Financial Shenanigans

Financial Shenanigans — Yatsen Holding Limited (YSG)

Forensic verdict — Watch (score 38)

Yatsen's reported numbers are broadly traceable to economic reality, but four pressure points keep this off "Clean": (1) two consecutive Q4 goodwill baths on Eve Lom — RMB354.0M in Q4 2023 and RMB403.1M in Q4 2024 — that all but eliminate acquired skincare goodwill and mechanically deliver the FY2025 "non-GAAP profitability" narrative [1] [2]; (2) escalating related-party purchases from companies under Yatsen's "significant control" — RMB211.0M → RMB285.5M → RMB372.0M across 2023-2025 — plus a March 2026 US$120M convertible note placed with a vehicle affiliated with the founder/CEO [3] [4]; (3) explicit Q4 2025 "out of period adjustments" of RMB14.6M (RMB7.4M tied to prior years) flowing through current-quarter revenue and cost of revenues [5]; and (4) founder/CEO Jinfeng Huang controlling 90.4% of voting power through Class B super-voting shares, with no chair/CEO separation [6]. Cash from operations has been negative for three straight years (-RMB107M FY23, -RMB244M FY24, -RMB95M FY25) and the cash and short-term-investment war chest has shrunk from RMB2.59B (FY22) to RMB1.05B (FY25) [2] [1]. What would change the grade: a clean closing of the founder convertible note on third-party-comparable terms with majority-shareholder consent, FY2026 operating cash flow positive on lower related-party purchase volume, and FY2026 with no new "out of period" adjustments would push this to Clean (15-20); failure to close the convertible while burn continues, or a third consecutive Q4 impairment (DR.WU is the FY2024 critical-audit-matter unit, with RMB134M residual goodwill), pushes this to Elevated (45-55) [7].

Forensic Risk Score (0-100)

Red Flags

Yellow Flags

Clean Tests

CFO / Net Income (3y, -ve over -ve)

-0.29

FCF / Net Income (3y)

-0.30

FY24 Non-GAAP Gap (RMB M)

-582

Cash + STI Decline FY22→FY25 (RMB M)

-1,574

The headline pattern: the GAAP net loss collapsed from RMB710M (FY24) to RMB92M (FY25) — an 87% improvement — and management labelled FY25 a "non-GAAP profitability turnaround" [2]. But ~RMB403M of the swing is the simple absence of the prior-year goodwill impairment, ~RMB101M is non-GAAP carve-outs (SBC RMB59M, intangible amortisation RMB43M, investment impairment RMB13M, less tax effects), and CFO still ran -RMB95M [2] [8]. That is the lens to read this page through.

The 13-category shenanigans scorecard

This is the standardised scorecard required for every forensic page. Citations for each row's evidence appear in the prose section that handles that category.

No Results

The big-bath engine — two Q4 goodwill impairments

The single largest accounting story across the 2021-2025 corpus is the staged write-off of acquired-skincare goodwill, almost entirely on the Eve Lom reporting unit. Reported goodwill peaked at RMB857M (FY22), fell to RMB557M (FY23) after the first Eve Lom charge, and collapsed to RMB155M (FY24) after the second — a roughly RMB702M cumulative write-down [9] [10]. Yatsen's FY2022 20-F explicitly stated that "no impairment of goodwill was identified" in the FY22 annual test — and the FY23 charge of RMB354.0M followed only six months later, with the CFO attributing it on the Q4 2023 call to "weaker operating results than expected at the time of acquisition" [11]. The second-leg RMB403.1M write-down in Q4 2024 — same brand, same dynamic — was flagged on the earnings call alongside the "strategic transformation" narrative [12] [1]. The result is that FY25's headline operating-margin improvement is mechanically explained by the absence of the bath: management itself disclosed that operating loss margin fell from 24.3% to 4.3% "primarily because there was no impairment of goodwill for the full year of 2025" [2].

