Long-Term Thesis

The 5-to-10-year underwriting question

Yatsen is a sub-scale, founder-controlled Chinese beauty house that has just completed the first leg of a deliberate transformation from a high-growth, cash-burning color-cosmetics company into a skincare-led multi-brand platform. The 5-to-10-year question is not whether the FY2025 turn happened — gross margin rose to 78.2%, revenue grew 26.7%, and the company printed its first non-GAAP profit since IPO. The question is whether that turn compounds into a structurally profitable branded skincare platform at roughly Proya-scale economics (US$1.0–1.5B revenue, mid-to-high single-digit non-GAAP operating margins) by 2030–2032, or whether platform pricing, sub-scale costs, and founder-controlled governance lock it permanently into the loss-making fringe of the world's most contested beauty market.

Read this page as an underwriting frame, not a near-term catalyst calendar. The reinvestment runway is real; the durability of the moat is not yet proven; and the gap between today's −4.3% GAAP operating margin and the ~17% earned by direct peer Proya on the same business model is the entire bull-versus-bear debate, expressed as one number.

Underwriting verdict — four dials. Thesis strength: Medium. Durability: Low-to-Medium (mix shift is real; segment-level profit is not yet earned). Reinvestment runway: Medium-to-High (capital-light model in a 10%-CAGR TAM, ¥1.05bn cash + US$120M convertible inflow). Evidence confidence: Medium — the FY2025 print is real, but Q1 FY2026's snap-back into wider losses on Douyin traffic-acquisition costs is a live counter-signal that has yet to be refuted.

Reporting note. Yatsen reports in Renminbi (RMB / ¥); USD references use Yatsen's stated convenience rate of ¥6.9931 per US\$1 at FY2025 close and ¥6.8980 per US\$1 at Q1 FY2026 close. ADS = 20 ordinary shares.

Five things that have to be true by 2030

A 5-to-10-year long position in YSG is a bet that all five of these are true. Each is independently necessary; together they describe the difference between a successful skincare-led re-platform and a stalled mid-cap beauty footnote.

No Results

The PM lens: only Condition 1 and Condition 2 are within management's control quarter-to-quarter. Conditions 3 and 5 are partially management-controlled but materially exposed to two third parties (the Eve Lom / DR.WU underlying brands, and ByteDance / Alibaba pricing). Condition 4 is binary on founder behaviour. Condition 6 is a tail risk that has not been priced down by the 96% drawdown — it has just been re-tested twice.

The reinvestment runway — the upper bound of the opportunity

A 5-to-10-year long underwrites the size of the prize first. For YSG that prize is the addressable China skincare market, compounding at ~9% in dollars, into which a 50%-skincare house with rising brand mix can grow ~3x revenue with capex below 1.5% of sales.

TAM and the per-capita gap

No Results

China was already the world's largest beauty market by 2019 (US\$38.8B retail-sales value) and CIC projected expansion to US\$68.7B by 2025 at a 10.0% CAGR — three times the US growth rate [1]. Within that, the facial skincare market (where Yatsen is now 53% concentrated) was sized at US\$51.0B by 2025 at 8.8% CAGR [1]. Crucially, China's per-capita beauty spend in 2019 was just US\$27.7 vs US\$96.6 in the US, US\$152.8 in South Korea and US\$184.4 in Japan — every comparable cohort suggests China spends multiples more on beauty as GDP per capita rises through US\$15,000 [2].

The next decade's structural tailwinds were already in place at IPO:

Domestic brands have been the share-taker. Domestic share of mass and mid-end beauty in China rose from 24.4% in 2010 to 46.3% in 2019, ahead of western brands at 38.3%, and Japan/Korea precedent says that, once domestic wins this market, it tends to keep it [3].
Channel is uniquely social-commerce native. Social e-commerce was projected to reach 47.3% of China online retail by 2025 vs 3.7% in the US — meaning the playbook is built for the platform mix YSG operates in, not the legacy Sephora/department-store playbook of multinationals [4].
Regulation favours scale. The NMPA cosmetics regime in force since January 1, 2021 imposes registrant responsibility, formal product registration, and monetary penalties up to 30x the value of non-compliant products — a floor under brand launches that disadvantages cottage challengers and advantages established multi-brand operators [5].

