A fast-fashion growth engine went to 1.1% and posted a quarterly loss — in its own Hong Kong listing prospectus (SHEIN, 24 Aug 2026)

Why it matters (spine). The garden's fashion and resale veins rest on a bet that item-level truth — what a garment is made of, what is in it, what condition it is in — becomes valuable as the ultra-cheap new-goods flow gets more expensive. This document is the other half of that bet, stated by one of the largest operators of that flow in the one place a company cannot spin: a listing prospectus, under exchange rules, with audited comparatives. No market-share figure appears on this card, because none appears in the sections read — "one of the largest" is a description, not a measured rank. And the non-toxic spine has a specific stake here, which this card has to report as an absence rather than a finding.

What the document is. "Global Offering" of SHEIN Global Holdings Limited (希音國際控股有限公司), "A company controlled through weighted voting rights and incorporated in the Cayman Islands with limited liability", stock code 00625, dated 24 August 2026 on the face of the IMPORTANT page, filed to HKEXnews. The offering: 279,992,500 Class B Shares under the Global Offering "(subject to reallocation and the Over-allotment Option)", split 27,999,300 Hong Kong Offer Shares and 251,993,200 International Offer Shares. "Maximum Offer Price: HK$49.50 per Offer Share plus brokerage of 1%, SFC transaction levy of 0.0027%, AFRC transaction levy of 0.00015% and the Stock Exchange trading fee of 0.00565%." Nominal value US$0.000002 per share. On timing: "The Price Determination Date is expected to be on or before Friday, 28 August 2026 and, in any event, not later than 12:00 noon of Friday, 28 August 2026. The Offer Price will be not more than HK$49.50 and is currently expected to be no less than HK$47.60 per Offer Share unless otherwise announced."

The financial record, as printed (US$ millions; years to 31 December, plus the three months to 31 March):

	2023	2024	2025	Q1 2025	Q1 2026
Total net revenues	32,103	38,748	41,847	8,952	9,052
— Product revenues	31,235	35,351	37,107	7,765	7,757
— Service revenues	868	3,397	4,740	1,187	1,295
United States	9,454	10,464	10,101	2,380	2,040
Europe	10,214	13,603	14,802	2,843	2,908
Rest of the world	12,435	14,681	16,944	3,729	4,104
Apparel	22,080	25,068	26,707	5,548	5,562
Others	10,023	13,680	15,140	3,404	3,490
Net income / (loss)	2,789	3,365	2,064	395	(99)

And the company's own key ratios table: net revenues growth 41.1% · 20.7% · 8.0% · N/A · 1.1%; operating margin 4.3% · 2.5% · 4.1% · 3.9% · 2.9%; net margin 8.7% · 8.7% · 4.9% · 4.4% · (1.1)%. The Q1 2026 column is marked "(unaudited)".

What the numbers say when you read them together. Four things, none of them requiring interpretation beyond arithmetic:

• Growth halved, then more than halved, then all but stopped: 41.1 → 20.7 → 8.0 → 1.1. The step ratios are 0.50, 0.39 and 0.14 — only the first step is "roughly half", and saying otherwise would flatter the shape of the decline.
• Product revenue is now shrinking — US$7,765m to US$7,757m year on year in the March quarter — while service revenue grew from US$1,187m to US$1,295m. All of the (slight) top-line growth came from services; the goods business went backwards.
• The United States is in absolute decline, 2024 to 2025 (10,464 → 10,101) and much more sharply in the quarter (2,380 → 2,040, a fall of US$340m). Rest of the world rose to 45.3% of quarterly revenue.
• Apparel is a shrinking share of the mix: 68.8% → 64.7% → 63.8% of annual revenue. The like-for-like quarterly pair is 62.0% (Q1 2025) → 61.4% (Q1 2026), a move of about half a percentage point — smaller than the annual trend, and reported here as its own comparison rather than spliced onto the end of the annual series.

