Product Engine

Product Engine

Yatsen brands itself "R&D-led," and the science is not a fiction: certified labs, a 269-item patent portfolio, medical and university partnerships, and proprietary ingredients. But the spending and the intellectual property describe a company that mostly buys premium credibility rather than one that builds a defensible product moat. Product research is 3.2% of revenue (¥137.3 million) and has not passed its 2021 peak; selling and marketing outspends it 20.8 to 1; and the patent book tilts toward packaging design.

The marketing economics chapter established that Yatsen's entire operating-margin gap sits in the selling and marketing line — 66.3% of revenue, versus a profitable peer near 46%. That was the demand side: what the company pays to acquire a sale. This chapter is the supply side. The escape route the upside case needs — a premium, skincare-led mix that eventually needs less marketing per sale — has to be powered by something. Management says that something is research. The filings let us size it.

Twenty-to-one

The clearest way to read Yatsen's priorities is its own 2025 operating expense line, expressed against revenue.

Source: FY2025 Annual Report (Form 20-F), Consolidated Statements of Operations [1].

For every ¥1.00 Yatsen spent on research in 2025 it spent ¥20.80 on marketing. That ratio is not a one-year artifact of a growth push: it was 20.0x in 2023 and 20.8x in 2024 [2]. A model that acquires demand at twenty times the rate it develops product is, structurally, a marketing model with a research function attached — not the reverse.

Sources: FY2025 Annual Report (Form 20-F), Results of Operations [3] and Business — R&D Capabilities [4].

Rising as a share, flat as a sum

Read on its own, the research budget looks like a rising commitment: from 0.8% of revenue in 2019 to a 3.2% plateau since 2023 — roughly the ~3% of sales that L'Oréal, the industry's largest research force, spends [5]. By that intensity measure Yatsen looks normal for a science-led beauty company.

The absolute figure tells a quieter story. Research spending peaked in 2021 at ¥142.1 million and, four years later, at ¥137.3 million, has still not returned to it [6] [7]. The jump in intensity from 2.4% to 3.4% between 2021 and 2022 was mostly a denominator effect — revenue fell from ¥5.84 billion to ¥3.71 billion in the crash — not a spending surge [8] [9]. In US dollars the 2025 research budget is US$19.6 million.

Sources: FY2021 20-F (2019–2021) [10]; FY2024 20-F (2022) [11]; FY2025 20-F (2023–2025) [12].

Sources: research spend and ratios per the FY2021, FY2024 and FY2025 20-F Results of Operations [13] [14]; patent totals per each year's Business section [15] [16] [17] [18].

Patents that protect the label

The patent count has more than doubled since 2021 — 118 to 269 — which reads as an innovation engine gaining momentum [19] [20]. The composition points the other way. Of the granted patents, design patents — which protect the appearance of packaging, not a formulation — are the majority and the fastest-growing type: 74 in 2021, 158 in 2025. Invention patents, the category that covers proprietary chemistry, grew from 39 to 72 over the same span, staying under a third of the granted book.

Sources: granted-patent counts by type, FY2021–FY2025 20-F Business sections [21] [22] [23] [24] [25].

Yatsen is candid about where its defensibility lies. The FY2025 filing states plainly that "our most valuable intellectual property is our brand names," backed by 2,197 trademarks registered in China and 2,256 outside it [26]. A second qualifier sits in how the products are made: much of Yatsen's formulation work is co-developed with third-party original-design manufacturers — Cosmax, Intercos, Kolmar, Sensient Technologies and Naolys — the same contract labs that serve competitors [27]. Innovation routed through shared suppliers is harder to fence off than innovation done in-house.

What the premium credentials are

The bear reading should not be overstated, because the science infrastructure is real and it is being pointed at the right segment. Yatsen operates a global research center in Shanghai, a certified lab in Guangzhou, a joint laboratory with Ruijin Hospital, and a France facility for the Galénic brand, organized under a "1-3-4-6-20" framework of one manufacturing hub, three research centers, four focus areas, six joint laboratories and over twenty joint projects [28]. It runs partnerships with Huazhong University, Sun Yat-sen University and Lubrizol, and — through the Galénic acquisition — a long-term collaboration with Pierre Fabre, a major European dermo-cosmetics group [29].

Crucially, the acquired premium brands carry formulation heritage Yatsen did not have to build: DR.WU's clinical skin-renewal expertise and Galénic's Pierre Fabre link are genuine credentials, whatever the in-house patent mix looks like [30]. And the strategy is showing up on the one margin line research most directly touches. Gross margin climbed from 73.6% in 2023 to 78.2% in 2025, which management attributes to the growing skincare mix, a category it says "typically feature[s] higher gross margin and greater customer loyalty" than color cosmetics [31] [32].

Source: FY2025 Annual Report (Form 20-F), Results of Operations [33].

Management is explicit that research is the engine, not the accessory. On the Q4 2025 call, the chief financial officer named research as "the most important thing that we're going to do to grow our Skincare business," adding that over "the last 5, 6 years, we've been investing aggressively in our R&D capabilities" and that recent skincare growth "has largely been due to the contribution of our R&D team in terms of better products" [34]. Product wins like Galénic's VB serum and DR.WU's PDRN serum are offered as proof [35].

The read

The product engine is genuine but sub-scale, and on the evidence it is doing the work of lifting gross margin — not the work of lowering the cost of a sale. A US$19.6 million research budget, flat since 2021 and dwarfed 20.8 to 1 by marketing, is enough to differentiate a premium serum and support 78% gross margins; it is not, on its face, the kind of proprietary product advantage that lets a brand pull back on customer acquisition. The premiumization case therefore rests more on brand-building and acquired labels than on a self-built research moat — a distinction that matters because the operating-margin gap sits almost entirely in the marketing line (Marketing Economics).

The strongest fact against this read is on the gross-margin line: the mix shift is working, and management has tied it explicitly to research-led products [36]. What would change the read is a step the filings have not yet shown: the research budget scaling in absolute terms alongside the skincare mix, invention patents overtaking design in the granted book, and — the input the corpus cannot supply — evidence that the higher "customer loyalty" management claims for skincare actually translates into a lower marketing bill per sale. Until then, the product story is real enough to defend pricing, and too small to escape the marketing model.