The capital-light shape of the model

The runway is not just demand-side; the model itself is light enough to compound on relatively modest equity. Capex has run at RMB 44–59 million annually for three years — just 1.0% of FY2025 revenue — so the question is operating-margin runway, not capacity capex. R&D commitment has been deliberate: R&D expense above 3.0% of revenue every year since 2022 (3.3% / 3.2% / 3.2% across FY2023–FY2025), with a patent portfolio of 269 items at end-2025 and a manufacturing-and-R&D campus of 66,462 sqm built jointly with Cosmax (the world's largest contract cosmetics manufacturer) that came online in August 2023 [6]. Strategically, management has codified three pillars — R&D-led innovation, multi-brand equity, and margin expansion — as the durable framework since 2022 [7].

What the runway looks like as a glide path

The skincare-share line is the single most important picture on this page. It has moved from 3.8% in 2020 to 53.0% in 2025 — a three-year skincare brand CAGR of 22.4%, against company-wide revenue at flat-to-negative growth — and it carried gross margin from 64.3% to 78.2% over the same window [7]. The illustrative base path takes skincare share to 60% in 2027 and 70% in 2030 — a deceleration from the post-2022 pace, on the view that Color Cosmetics revenue stabilises at the FY2024–25 ~¥2.0B floor while Skincare continues to scale on Galénic and DR.WU. Even at that trajectory, gross margin gains taper to ~80% — close to global prestige ceilings — meaning the next 500 basis points of consolidated margin have to come from the cost line, not the mix line.

The history that drives the underwriting

Five years of public history are now long enough to draw a track record from. The cycle has lived three lives; the next five years are where the fourth gets written.

Three multi-year facts an underwriter should commit to memory. First, revenue compressed from a ¥5.84bn peak in 2021 to a ¥3.39bn trough in 2024 — a 42% peak-to-trough collapse over three years [8]. Second, gross margin expanded every single year through that drawdown, rising 14 percentage points from 64.3% (2020) to 78.2% (2025) [9]. Third, GAAP operating margin has never been positive in YSG's public history — even at the 2021 revenue peak the company ran a 28% operating loss; the company is not a once-profitable business stumbling, it is a still-pre-profit business inflecting. The investment debate is therefore whether the FY2025 inflection is the bottom of the J-curve or a head-fake.

The transformation as management has framed it

Management has been explicit, on the record, since FY2022 about what they are doing: "Starting early 2022, we have carried out a comprehensive strategic transformation plan…driving R&D-led product innovation…rapid expansion of our skincare brands…consistent margin expansion through stricter pricing and discount policies" [7]. The track record against that framing:

Pillar 1 — R&D commitment. Delivered. R&D held at or above 3.0% of revenue across the trough; new R&D facilities added in Shanghai (May 2024) and France (Galénic); 269-item patent portfolio at end-2025 [6].
Pillar 2 — Brand portfolio. Mostly delivered. Skincare contribution went from 33.5% (2022) to 53.0% (2025) at a 22.4% three-year CAGR [7]. The pillar's weak link is Eve Lom, which has needed two consecutive goodwill impairments; the strong link is Galénic, which is now the FY2025 growth engine.
Pillar 3 — Margin expansion. Partially delivered. Gross margin lifted 1,020 basis points from FY2022 to FY2025 [9]. But GAAP operating margin still negative, and Q1 FY2026 stepped backward on Selling & Marketing intensity.

CFO Donghao Yang has been explicit about which lever drives the next leg: "with Skincare business, the gross margin, net margin are typically much higher than Color Cosmetics brands. So by doing that, we're going to be able to improve our margin profile. And secondly, our top line will continue to grow this year. And as a leveraging effect, we do believe that our net margin will improve accordingly" [10]. That is the management thesis, exactly as it would have to play out for a 5-year long to work.

The peer benchmark — the entire bull-bear gap, in one chart

The single most useful 5-year reference point is not the global majors (Estée Lauder, L'Oréal, Shiseido) — those are scale references, not substitutes — but Proya, a China-listed, digitally-led, multi-brand skincare-led peer that is two to three years ahead of Yatsen on exactly the same playbook.