What this leaves on the balance sheet is RMB155M of residual goodwill — and the FY2024 auditor critical-audit-matter is now the DR.WU reporting unit (RMB134M of the residual), the next domino auditors elected to highlight for fair-value sensitivity [7]. For FY2025, the auditor swapped the goodwill CAM out and replaced it with Impairment of Inventories (RMB38M allowance on RMB509M of inventory), which itself implies the audit committee considers inventory marks the next significant judgment area [13]. A third Q4 impairment — on DR.WU, or via inventory — is the single most important read-through to monitor in the next two annual reports.

Earnings quality — non-GAAP narrative vs GAAP cash

Yatsen presents FY2025 as a "non-GAAP net income turnaround" of RMB8.4M, against a GAAP net loss of RMB92.4M [2]. The gap between the two is RMB100.7M — and it is built from add-backs whose "non-recurring" framing varies sharply by item. Share-based compensation (RMB59M in FY25) is a recurring labour cost, not a non-cash one-off; intangible amortisation from acquisitions (RMB43M) is by construction a multi-year cost; "impairment of investments" (RMB13M, a new FY25 add-back category) is small but it expanded the definition vs FY24 [2]. The FY24 non-GAAP gap was much larger: GAAP loss RMB710M, non-GAAP loss RMB128M, a gap of RMB582M dominated by the RMB403M goodwill add-back, RMB106M intangible amortisation, and RMB91M SBC [10].

The non-GAAP definition itself has drifted across years. The Q4 2024 release adds back share-based compensation, amortisation of acquired intangibles, revaluation of equity-method investments, impairment of goodwill, and tax effects [10]. The Q4 2025 release adds a sixth item — impairment of investments — to the net-income definition, broadening the carve-out list at the exact moment the headline pivots from loss to "non-GAAP income" [2]. Read literally, every time a new non-cash charge appears, management has reserved the right to exclude it. That is the KM1 pattern: the headline metric drifts to fit the result, not the other way around.

A second earnings-quality flag is the contribution of below-the-line items to the reported number. FY25 financial income (RMB40.7M), FX gain (RMB13.4M), equity-method income (RMB5.9M) and other income (RMB46.7M) together totalled RMB106.7M — larger than the entire pre-tax loss of RMB92.5M [2]. Without those non-operating tailwinds, the FY25 pre-tax loss would have been close to RMB200M (i.e., the disclosed operating loss of RMB186M), and the "narrowed loss" narrative weakens.

Cash-flow quality — three straight years of OCF burn

The cleanest forensic test on this company is the cash-flow statement, and it does not flatter the income statement. Operating cash flow has been negative for three consecutive years — even as gross margin expanded from 68.0% to 78.2% and revenue accelerated to +26.7% in FY25 [2]. FY22's positive RMB136M of CFO has not repeated.

Two CFO mechanisms matter here. First, FY25's working capital absorbed cash on net even with payable stretch. Accounts and notes payable rose from RMB72M to RMB149M (+RMB77M) — the line is now disclosed as "Accounts and notes payable" rather than just "Accounts payable", suggesting bank-acceptance financing has been introduced [9]. At the same time inventory ballooned from RMB386M to RMB509M (+RMB123M), an outflow that nearly cancelled the payable stretch. Without the AP/notes-payable expansion, FY25 CFO would have been closer to -RMB170M. That is a CF4 working-capital lifeline.

Second, the investing line has been a liquidity source, not a use. FY24 investing activities generated +RMB592M of cash, driven almost entirely by selling short-term investments out of a previously RMB1.2B portfolio [10]. FY25 still drew on STI, with the portfolio falling from RMB539M to RMB246M [9]. This is one-time monetisation of wealth-management product holdings, not recurring operating cash generation. The result is the cash war chest — cash, restricted cash and short-term investments combined — has fallen from RMB2.59B at FY22 year-end to RMB1.05B at FY25 year-end [1] [2].

