PRINTABLE VERSION HERE 

Dynamic pricing deserves attention now because three developments have converged. Artificial intelligence has made individualised pricing accessible, allowing for the possibility of a different price for every consumer. The growth of e-commerce has supplied both the behavioural data and the repricing infrastructure to act on it. Meanwhile, regulation is beginning to take shape, and the rules that will govern individualised pricing are being written before most retailers have decided where they stand.

As individualised pricing spreads, the question for department stores is not whether to run a data operation — they already run one, through their loyalty programmes, apps and data partnerships — but what to do with it. Two answers follow, and only one is defensible. The first is to point that operation at customers the way big tech does, personalising prices to extract the most each will pay but its failure mode is defeat by big tech itself. IADS argues that the second is to run the same machinery under a public commitment that it will only ever move a customer's price downward, and to make that transparency the differentiator.

The taxonomical pricing debate

The fundamental source of confusion in the contemporary pricing debate stems from the conflation of four concepts by media, regulators, and retailers. That many retailers have never resolved the confusion inside their own organisations is why so few can take a clear position on it. Algorithmic pricing, often powered by AI and machine learning, recalibrates prices across any combination of factors including, but not limited to, demand, competitor data, seasonal patterns, supply costs, and customer behaviour. NRF contends that this is simply the digitised and accelerated version of what retailers have always done manually.

Algorithmic pricing spans a spectrum in terms of data input, transparency, and degree of personalisation. Dynamic pricing lies on one end, differential pricing somewhere in the middle, and personalised (or surveillance) pricing on the opposite end:

•  Dynamic pricing is the real-time adjustment of prices based on immediate demand signals, inventory levels, or competitor pricing. It is market-facing, with prices fluctuating for everyone based on these conditions, not depending on the specific customer. Given its scale of digital and infrastructure capabilities, Amazon reportedly made over 2.5 million price changes per day as far back as 2013. The business models of airlines, hotels, ride-hailing apps, and fuel retailers are based on dynamic pricing.
• Differential pricing refers to charging different customer segments different prices such as seniors' discounts, student rates, and trade pricing. This is legally well-understood as it price discriminates based on market-facing customer characteristics.
•  Personalised pricing (also called surveillance pricing or individualised pricing) is the most contested variant. It uses personally identifiable data such as browsing history, location, credit score, purchase history, and mouse-movement patterns to infer each individual customer's maximum willingness to pay and price accordingly. Fordham Law professor Zephyr Teachout, explains it as using "the same data infrastructure that social media uses to sequence feeds, only here the information doesn't determine what someone sees but what they pay" The aim is to charge each buyer the most they can be induced to pay rather than a single market price.

The mechanism of individualised pricing

While modern dynamic pricing is proliferating across online storefronts, it is increasingly operationalised in physical stores, with the same interlocking technologies – in-store and online price labels, AI systems, and consumer data pipelines - enabling both. In either channel, the shopper's inability to observe the mechanism is engineered in rather than incidental.

In-store, electronic shelf labels (ESLs) are the enabler. The digital displays remove the need to manually re-sticker shelves, allowing prices to change in real time across an entire store silently, and leaving no record a shopper can inspect of what an item cost an hour earlier. While NRF argues that ESLs exist primarily for price accuracy and consistency, and to reduce compliance risk, rather than as a vehicle for surge pricing, they are also vehicles of enabling dynamic pricing when combined with backend analytical modelling, enhanced by AI.

AI systems learn over time how to find ‘optimised prices’ by running elasticity models examining how price changes affect demand across different SKUs, customer segments, and time windows. Digitalisation has resulted in the reduction of ‘menu costs’, allowing platforms to change prices as a function of demand at negligible cost. This can look like Uber and Deliveroo’s algorithmic surge pricing and discount coupons based on demand.

Consumer data pipelines power the entire process and separate algorithmic pricing from individualised pricing. None of the inputs they draw on are visible to the shopper being priced and include credit scores, GPS location patterns, search duration and frequency, individual purchase histories, browser cache and cookie data, and keystroke dynamics that correlate with different emotional states. The US Federal Trade Commission’s surveillance pricing study on intermediary firms such as McKinsey, Mastercard and Accenture, found that behaviours including mouse movements and items left in shopping carts are tracked and fed into individualised pricing models and used by at least 250 retail clients from grocery to apparel.

