Dynamic pricing in fashion: where does AI help and where does it harm the brand?
Algorithmic pricing can clear stock and protect margin, but frequent or personalised price changes carry brand, trust and legal risks. Where AI pricing earns its place in fashion and where it should stop.
KEY TAKEAWAYS Summary by the editors
- AI pricing in fashion delivers its clearest documented value in markdown and clearance decisions, where Zara's published field experiment found clearance revenue rose by about 6%.
- Frequent, highly visible price changes on full-price products risk teaching customers to wait for discounts and can undermine a brand's price positioning.
- Under EU rules applying since 28 May 2022, any announced price reduction must state the prior price, defined as the lowest price applied in a period of at least 30 days before the reduction.
- EU consumer law also requires traders to inform consumers when a price has been personalised on the basis of automated decision-making, and US regulators and states are scrutinising personalised pricing.
- A sound fashion pricing policy sets guardrails first: which products, channels and price moves the algorithm may touch, and which remain human decisions.
AI helps fashion pricing most in markdowns, clearance and promotion planning, where it decides how much and when to discount stock that is not selling. It harms the brand when prices move so often, so visibly or so personally that customers learn to wait, compare or distrust. The difference lies less in the algorithm than in the guardrails a brand sets around it.
What is dynamic pricing in fashion?
Dynamic pricing means adjusting prices frequently in response to demand, stock, competitor prices or time remaining in the season. In fashion it ranges from a gentle, rules-based markdown cadence to near real-time changes in online channels. Three variants are worth separating, because their risks differ sharply.
- Markdown optimisation: deciding the timing and depth of discounts on seasonal stock.
- Demand-based price changes: moving prices up or down in-season according to sell-through.
- Personalised pricing: different customers seeing different prices for the same item based on their data.
Where does AI pricing help fashion businesses?
The best-documented case is clearance. Caro and Gallien, working with Zara's pricing team and publishing in Operations Research in 2012, replaced a manual, informal markdown process with a formal forecasting and price optimisation model. In a controlled field experiment across all Belgian and Irish stores in the 2008 autumn-winter season, the new process increased clearance revenues by approximately 6%, and Zara subsequently used it worldwide for clearance markdown decisions.
Online, research with the flash-sale retailer Rue La La found that demand for an item depended on the prices of other items in its category, so prices had to be optimised together. Harvard Business School's Working Knowledge reports that a January 2014 field experiment raised revenue on the experimental products by about 9.7% with minimal impact on units sold. In both cases the algorithm worked within a business model that already accepted price variation.
How can dynamic pricing harm a fashion brand?
Fashion brands sell an idea of value as much as a product. When prices move constantly, several risks appear.
| Use case | Typical benefit | Main brand risk | Risk level |
|---|---|---|---|
| Markdown timing and depth | Higher clearance revenue, less leftover stock | Over-early discounting trains customers to wait | Low to medium |
| Promotion planning | Better targeted offers | Promotion dependency and eroded full-price sell-through | Medium |
| In-season price increases on bestsellers | Captures demand on scarce items | Perceived opportunism, social media backlash | Medium to high |
| Personalised prices per customer | Theoretical revenue gain | Fairness complaints, regulatory exposure | High |
| Wholesale and marketplace price parity | Consistent positioning | Algorithm undercutting retail partners | Medium |
For premium and luxury labels the risk is sharper, because price consistency itself signals quality. A brand that discounts automatically to clear stock may win this season's margin and weaken next season's full-price sell-through. That trade-off rarely appears in the algorithm's objective function unless someone puts it there.
Brands that sell through wholesale face an additional risk. If an algorithm lowers prices in the brand's own online store, retail partners selling the same products at recommended prices lose sales and margin, and may respond by cutting their orders or discounting in turn. Price moves in direct channels therefore need to be coordinated with wholesale terms and with marketplace sellers, otherwise the brand's own channel becomes the reason its price positioning erodes across the market.