That RMB1.54B decline coincides with treasury stock building from RMB865M (FY23 year-end) to RMB1,276M (FY24) and back to RMB1,251M (FY25) — i.e., FY24 buybacks alone consumed roughly RMB412M of cash, with FY25 net activity broadly flat after some share retirement / share-plan flows [10] [9]. The company is shrinking the cash base by financing buybacks out of investing proceeds while operations bleed — a defensible capital-allocation choice given the share price, but not a story management should be allowed to retell as "improved cash conversion".

A smoking-gun out-of-period adjustment in Q4 2025

The Q4 2025 income statement carries an explicit footnote disclosure that should be flagged: "In the fourth quarter of 2025, we made certain out of period adjustments mainly relating to revenues and cost of revenues to correct certain prior periods errors mainly occurred during the sales return and inventory receipt processes, which reduced quarterly profit by RMB14.6 million. Out of the RMB14.6 million adjustments, RMB7.4 million adjustments were related to prior years." [5]. Management's own conclusion is that the errors are "not material" individually or in aggregate.

This is a textbook EM6 pattern — prior-period errors swept into the current period rather than separately restated. The amounts are small relative to the RMB4.3B revenue base (0.34%), and the disclosure is transparent, so this does not justify a "red" verdict on EM5/EM6 standing alone. But it sits inside a company whose Q4 2025 net income of RMB3.0M is roughly five times smaller than the RMB14.6M adjustment — so the swing from a small reported profit to a small loss would have moved the headline number without the prior-period sweep. The disclosure is honest; the magnitude relative to a thin profit is what makes this material.

The lift-out matters because Q4 2025 is the quarter management used to declare "the first full-year non-GAAP profitability since 2020." If the RMB7.4M prior-year adjustment is added back to Q4 2025 (the right comparison for an investor underwriting underlying run-rate), the quarter's reported RMB3.0M net income flips to a small net loss, and the FY25 non-GAAP net income of RMB8.4M shrinks toward zero [2].

The breeding-ground / EM2 risk on Yatsen is concentrated in one disclosure: the company's purchases of "inventories and services from companies over which we exercise significant control." Those purchases have risen RMB211.0M → RMB285.5M → RMB372.0M over 2023-2025, a 76% increase in two years — well ahead of revenue growth over the same period [3]. Sales of inventory to a company controlled by the CEO were RMB20.1M in 2023 and RMB10.4M in 2024 before going to nil in 2025; the IPO baseline disclosure showed RMB67.9M founder advances and RMB349M shareholder amounts, evidencing this is a long-standing affiliated ecosystem rather than an isolated event [3].

The escalation matters because (i) it is large relative to non-cash cost of revenues (FY25 COGS RMB937M; affiliated purchases of RMB372M imply roughly 40% of cost-of-revenue plus services are sourced from a related party), and (ii) the disclosure says the amount due to that affiliated company at year-end was RMB21.3M, "unsecured and interest free" — i.e., the affiliated party is extending working-capital credit to Yatsen [3]. That is exactly the kind of two-way flow Financial Shenanigans literature flags as worth arm's-length testing. The audit committee's stated duty is to review and approve all related-party transactions; the size of these flows means the audit committee's process is now a load-bearing control, not a procedural one.

Governance breeding ground — founder dominance + post-IPO convertible

Three structural conditions amplify the accounting flags above. First, founder-CEO Jinfeng Huang holds all the Class B super-voting shares, equal to 32.0% of issued shares but 90.4% of aggregate voting power [6]. The roles of Chairman and CEO are not separated [6]. The board has five directors, three independent, two executive — the minimum permissible NYSE configuration, with the founder, CFO, and three independents [6].

Second, as a foreign private issuer, Yatsen discloses executive compensation only in aggregate — RMB9.5M cash and RMB0.5M benefits for the executive group in FY2024 — not individually. Investors cannot test whether CEO or CFO incentive payouts moved with the goodwill bath or the non-GAAP swing because the compensation framework is not visible at named-executive level. This is a transparency limitation rather than a misconduct flag, but it removes one important triangulation channel.

