Shein produces its first run of a new garment in batches of 100 to 200 units.
By any manufacturing logic, that is a terrible idea. Per-unit costs on a run that small are dreadful, setup barely amortizes, and no factory wants the order. Shein does it anyway, launching close to 4,700 new garment designs a day in the first quarter of 2026.
The reason shows up in one number that has nothing to do with fashion.
The number that explains the whole business
In 2025, Shein's inventory turnover ran at 36 days. Zara sat around 88 days. H&M sat around 138.
Turnover measures how long capital sits inside a garment before converting back into cash. At 36 days, a dollar cycles through the business roughly ten times a year. At 138 days, under three.
Two companies with identical margins and identical capital produce wildly different returns on that gap alone. Small batches are how Shein bought the speed.
Forecasting became measurement
Traditional apparel commits capital to a prediction. Buyers decide six to nine months ahead what will sell, place large orders to earn per-unit pricing, and then discover whether they were right. Being wrong shows up as markdowns, where the industry's margin quietly dies.
Shein's disclosed operating model, which it calls Large-scale Automated Test and Reorder, removes the prediction step. A tiny batch goes live, real customer behavior gets measured against it, and products that perform get restocked in as few as five days.
The 100-unit run is the price of an experiment. Shein accepts a worse per-unit cost on the first batch to buy information, then commits real capital only to demand it has already observed.
That trade shows up cleanly in the financials. Gross margin climbed from 60.2% to 67.9% across two years, driven substantially by charging closer to full price and clearing far less stock at a discount.
Fragmentation is the enabling infrastructure
Running thousands of 100-unit batches requires a supplier base almost nobody would design on purpose.
Shein worked with roughly 7,500 contract manufacturers in 2025, predominantly small and medium enterprises. Its largest single supplier accounted for 6% of total purchases, and the top five together came to 16.4%. Suppliers are non-exclusive and free to serve other clients.
Vertically integrated factories built for container-load economics cannot profitably accept a 150-unit order. Small workshops with idle capacity can, because that order fills a gap in the schedule.
Shein supplies those workshops with cloud software covering fulfillment, production, quality control, and shipping, then allocates orders algorithmically by expertise, price, and capacity. The technology's job is coordination across a deliberately fragmented network. Concentration would have made the network cheaper per unit and useless for the actual model.
The part the operating model never solved
Small batches produce small parcels, and small parcels cross borders one at a time.
For years that was close to free. The US de minimis exemption let shipments under $800 enter duty-free, and the EU had a comparable threshold. The entire cost advantage of shipping individual orders from Chinese warehouses to Western doorsteps rested on those rules.
The US removed its exemption in 2025 and the EU added a €3 charge on low-value e-commerce imports in July 2026. Chinese-origin goods Shein ships to American customers now face tariff rates between 10% and 87.5%.
The damage lands in one line. Fulfilment costs ran 42.1% of net revenue in 2023, 45.6% in 2025, and 47.7% in the first quarter of 2026.
Two directions at once
Shein's IPO filing is unusual because its two headline margins move in opposite directions.
Gross margin improved to 67.9%, evidence the LATR machine works exactly as designed. Net profit margin fell from 8.7% in 2024 to 4.9% in 2025 and to negative 1.1% in the first quarter of 2026, producing a $99 million net loss against a $395 million profit in the same quarter a year earlier.
Revenue growth tells the same story: up 20.7% in 2024 to $38.7 billion, up 8% in 2025 to $41.8 billion, and up 1.1% in the first quarter of 2026. US revenue fell 14.3%.
Marketing has climbed from 10.8% of revenue in 2023 to 15.8% in the first quarter of 2026, with order frequency flat near four purchases a year. Growth now depends on acquiring new shoppers, and the remaining headroom sits in markets with the lowest order values and fiercest competition.
The lesson is about which margins you own
Shein built a genuinely superior operating model. Turning demand forecasting into demand measurement is a real structural advantage, and the inventory numbers prove it against the best operators in the category.
That advantage was also sitting on top of a regulatory subsidy nobody at Shein controlled. When the subsidy disappeared, the operating model kept performing, and the P&L still broke.
Every business has both kinds. One comes from decisions you made and can defend. The other comes from conditions that happen to be true right now: a policy, a platform's algorithm, and an underpriced ad channel. The second feels identical to the first on a spreadsheet until the day it stops existing.
Shein priced its Hong Kong listing near $26.5 billion with both stories on the table.
What founders can take from this
If you have your own startup, here’s what you can learn from all this:
- Buy information before you buy inventory. An uneconomic first run that tells you the truth beats an efficient large run built on a guess.
- Track turnover, not just margin. How fast capital cycles often matters more to returns than what you make per unit.
- Design your supplier base around your batch size. Flexibility comes from partners with spare capacity, which usually means smaller and more numerous.
- Separate earned margin from granted margin. Write down which parts of your unit economics depend on a policy, platform, or subsidy you do not control, then model the version without them.
- Watch the line item nobody is optimizing. Shein's production side kept improving while Fulfillment ate the business.
Know an operator still forecasting demand six months out? Pass this on.