Loyalty programmes have historically been the most socially accepted form of individualised pricing as a straightforward exchange of data for rewards. However, that social contract has been hollowed out: in many cases the programme is now less a reward mechanism than the primary data-collection infrastructure that makes surveillance pricing possible. The clearest result is the reverse loyalty programme. Once a retailer has accumulated enough behavioural data to recognise a frequent customer's demand as inelastic, it can offer that customer fewer discounts rather than more resulting in the most loyal ending up being charged the most because their loyalty looks like captive demand. Reporting on Starbucks, whose loyalty data was found to have been shared with 64 third parties without customers' awareness, showed how loyal customers ended up paying more than others. The former director of the FTC's Bureau of Consumer Protection found that companies mine loyalty data to gauge how much customers will tolerate, and that programme designs are structured to obscure whether consumers save anything at all.

Why department stores cannot win the data war

The data asymmetry between big tech and any other retailer is structural and uncloseable. Amazon sees the full browsing and purchase graph of hundreds of millions of shoppers across every category plus the sales data of the third-party sellers who compete on its own marketplace. Through initiatives reportedly grouped under Project Starfish, a unified internal data system, Amazon ingests the product listings, prices, and transaction data of thousands of third-party brands, surfacing their goods and completing checkout inside its own app, even as it sends cease-and-desist letters to those who scrape its marketplace in return. Because many of those brands depend on Amazon's own cloud, the cost of resisting falls entirely on them.

Every transaction adds to Amazon's data, that data sharpens its next price, and the sharper price wins the next transaction, compounding the lead it already holds. A department store entering the same contest starts with less data, learns more slowly from what it has, and so improves its pricing more slowly than Amazon improves its own; the gap does not hold steady but widens with every cycle. The contest is also frequently run on Amazon's own ground: a department store that lists on its marketplace or builds on its cloud feeds the same transaction and cost data into the engine it is trying to beat, so the effort to compete on data can end up subsidising the competitor. A gap of this kind is not closed by spending more; spending only buys a place on a faster treadmill.

The data feeding Amazon's pricing engine therefore includes competitor and brand data that a department store could neither lawfully nor practically acquire. It reprices millions of times a day and runs continuous live experiments at a volume that produces reliable elasticity estimates for each SKU at a given moment. A department store simply lacks the traffic to learn at that scale. And because Amazon spreads the fixed cost of pricing infrastructure and data-science talent across a sales base no single retailer can rival, each increment it invests buys a sharper model than a department store's can. A retailer that competes on the platform's own terms therefore builds a permanently inferior copy of Amazon's machine and loses the data war anyway while giving up the one advantage Amazon cannot buy - a customer's trust that they are not being milked.

How individualised pricing erodes trust

The sharpest critique of personalised pricing is that it targets vulnerability and circumstance rather than preference. It prices the moment a shopper's bargaining power is lowest — the borrower with a poor credit score, or the parent buying for a sick child at midnight — pricing the weakness of the position, not a preference for better service.

Opacity is a core feature that makes individualised pricing sustainable. Because its operation is almost impossible to verify from the outside, a shopper has no basis on which to challenge it. As Teachout writes, "When different prices show up for different people, companies tend to claim it is not the work of surveillance pricing." Amazon and Delta have attributed observed price variation to "market dynamics, third-party sellers or A/B testing."

Economist and CEO of aiRESULTS, Matt Hasan argues that an algorithm may estimate what a customer can pay but cannot measure how they feel about paying it. Once a shopper discovers, or even suspects, that a price reflects what the system knows about them, the relationship turns adversarial. Those tools are already mainstream: price-history trackers such as CamelCamelCamel and Klarna's PriceRunner have made comparison a reflex, and a 2025 Deloitte survey found that 56% of US consumers planned to use AI assistants to compare prices and hunt deals. As that capability spreads, a growing share of consumers will police prices and disengage until they fall. This is still a trajectory, not a settled state, but it runs one way. As opacity grows costlier to defend, the endpoint is agentic commerce: buyer and seller algorithms manoeuvring against each other until the human relationship with the brand thins to nothing. A customer who starts tracking prices is no longer loyal but transactional, and the extra margin taken from them now is lost when they disengage. Nearly 70% of consumers said they are very or extremely protective of their personal data and would trust a merchant less for covertly using it.

This is what makes transparency decisive, and it answers the obvious objection that a platform like Amazon is trusted despite its data practices. Consumers do not process prices as rational negotiators but as signals of fairness and intent. Once a shopper starts to feel like a target rather than a customer, concealment cannot undo it — and suspicion alone, not proof, is enough to trigger the shift. Personal income shapes both how unfair a price feels and how strongly consumers act on it, so for households with the least budgetary room a shifting price reads as exploitation rather than just annoyance. The only long-term control is to make the logic of prices legible. Amazon is tolerated on convenience and price, not on pricing candour. This is the uncomfortable point: its scale is proof that most shoppers reward convenience over candour, so transparency is a bet on the segment that values being dealt with straight.