No Results

Two things this chart says. First, Yatsen has higher gross margin than every comparator — gross-margin position is not the problem. Second, Yatsen and Proya share a business model (digitally-led, multi-brand, China-domiciled, skincare-anchored, no shelf moat), yet earn 22 operating-margin points apart. Proya printed FY2025 net income of approximately ¥1.50 billion on ¥10.6 billion revenue at 17.6% operating margin and 24.4% ROE [11], while leading multiple Tmall sub-categories (#1 face cream, #1 mask, #2 sets, #3 serum in 2025) [12]. YSG has comparable gross margin but no disclosed category leadership. The bull case is "Yatsen is Proya, two years late." The bear case is "Proya already took the share Yatsen needed."

Scenario fan — what mid-cycle profit looks like

The scenario fan is the actual 5-year debate. A single assumption — mid-cycle non-GAAP operating margin three years out — moves implied operating profit by ¥1 billion across the range real history makes plausible. At Proya's current margin, Yatsen earns ~¥950M EBIT against a market cap of ~US\$295M (≈¥2.06B); at half-Proya, ~¥540M; at the FY2025 actual non-GAAP margin of −2%, nothing. Every other input — revenue growth, share count, valuation multiple — moves the answer by far less. The size of the prize is large; the path is narrow.

The capital allocation track record — five years, four distinct chapters

A 5-to-10-year holder is underwriting the operator, not just the cycle. Yatsen's public capital-allocation record is five years long and tells four chapters in sequence.

No Results

The aggregate is unusual for a pre-profit company: 40.2 million ADSs repurchased for US\$202.2 million through February 28, 2026 under four successive authorisations (US\$100M → US\$150M → US\$200M, then a fresh US\$30M in May 2025) [13]. Total IPO proceeds deployed to buybacks reached US\$499.3M by year-end FY2025 [14]. On a 93.9M-ADS base today, that retired roughly 30% of the post-IPO float and meaningfully concentrated the remaining holders.

Two honest reads on this record:

Discipline argument. Returning IPO equity to shareholders at average prices well below the FY2021 peak was a defensible signal that management did not spend the cash on further low-return marketing or acquisitions. Capital allocation revealed an underlying preference for shrinking equity over redeploying into a broken-economics business.
Funding argument. None of this was funded from operating cash flow — operating cash flow has been negative in every year since FY2022 ([−¥107M / −¥244M / −¥95M] in FY2023-FY2025). The buybacks were funded by dwindling IPO cash, and the consequence was that liquidity fell from ¥2.59B at end-FY2022 to ¥1.05B (US\$150.7M) at end-FY2025, a 60% reduction in three years. This was a self-liquidating buyback, not a sustainably-funded one.

The March 2026 convertible — a governance turning point

The single most important capital-allocation decision in the five-year record happened in March 2026. Yatsen signed a definitive agreement on March 11, 2026 with an investment vehicle affiliated with Trustar Capital and Mr. Jinfeng Huang (the founder and CEO himself) to issue approximately US\$120 million of RMB-denominated convertible senior notes plus warrants [15]. The structural terms matter for an underwriter:

Coupon and tenor: 1.5% per annum; First Note 364 days extending automatically to 5 years upon receipt of an NDRC certificate; Second Note 5 years from issuance.
Conversion: at US\$4.63 per ADS (a 46% premium to the recent ~US\$3.17 spot), beginning 364 days from issuance.
Warrants: entitle the holder to acquire ordinary shares equal to one-tenth of conversion shares at the equivalent of US\$10.00 per ADS — long-dated upside with very low strike risk.
Put protection: holders may require the company to repurchase notes on the third anniversary of the First Note at a price reflecting a 4% internal rate of return to the holder — a floor that protects the founder vehicle from price downside [15].
Friction: a "significant shareholder" formally objected to the transaction following announcement [15]. Hillhouse subsequently joined the first closing, completed May 21, 2026, alongside Trustar and the founder vehicle.

