Variant Perception

Where We Disagree With the Market

The market is pricing Yatsen as a melting Perfect Diary color-cosmetics business with a one-time goodwill-comp benefit and a pending dilution overhang — and the report's evidence says by Q4 2025 it had already become a 61%-skincare business whose convertible is convex, not concave. The embedded ¥19.64 / US$2.89 ADS price sits at 0.25× EV/Sales (vs a 2.1× peer median), reflects four years of institutional exit, and treats the Trustar/founder convertible struck 60% above spot as if dilution is certain rather than contingent on the bear case being right. Our three sharpest disagreements are with the segment frame, the quality-of-liability read on the convertible, and the time horizon assumed for operating leverage. No variant view says the stock is "cheap" — variant views say the path to resolution runs through three specific prints (Q1 2026 on May 14, 618 in mid-June, and Q2 2026 in mid-August), each of which is observable and dated. The first thing a PM should know is that the published sell-side target average of ¥51.97 / US$7.65 (+160% upside) is meaningless because the institutional buyer base has already left — the market price reflects the marginal seller, not the marginal sell-side analyst.

Variant Perception Scorecard

Reading the score. Variant strength of 58 is "real edge, not a layup." The evidence supports a material disagreement with embedded market pricing, but the two highest-leverage proofs (Q1 2026 S&M ratio and trailing-four-quarter OCF) are forward, not realized. Consensus clarity is the lowest input because the buyer base is fractured: the published sell-side average target ¥51.97 implies +160% upside, the tape says -73% from August 2025 high, and the price action since the December 2025 death cross says the marginal holder gave up. Evidence strength sits at 62 because three concrete pieces of report evidence (Q4 2025 skincare mix at 61.1%, the convertible struck at ¥31.5/ADS conversion price ≈60% above spot, and S&M growth +25.5% YoY trailing revenue +26.7% YoY) directly contradict the market's implied assumptions. Time-to-resolution of 4 months reflects the May 14 Q1 print plus the June 618 read; one clean confirmation print resets the debate, one disconfirmation locks the discount in.

Consensus Map

The institutional consensus and the sell-side consensus disagree with each other. Both views need to be on the map because the tape reflects one and the published targets reflect the other.

The signals that matter for our disagreements are issues #1, #2, and #3 — each has a single observable line item that resolves it, and the resolution windows are clustered between mid-May and mid-August 2026. Issue #4 (cash burn) is real but the convertible already addresses it for the bull case; issue #5 (thin coverage) is an institutional fact rather than an analytical claim; issue #6 (governance) is a permanent discount we do not contest.

The Disagreement Ledger

Disagreement 1 — Wrong segment. Consensus reads YSG as a Perfect Diary color stock and uses Coty (0.93× EV/Sales) or worse as the anchor. The math has already passed that anchor: Q4 2025 skincare was 61.1% of revenue, growing 51.9% YoY at gross margins above 80% — that is a Yunnan Botanée profile (2.63× EV/Sales) on the majority of the income statement. If we are right, the market has to concede that the multi-year skincare premiumization is the franchise, not a side-bet — and the appropriate multiple is a blend, not the worst-case anchor. The cleanest disconfirming signal is Q1 2026 skincare share dropping back below 50%, OR Galénic/DR.WU YoY growth decelerating into the single digits while color grows — neither of which would be consistent with the trajectory the report's data shows.

Disagreement 2 — Wrong quality of liability. Consensus treats the Trustar/founder convertible as 25–35% dilution on terms minorities cannot negotiate, and priced 27% out of the equity in three weeks. The misread: the conversion price is 60% above current spot and the warrants are 240% above spot, so dilution is contingent on the bear case being right — in the bull case, conversion lands at a stock price 60%+ above today, which is the holder selling into appreciation, not minorities being diluted at a low. If we are right, the market would have to concede that the convert is insurance financing (Audit-Committee-approved, founder-aligned) rather than forced financing, and the 27% drawdown should reverse on terms clarification. The disconfirming signal is the Second Note tranche closing at terms materially worse than the First Note (which would suggest the founder used the optionality to renegotiate down), or the NDRC certificate denied/delayed beyond Q3 2026, which would force First Note refinancing in 2027.

Disagreement 3 — Wrong time horizon. Bear case argues the S&M ratio failed to bend in "the best year of the cycle," concluding the moat does not reach the operating line. We argue the bear is reading the ratio rather than the rate of change: absolute S&M grew 25.5% in FY25 against revenue +26.7%, the first sub-revenue print in three years, and the Q4 spike to 64.8% was Double-11-specific KOL inflation, not structural reset. If we are right, the market has to concede that operating leverage is slow but live, and the first clean two-quarter sequence (Q1 + Q2 2026, landing by mid-August) is enough to reset the discount. The cleanest disconfirming signal is Q1 2026 S&M ratio above 67% on revenue growth at or below 18% — that combination would confirm marketing intensity is structural rather than cyclical and end the variant view in one print.