What do regulators say about dynamic and personalised pricing?
In the EU, the Omnibus Directive (EU) 2019/2161 inserted Article 6a into the Price Indication Directive: any announcement of a price reduction must indicate the prior price, defined as the lowest price applied during a period of not less than 30 days before the reduction. Member States had to apply the rules from 28 May 2022. Rapid up-and-down pricing therefore directly affects which reference price a retailer may show next to a discount.
The same directive amended the Consumer Rights Directive so that traders must inform consumers, where applicable, that a price was personalised on the basis of automated decision-making. In the US, the Federal Trade Commission published initial findings from its surveillance pricing study in January 2025, noting that intermediaries can draw on a wide range of consumer data to influence the prices people see. The American Bar Association reports that New York enacted an Algorithmic Pricing Disclosure Act requiring a label when an algorithm sets a price using personal data, and that a federal judge dismissed a retail industry challenge to it in October 2025.
Where should a fashion brand draw the line?
A workable policy starts from brand position, not from technology. Most fashion businesses can justify algorithmic support for markdown depth and timing on seasonal stock, for promotion planning and for price testing of new products in controlled ways. Many will choose to keep full-price changes on core lines, price increases on scarce bestsellers and any form of individual price personalisation outside automation, or exclude them entirely.
- Define which categories, channels and markets the algorithm may price, and which it may not.
- Set floors, ceilings and a maximum frequency of change per item.
- Add brand metrics to the objective, such as full-price sell-through share, not only revenue.
- Log every automated price change with the reference price required by law.
- Review outcomes each season with merchandising, brand and legal teams together.
Is dynamic pricing right for every fashion brand?
No. Off-price, flash-sale and value retailers already compete on price movement, so algorithmic pricing aligns with their model. Brands built on full-price integrity gain more from using AI to set the right initial price and buy the right depth, which reduces the need for markdowns in the first place. Dynamic pricing is a tool for managing mistakes as well as opportunities; the better the upstream decisions, the less of it a brand needs.
Whatever the model, the decision about how far algorithms may move prices is a governance question. It belongs with the people accountable for brand, margin and legal compliance, and it should be written down before a system goes live, not discovered afterwards in a customer complaint or a regulator's letter.
Frequently asked questions
Do fashion brands use dynamic pricing?
Many fashion retailers use algorithmic tools for markdown and clearance decisions, and online channels allow more frequent changes. A published example is Zara, whose clearance pricing model raised clearance revenues by about 6% in a 2008 field experiment. Real-time pricing on full-price lines is less common because of brand and trust risks.
Is personalised pricing legal in the EU?
Personalised pricing is not banned outright under the Omnibus Directive, but traders must inform consumers when a price has been personalised on the basis of automated decision-making. Data protection and unfair commercial practice rules also apply, so legal advice is essential before any implementation.
What is the 30-day rule for discounts in the EU?
Under Article 6a of the Price Indication Directive, inserted by Directive (EU) 2019/2161, an announced price reduction must show the prior price, meaning the lowest price applied in a period of at least 30 days before the reduction. The rules apply from 28 May 2022.
Does dynamic pricing damage luxury brands?
It can, because consistent pricing is part of how luxury signals value. Frequent automated discounts can train customers to wait and weaken full-price sales. A cautious approach for luxury is to limit algorithmic pricing to controlled clearance channels, or not to use it at all.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
SOURCES
- IDEAS/RePEc: Clearance Pricing Optimization for a Fast-Fashion Retailer (Caro and Gallien, Operations Research 2012)
- Harvard Business School Working Knowledge: How do you predict demand and set prices for products never sold before?
- EUR-Lex: Directive (EU) 2019/2161 on better enforcement and modernisation of Union consumer protection rules
- US Federal Trade Commission: Surveillance pricing update and the work ahead
- American Bar Association: When pricing gets personal