Three reads, in declining order of generosity. (i) A position of choice. The financing came after the FY2025 non-GAAP turn, not before; the balance sheet still had ¥1.05B of cash. So management was funding the next phase of brand investment from a position of relative strength, with a smart-money anchor (Hillhouse, Trustar) rather than a dilutive equity offering. (ii) Real validation, real dilution. Hillhouse choosing to participate after publicly objecting is the most credible third-party endorsement of the strategy in the public record — and the structure is convertible, not common, so existing equity is not immediately diluted. (iii) An overt self-deal element. The founder's own vehicle is among the lenders, with a 4% IRR put floor that effectively underwrites his downside while the public float bears it. This is precisely the friction the dual-class voting structure (Class B = 20 votes, 90.7% / 34.3% voting / economic split) enables [16]. A 5-year holder has to make peace with that structure, because nothing about the share class architecture changes within the underwriting window.

The five failure modes that survive contact with the bull thesis

A long-term thesis is built as much on what would break it as on what would make it work. Five durable failure modes deserve real weight.

No Results

Failure mode #1 deserves separate emphasis

The single biggest threat to the 5-year thesis is the platform-pricing mechanic, because Yatsen has already demonstrated it can flip the operating margin by six percentage points in a single quarter. In Q1 FY2026, revenue grew 22.5%, skincare grew 58.5%, gross margin hit an all-time-high of 80.2%, and the operating loss margin widened from −4.1% to −9.7% because Selling & Marketing rose to 72.2% of revenue (from 66.4%) — attributed by management explicitly to "higher traffic acquisition costs on the Douyin platform" [17]. A wide-moat consumer brand does not see its operating margin set quarter-by-quarter by a third party. Until Yatsen demonstrates two-to-three quarters of S&M intensity holding below ~65% of revenue while gross margin continues to expand, the underwriting case for an EBIT-margin runway above 5% has not been proven — only narrated.

Multi-year watch signals — the quarterly PM scorecard

The five-year thesis is right if and only if a specific set of multi-year signals trend the right way. The scorecard below is what a PM should re-rate against every reporting cycle. Not all signals are equal in weight; the bold ones are the load-bearing inflection variables.

No Results

A practical reading guide: the load-bearing signals are #2 (non-GAAP OM), #3 (S&M intensity), and #7 (OCF). If all three trend the right way for four consecutive quarters, the 5-year thesis is working and the scenarios shift up. If signal #3 stays above 70% of revenue for two consecutive quarters, the 5-year thesis is broken — at that point, gross-margin gains are being given back as fast as they are earned and the structural moat has been disproved.

What would change my mind

A high-conviction long needs both a confirming set and a disconfirming set, both specific enough to be falsifiable.

No Results

The synthesis

Strip away the 24-month noise and what an underwriter is left with is this:

The size of the opportunity is real. China beauty is a US\$68.7B market growing ~10%, skincare specifically is US\$51B growing ~9%, domestic brands have already taken decisive share, the model is capital-light at ~1% capex, and the company has built a credible R&D footprint (US\$100M cumulative, 269 patents, Cosmax JV, three R&D centres).
The transformation is real. Five years of public history show gross margin up 14 percentage points, skincare share up 49 percentage points, and the operating-margin deficit narrowed from −51% (2020) to −4.3% (2025). The FY2025 non-GAAP profit is real, even allowing for the absence of FY2024's impairment charge.
The path is narrow. Two third-party variables — platform pricing and brand-equity durability — have already produced backslides (Q1 FY2026 traffic costs; two Eve Lom impairments). The 22-point operating-margin gap to Proya is what proves whether Yatsen is a sub-scale also-ran or an early-cycle compounder.
The governance overlay is the un-fixable. A founder controls 90.7% of votes from 34.3% of economics, has just arranged a self-funded convertible with a 4% IRR put floor, and operates inside a controlled-company NYSE exemption. A 5-year holder is signing up to that structure for the entire underwriting window.

The cleanest single statement of the underwriting question: "is Yatsen capable of earning a Proya-half operating margin by 2030 on the brand portfolio it already owns, without further dilutive financing or a third goodwill impairment?" A yes makes today's ~0.24x EV/Sales and ~0.7x book look hard to justify. A no makes the FY2025 inflection a cyclical print on a structurally pre-profit business carrying a non-trivial probability of a binary downside event over the holding period. The honest mid-case answer is that two more reporting cycles will define it — which is why position-sizing discipline must lead the thesis, not the other way around.

References

Generated Jun 19, 2026 10:16 UTC