveillance pricing is not competing on a level field with one that does. The optimizing retailer extracts more revenue from each transaction; it can reinvest that margin into lower headline prices that attract more customers, while recouping the revenue through personalized premiums on the back end. The non-optimizing retailer, showing the same price to everyone, looks more expensive on the comparison sites and less profitable in the books. The market incentive is not neutral. It systematically rewards the surveillance model.

The Habit the System Is Installing

The behavioral consequence of surveillance pricing that gets the least attention is not what it does to individual transactions. It is what it does to the act of shopping itself. When you cannot trust that the price you see is the price anyone else sees, comparison becomes harder. The reference point dissolves. Price anchoring — the cognitive mechanism by which a stated price shapes your sense of what is reasonable — stops working reliably, because the anchor is personalized to your profile rather than set by market conditions. You are making decisions against a benchmark that was constructed specifically to influence your decision.

This is a different kind of attention economy problem. Recommendation algorithms shape what you want. Surveillance pricing shapes what you pay for what you already want. Together, they close the loop: the platform influences desire, then extracts maximum value from the desire it influenced. The user's role in this system is not to make free choices. It is to generate signals that the system can monetize, twice — once through advertising, once through pricing. The shopping cart is not a neutral tool. It is a behavioral data collection device that happens to sometimes result in a purchase.

The FTC study will generate policy debate, some of it productive. But policy moves slowly, and the infrastructure being described is already mature. The more immediate shift is perceptual: understanding that the price on a checkout screen is increasingly a proposition, not a fact — a number the algorithm believes you will accept, derived from a profile you did not know you were building, managed by a company you have never heard of, on behalf of a retailer that prefers you not think about any of this while you decide whether to click buy.

References

1. Accepting cookies: Nudging, deceptive patterns and personal preference (sciencedirect.com)
   Demonstrates that cookie consent acceptance rates drop sharply when interfaces are designed to be genuinely neutral rather than manipulative.
2. FTC Surveillance Pricing Study Indicates Wide Range of Personal Data Used to Set Individualized Consumer Prices (ftc.gov)
   FTC's 2025 market study documenting that third-party intermediaries use behavioral data like location, browsing patterns, and shopping history to set individualized prices for consumers.
3. Measuring the Factors Influencing Purchasing Decisions: Evidence From Cursor Tracking and Cognitive Modeling (pubsonline.informs.org)
   Provides research evidence that cursor movement patterns correlate with consumer deliberation, ambivalence, and price sensitivity.