Enumerated so a missing item is detectable. Derived in this section: the three subtractions (7,765 − 7,757; 10,464 − 10,101; 2,380 − 2,040 = 340); the three step ratios (0.50, 0.39, 0.14); the three annual apparel shares (68.8, 64.7, 63.8); the two quarterly apparel shares (62.0, 61.4) and the 0.6-point difference between them; and the rest-of-world quarterly share (45.3). Derived elsewhere on this card: the eight ratio recomputations in the arithmetic-audit section below, the two counts stated there ("six of the fourteen printed ratios", "four of the five periods printed"), and the two gross-proceeds figures under Money angle. Everything else is read off the printed tables.

The cause, in the company's own words. "Following the removal of the U.S. de minimis exemption and the EU's EUR 150 customs duty exemption for low-value consignments, we adopted formal customs clearance procedures and a framework of price increases, localised inventory and fulfilment, and enhanced trade compliance in the U.S., with similar measures expected in the EU. Since May 2025, the removal of the U.S. de minimis exemption has had an adverse impact on our sales in the U.S. and the overall growth of our net revenues, and has contributed to an increase in our fulfilment expenses as a percentage of net revenues, although we have since observed signs of normalisation in consumer purchasing behaviour and sales trends in the U.S." On Europe: products that previously relied on the EUR 150 exemption "are subject to increased costs, which may have a material adverse effect on our business", with the company noting it generated "approximately one-third of our net revenues from Europe in 2025 and for the three months ended 31 March 2026." Its own hedge, quoted: "Although it remains too early to fully assess..."

Arithmetic audit, run and reported — including what could not be checked. The ratios that can be checked against the printed lines all reconcile: 38,748 ÷ 32,103 − 1 = 20.7%; 41,847 ÷ 38,748 − 1 = 8.0%; 9,052 ÷ 8,952 − 1 = 1.1%; net margins 2,789/32,103 = 8.7%, 3,365/38,748 = 8.7%, 2,064/41,847 = 4.9%, 395/8,952 = 4.4%, −99/9,052 = −1.1%. Six of the fourteen printed ratios cannot be checked here and are quoted on the company's authority: the 41.1% growth rate for 2023 needs a 2022 revenue line that is not in the table, and all five operating margins need an operating-income line that is not either. Both live in the Financial Information section, which was not read. The table holds fourteen ratio figures in total — four growth rates, five operating margins, five net margins — of which the eight recomputed above check out.

One oddity is flagged rather than explained: net margin exceeds operating margin in four of the five periods printed — 2023 (8.7 vs 4.3), 2024 (8.7 vs 2.5), 2025 (4.9 vs 4.1) and Q1 2025 (4.4 vs 3.9) — with only Q1 2026 inverting. That implies material non-operating income across the whole period, not just the early years. The sections read here do not explain it, and this card does not guess. Reconciling it needs the income statement.

Where this touches the garden's resale thesis, and where it does not. The cabinet's resale work argues that item-level data becomes the defensible layer as goods circulate. This document supplies the missing pressure term — the cheap new-goods channel's growth being removed by trade policy, not by consumer conscience — which is a more durable driver than sentiment, because a customs rule does not change its mind. But three cautions belong on the card. The garment volume is not disclosed here, so "cheaper clothes got more expensive" is an inference from a revenue table, not a measured price or unit figure. The company reports "signs of normalisation", so its own read is a shock absorbing rather than a permanent reset. And the shrinking apparel share may reflect diversification into other categories rather than any decline in clothing sold.

On non-toxicity, fibre composition and condition: an absence, reported as an absence. The sustainability language in the sections read is at the level of principle, not data: a "technology-enabled, demand-driven operating model that features small-batch production" which "allows us to minimise overproduction, reduce inventory waste", and initiatives "covering responsible design, material sourcing, manufacturing, circularity programmes and decarbonisation goals." No fibre-composition figure, no restricted-substances or chemical-testing disclosure, no non-toxicity claim and no take-back or resale programme detail appears in any passage read for this card. That is an absence in what was read — an extraction of the front matter, summary and risk sections — and explicitly not a claim that no such disclosure exists elsewhere in a document of this size. The garden's three untouched data layers remain untouched by this card.