Third, the proposed March 2026 US$120M convertible-note financing was placed with a vehicle affiliated with Trustar Capital and the founder/CEO — a related-party financing rather than a third-party round [4]. The annual report discloses that "a significant shareholder of ours raised objections to, and concerns regarding, the consummation of the transaction" and that closing of the first tranche has not occurred as of the FY25 filing date [4]. The notes carry an aggregate principal of approximately RMB830M with warrants attached — and the precise economics are documented in the Note Purchase Agreement filed as Exhibit 4.10 to the FY25 20-F [3]. An investor underwriting position size should track this transaction's closing terms closely: any change to the warrant strike, conversion price, or note coupon below an independent third-party benchmark would meaningfully shift the related-party-fairness assessment.

Where the clean tests come back clean

It is fair to call out what is working. Revenue-recognition cut-off is well-described — the FY25 critical-audit-matter is inventory, not revenue, and DSO is lower in FY25 (18.8 days) than at any year-end in the sample, which is the opposite of the EM1 pattern [2]. Capex is modest (RMB44-57M annually) and well below depreciation and amortisation, ruling out the CF2 "operating costs disguised as investing" pattern. Borrowings have been negligible, so CF1 (financing inflows mislabelled as operating) does not apply. The auditor — PricewaterhouseCoopers Zhong Tian LLP, engaged since 2019 — has issued unqualified opinions throughout and continues to flag a single CAM per year [13]. Internal control over financial reporting was concluded effective for FY25 by the auditor [14]. And the company has been entirely transparent about its non-GAAP definitions — both the Q4 2024 and Q4 2025 releases disclose the full add-back list and reconciliation [10] [2].

The honest read is: Yatsen is a transparent, audited, US-listed company whose accounting choices are documented and visible, but whose reported earnings improvement story leans heavily on the absence of repeat goodwill baths, on broadening non-GAAP definitions, and on related-party purchase volumes that have outgrown revenue. The substance is checkable; the incentives are pointed in a worrying direction.

What to underwrite next

The four diligence items worth bandwidth between now and the FY2026 20-F:

DR.WU goodwill recoverability. FY24 critical-audit-matter unit; residual goodwill RMB134M of total RMB155M. Read the FY26 critical-audit-matter list first when the annual report is filed [7]. A third Q4 impairment downgrades the grade to Elevated.
Closing of the founder convertible note. US$120M principal in two RMB-denominated tranches with warrants; a significant shareholder has objected to the structure [4]. Track whether the closing terms are amended (e.g., a co-investor brought in, or strike adjusted), whether the founder steps back from the purchaser vehicle, and whether the proceeds are correctly classified in financing activities — not operating — in FY26 cash-flow statements.
Related-party purchase volume. RMB372M in FY25, +RMB87M YoY against RMB905M of revenue growth [3] [2]. FY26 disclosure that materially exceeds the affiliate's normal supplier scale, or that grows faster than COGS, is an upgrade to red on EM2.
Out-of-period adjustments frequency. FY25 logged RMB14.6M in Q4 [5]. A repeat in FY26 — even of similar size — would move EM6 from yellow to red, because pattern is what distinguishes immaterial true-up from systematic smoothing.

How this should affect valuation and sizing: this forensic profile is a position-sizing limiter and a valuation-haircut item, not a thesis breaker. A clean balance sheet (no debt, RMB1.05B of cash + STI), a transparent auditor, and rapidly growing revenue with expanding gross margin keep the business out of the "Elevated" or "High" bucket. But every reported earnings figure should be deflated to acknowledge (i) recurring SBC and acquired-intangible amortisation belong in operating costs, and (ii) the FY25 swing to "non-GAAP profit" was largely the absence of a one-off charge plus broadened add-backs. Treat the underwriting as: this is a business burning roughly RMB100M of operating cash per year on a shrinking war chest, with a founder convertible note pending and a related-party supplier ecosystem worth RMB372M annually that requires standing audit-committee oversight. Margin of safety on multiples should reflect that reality, not the non-GAAP turnaround narrative.

References

Generated Jun 19, 2026 10:16 UTC