Disagreement 4 — Wrong quality of earnings on cash conversion. A modest variant view. Market anchors on -¥95M FY25 OCF; the data shows ~¥122M of deliberate inventory build for the skincare ramp consumed most of it. Strip the build and operating cash was roughly neutral in the worst quarter of the cycle. The materiality is lower because this is a translation correction rather than a structural call, and the resolution signal (FY26 H1 inventory days) is already on the standard watch list. We carry it as supporting context for the bigger three, not as a standalone investment thesis.

Evidence That Changes the Odds

The single most decisive piece of evidence on the table is line 4 — the rate-of-change read on the S&M line. The market is anchored on the ratio (66.3% FY25 vs 66.9% FY24 — "stuck") and the variant is anchored on the slope (S&M absolute growth dropping below revenue growth for the first time in 3 years). Both views are working from the same data; the only thing that resolves it is a second quarter where the slope continues. That is mid-August 2026.

How This Gets Resolved

The resolution stack is uncommonly dense in the next 100 days. Signals #1 and #2 land on May 14 (3 days from now), signal #3 lands in mid-June, and signal #7 lands in mid-August. Any two of those four going the variant way would materially challenge the discount; any two going against would close the variant view as wrong. Signals #4, #5, #6, and #8 are slower-cycle confirmation reads that matter only after the first two are in.

What Would Make Us Wrong

The disagreement closest to the bone is #1 — that YSG is already a skincare business priced as a color stock. The thing that could make us wrong here is not that skincare growth slows, but that the per-customer marketing cost of skincare turns out to be just as high as the per-customer cost of color. We have no published cohort retention data for Galénic, DR.WU, or Eve Lom — that is the single missing disclosure for the entire moat thesis. If skincare CAC equals or exceeds color CAC (because the platforms charge the same take rate regardless of basket type, and KOL pit fees do not discount for prestige), then the segment mix shift mechanically lifts the gross margin but leaves the operating margin trapped at the same -4% line we see today. The bull case asks the reader to trust that prestige-skincare repeat economics are structurally better than mass-color repeat economics. That is plausible but unproven from public disclosure, and the absence of disclosure is itself evidence — a beauty company with strong cohort retention typically discloses it. Yatsen has had eight years to publish a single cohort number and has not.

The disagreement second-closest to the bone is #2 — that the convertible is convex financing not flat dilution. The thing that could make us wrong here is that the Second Note tranche closes at substantially worse terms than the First Note, or that the NDRC certificate is denied, or that the founder uses the optionality from the 90.7% voting control to renegotiate the conversion price down at a future cycle low. None of those would be malfeasance — they would simply be the rational use of control by a controlling shareholder. The risk-factor cross-reference in the 20-F already flags that the financing exists because operating cash is insufficient, which is the company's own statement that this is bridge capital, not offensive capital. If the bridge needs to be re-priced, the bear's "minority dilution at terms minorities cannot negotiate" read becomes the right read.

The disagreement furthest from the bone is #3 — that operating leverage has started, slowly. The thing that could make us wrong here is platform take-rate inflation. Tmall, Douyin, RED, JD set the rules on 84.9% of YSG's revenue; a 200 bps take-rate hike costs ~¥75M of margin on the FY25 base, which is roughly the entire FY26 operating-leverage budget the bull case is underwriting. We have no visibility into platform pricing, no ability to forecast it, and no offsetting hedge inside the company's distribution model. The Q4 2025 S&M ratio jump to 64.8% may be the start of that story rather than a one-cycle KOL inflation episode. We cannot tell from a single quarter, and a second consecutive elevated ratio in Q1 2026 would force us to concede the structural read.

The fourth thing that could make us wrong is the simplest and most uncomfortable one. Yatsen is genuinely illiquid. Even if the variant view is right, generalist long-only and long/short funds cannot build or exit positions at meaningful size — the Tech tab calculates that a 0.5%-of-market-cap position takes 24 trading days to clear at 20% ADV. That means even a 100%-right variant view may take years to monetize, because the marginal buyer who would close the multiple gap is structurally absent from the order book. A PM who underwrites the variant view also has to underwrite that they will hold through illiquidity and possibly through one more cycle of forced flow if the next operating print is mixed.

The first thing to watch is the Q1 2026 S&M ratio printed at the May 14, 2026 release.