Money angle and metacrisis lever. Money angle, stated as an option rather than a forecast: a listing turns a very large ultra-fast-fashion operator into a company on a disclosure calendar, which converts a private black box into a twice-yearly primary source for the fashion vein — that is worth more to RDKG than the offer price. At maximum HK$49.50, the 279,992,500 shares imply gross proceeds before expenses of about HK$13.86bn; at the HK$47.60 lower end, about HK$13.33bn. Both figures are derived on this card, not printed in the document, and exclude the over-allotment option and all fees. Metacrisis lever: in my judgement this is the cabinet's most direct example of an environmental outcome arriving through a customs rule rather than an environmental policy. The de minimis change was trade policy; its recorded effect is a contraction in a high-volume, ultra-low-price clothing channel — whose scale this card does not quantify, because the document as read gives no unit, volume or share figure. Whether that is a good outcome for materials throughput is untested here for the same reason.

What would prove it wrong (falsifier). Three limbs.

• The figures fail if this document is superseded, corrected or restated, or if the printed table differs from what is quoted — the Q1 2026 column is marked unaudited, which is the weakest number on the card and also the most quoted.
• The offering fails if it does not proceed. The document says so itself: "If, for any reason, we do not agree with the Sponsor-Overall Coordinators ... on the pricing of the Offer Shares, the Global Offering will not proceed and will lapse." The price determination date is on or before 28 August 2026 — today, as this card is written — so the outcome is unknown here and no listing is asserted.
• The deceleration-as-trend reading fails if the next two reporting periods show growth reaccelerating, which the company's own "signs of normalisation" language invites. One quarter at 1.1% against a tariff shock is a data point; the trend needs the first post-listing interim report.

Provenance (what I actually read). The offering document PDF at www1.hkexnews.hk/listedco/listconews/sehk/2026/0824/2026082400019.pdf, fetched from the main thread on 28 August 2026 and read as a saved text extraction. Two independent date signals: the HKEXnews path segment /2026/0824/ and document ID prefix 20260824, which are the exchange's own machine-generated filing convention, and the line "24 August 2026" printed on the IMPORTANT page of the document itself. I read a partial extraction, not the whole prospectus — approximately the cover and IMPORTANT pages, the SUMMARY section (the financial and key-ratio tables above carry the page marker "SUMMARY – 18 –"), the business-highlights and sustainability bullets, and parts of the risk-factors and management-discussion text. Every quotation of the offering document, and every printed figure, comes from that extraction. Two cabinet cards were also opened in the repository and quoted or characterised directly: ebay-depop-acquisition-closed (30 July 2026), the source of the "a gap worth testing, not an established fact" hedge quoted in so_we_should, and ebay-q2-2026-provenance-layer-depop-price, the source of the five-layer data-layer taxonomy and of which two layers an incumbent has built. I did not read the Financial Information section, the Appendix I accountants' report, Business, Regulatory Overview, Underwriting, Structure of the Global Offering, or the appendices — which is why the operating-versus-net-margin oddity is flagged and not resolved, and why the absence of chemical and fibre disclosure is reported as an absence in what was read. The allotment results and final offer price had not been published as of this fetch; no figure on this card comes from wire coverage, and a share-count figure circulating in trade coverage as "280 million" is stated here as the document's exact 279,992,500.

Update note (1 September 2026, appended per L64 — the card above is unchanged). The trigger this card's so_we_should set has fired: the company's "Announcement of Final Offer Price and Allotment Results", dated 31 August 2026 on its face, priced the offering at HK$48.56 and is carded at shein-hk-listing-priced-allotment. Of the falsifier's three limbs below, limb 2 (the offering failing to proceed) is closed by that document; limb 1 (the figures being superseded or restated) and limb 3 (the deceleration-as-trend reading, testable at the first post-listing interim report) remain open. The "no listing is asserted" framing above was correct when written and is left standing.