How a department store earns trust

Making that bet credible begins with an honest admission: department stores too operate loyalty programmes, apps, and data partnerships built on the same infrastructure described here. The real position is therefore not that a department store holds less customer data, but that it commits to using it in only one direction, to lowering prices, but not to raise them. The same behavioural signals that could locate a customer's ceiling can instead fund discounts, protect the price of essentials, and reward frequency, while the retailer refuses on principle to convert a rewards relationship into a map of who can be charged more. But a promise to hold that data and not weaponise it is not self-enforcing. The capability to find each customer's ceiling stays in-house, the margin from using it is permanent, and its use is nearly impossible to detect from the outside, so restraint is precisely the claim a shopper cannot verify.

Disclosure narrows that gap without closing it: a department store that states openly which data it collects and to what end, and commits that pricing intelligence will only ever move a customer's price downward, turns a private intention into a public standard the retailer can be measured against and caught breaking. Norway's REMA has used dynamic pricing to make about 2,000 price changes a day exclusively to lower them. What disclosure cannot do is remove the temptation. The department store is asking to be trusted with an instrument it declines to fire, and that trust is earned by making a breach visible and costly, not by the pledge itself.

This does not mean abstaining from personalisation, but being transparent about what it entails. Personalised discounts, openly offered, lower a specific customer's price and are welcome; personalised price floors, which raise it because the algorithm judges the shopper will pay more, are what customers punish. But 'visibly' is the hard part: from the outside, a discount someone else received and a price floor aimed at a customer look the same. Refusing the second cannot be shown by saying so; it has to be anchored in something checkable — a single public posted price that no shopper is ever charged above, with personalisation allowed to move only downward from it. That makes the floor's absence verifiable, because charging above the posted price would be detectable where a bare pledge would not be. Success depends less on how sophisticated the system is than on whether people feel it works for them rather than against them.

Why principled pricing survives the regulatory crackdown

The regulatory crackdown looks like a threat to personalised pricing, but its logic favours the retailer that has already committed to transparency. Disclosure laws like New York's requirement to display "This price was set by an algorithm using your personal data" risk sweeping in routine loyalty discounts and targeted promotions alongside surveillance pricing; if retailers pull back personalised offers to avoid compliance risk, the price-sensitive households, bulk buyers, and frequent shoppers who gain most from those discounts lose out. The NRF’s strongest argument is the ‘white box’ model of explainable algorithms, a ban on protected data, and oversight for bias.

New York is, for the moment, among the most heavily regulated jurisdictions on this issue and spans a range of regulation: the Algorithmic Pricing Disclosure Act (in force since November 2025 and upheld against the NRF's First Amendment challenge) mandates disclosure, while the pending One Fair Price Act would ban surveillance pricing yet preserve bona fide loyalty programmes and uniform discounts, and further state and New York City bills target electronic shelf labels and intra-day price changes. Whether a jurisdiction chooses a label or a ban, the instruments converge on the principle that a retailer may not use personal data, unilaterally and invisibly, to decide what a consumer should pay, but they act with very different bluntness.

That bluntness tells against the disclosure label most of all: it tars a benign, downward-only discount with the same ominous warning as extractive surveillance pricing, so even a transparent retailer pays a cost under it, badging its good offers as if they were suspect. The exposure is lowest, not nil, and lower under a well-drafted ban than under a blunt disclosure rule. A department store that has made transparency its differentiator is not scrambling to comply as it already meets the standard these laws are converging on.

Transparency as differentiator

Loyalty, trust, and individualised pricing are in tension as currently practised. A programme that rewards frequency and states its logic openly, and one that mines behavioural data to find the maximum a customer will bear, are structurally the same product with opposite intent, that consumers cannot tell apart from the outside. Transparency is what lets a department store resolve that tension without trying to out-compete Amazon on data, a contest it would lose. But that advantage is neither free nor guaranteed.

On the deepest question — whether pricing should be allowed to redistribute surplus from consumers to retailers at all — IADS's position is that neither retailers nor technologists should settle it on consumers' behalf. It is a distributional choice for regulators and democratic institutions, and the one place where self-policing is least trustworthy. Transparency is not a universal winning move. Its upside is genuine — studies of demand-based pricing find that disclosing the reason for a price raises how fair consumers judge it, so candour can command a premium among shoppers who value being dealt with straight. But the cost is just as real: Amazon's scale is standing proof that most shoppers, most of the time, still choose the cheaper, more opaque option. For the department stores that can hold that line, treating transparency as a differentiator rather than a compliance cost, opposing the Amazon model, is how they remain the retailers consumers still trust when every price is set by an algorithm.

Credits: IADS (Anchita Ranka)