How Shein uses data and algorithms: the on-demand model and its criticism

KEY TAKEAWAYS Summary by the editors

1. Shein describes an on-demand model in which new designs are ordered in small batches of 100 to 200 pieces to measure demand, then retired or replenished based on customer demand signals.
2. Shein says its supplier partners use its proprietary technology for real-time demand measurement, production planning and supply chain visibility.
3. Shein claims its model reduces inventory waste to single-digit percentages; this figure is a company claim that has not been independently audited in public sources.
4. In May 2025 EU consumer protection authorities said Shein breached EU consumer law through practices including fake discounts, pressure selling and misleading sustainability claims.
5. In April 2024 an artist filed a proposed class action in New York alleging that Shein uses algorithms and AI to find and copy trending designs; Shein has denied similar claims in other cases.

Shein uses data and algorithms to run what it calls an on-demand production model: it orders new designs in small batches, measures how customers respond online and passes those demand signals to suppliers, who scale up the items that sell. The approach is the core of Shein's business, and it is also at the centre of criticism from regulators, designers and sustainability advocates, who argue that the same speed and data-driven marketing encourage overconsumption and copying.

What has Shein built?

In its own description of the model, Shein states that "new designs are ordered in small batches of 100-200 pieces to measure demand" and that "customer demand signals guide whether designs are retired or replenished, and inform new designs". Its supplier partners, according to the company, "use our proprietary technology, which provides real-time insights into demand measurement, production planning, and supply chain visibility".

In practice this combines three elements:

• Test and repeat: very small first orders across a large number of new designs.
• Demand sensing: online behaviour on Shein's own app and site used as the signal for what to reorder.
• Supplier integration: a shared technology platform that sends those signals directly to manufacturers.

Why did Shein invest in data and algorithms?

The commercial logic is risk reduction at the level of the individual design. Traditional fashion buying commits to large quantities months before a season, and unsold stock is marked down or destroyed. By ordering very small quantities first, Shein limits the cost of a design that fails and only commits to volume when there is evidence of demand. Shein presents this as a sustainability benefit, saying it reduces inventory waste "to single digit percentages, far below the industry average".

Analysis: the model is also a growth engine. Testing many designs cheaply produces a constant stream of new products, and online engagement data shows quickly which ones to push. That is the part critics focus on: a system optimised to find what sells and produce more of it quickly can lower waste per design while still increasing the total number of garments made and sold.

Read also

How Levi Strauss & Co. uses AI: the data programme and its lessons

How does it work (data, models, process)?

Shein has not published technical details of its forecasting or recommendation systems, so the mechanics are known mainly from its own summary and from allegations in litigation. The table separates what Shein states from what others allege.

In a proposed class action filed on 5 April 2024 in the US District Court for the Southern District of New York, artist Alan Giana alleged that Shein uses "algorithms, AI, and related computerized monitoring systems" to identify trending designs online and transmit them to factories. These are allegations in a lawsuit, not findings of a court, and in response to an earlier lawsuit by three designers, reported by NBC News, Shein said it "takes all claims of infringement seriously" and would vigorously defend itself.

What results has Shein reported?

Shein's main published result for its model is the claim of single-digit inventory waste. We found no independent audit of that figure in public sources. Shein has not disclosed data on forecast accuracy, the share of tested designs that are scaled up or the environmental footprint per garment that would allow outsiders to test its sustainability argument.

What are the limits and open questions?

Regulatory pressure is the most concrete limit. On 25 May 2025 the EU's Consumer Protection Cooperation network, as communicated by Ireland's CCPC, said Shein breached EU consumer law through:

1. fake discounts not based on genuine prior prices;
2. pressure selling with false purchase deadlines;
3. missing information on consumer rights such as returns and refunds;
4. deceptive labels presenting legally required features as special;
5. misleading sustainability claims;
6. hidden contact information for customer service.

In July 2025 FashionUnited reported that France's consumer authority DGCCRF had fined Shein 40 million euros for deceptive practices, finding that 11 percent of advertised discounts were in fact price increases, and that a 150 million euro fine from the data protection authority CNIL over advertising cookies was pending a final decision.

Analysis: these cases concern marketing, pricing and data protection rather than the production algorithm itself, but they matter for the model. An on-demand system only reduces waste overall if it does not simply stimulate more buying, and a low unsold rate per design says little about total volume produced. The design allegations raise a separate question: when demand sensing relies on what is trending online, where does inspiration end and copying begin?

Read also

How adidas uses AI: from design archives to generative content

What can other fashion companies learn?

Several principles from the on-demand model are relevant well beyond ultra-fast fashion, provided they are applied with stronger safeguards:

• Small initial buys and fast reorders reduce the risk of each new style, if suppliers can produce small runs economically.
• Shared demand data with suppliers shortens reaction times more than any single forecasting model.
• Claims need evidence. Sustainability and pricing claims based on data must be verifiable, or they become a regulatory liability.
• IP checks belong in the pipeline. Any process that turns online trend signals into products needs explicit design and rights review.

Frequently asked questions

What is Shein's on-demand business model?

Shein orders new designs in small batches of 100 to 200 pieces, measures how customers respond and then retires or replenishes each item based on demand signals. Its suppliers use Shein's technology to receive those signals and plan production.

Does Shein use AI to design clothes?

Shein says customer demand signals inform new designs but has not published details of the systems it uses. A 2024 lawsuit in New York alleges that Shein uses algorithms and AI to find and copy trending designs; these are allegations that have not been decided by the court in the sources we reviewed.

Is Shein's model more sustainable?

Shein claims its approach keeps inventory waste in single-digit percentages, but this has not been independently verified in public sources. Critics argue that low unsold stock per design does not offset the high overall volume and consumption the model encourages.

Why has Shein been fined in Europe?

In July 2025 France's DGCCRF fined Shein 40 million euros for deceptive commercial practices, including misleading discounts, and a 150 million euro fine from the data protection authority CNIL over cookies was reported as pending. In May 2025 EU consumer authorities also found breaches including fake discounts and misleading sustainability claims.

GuideThe complete guide to AI strategy for fashion companiesRead the complete guide