3350 — Hankook Cosmetics Manufacturing Co., Ltd.

Deep-value analysis (REFRESH) — 003350 (한국화장품제조 / Hankook Cosmetics Manufacturing Co., Ltd.)

Analysis date: September 16, 2026 · Exchange: KRX / KOSPI (code 003350) · Yahoo symbol: 003350.KS Reference price: … (close 09.16.2026) · Reporting currency = price currency = KRW Reference analysis: 08.24.2026, at …, verdict SMALL BUY · Price move since then: … Regime: REFRESH — mechanical deep_delta.py trigger: falsifier pret > 12824 BROKE.

Why this isn’t a cosmetic refresh. The broken falsifier said: “20.01% of share capital is a latent overhang via the exchangeable bond, ignored above a price of ….” Between September 11 and 14, 2026 the overhang stopped being latent — holders began requesting the exchange, and the DART filing of 09.14.2026 documents it at the won level. In addition, the semi-annual report filed on 08.13.2026 (DART, 반기보고서 2026.06) became accessible and contains exactly the four pieces of information the August analysis had flagged as unavailable: the inventory breakdown by category, capacity utilization, related-party concentration, and the exact bond terms. All four were obtained this time from the primary filing, not from aggregators.

Executive summary (1 page: thesis, estimated value, verdict)

The underlying thesis hasn’t changed and I inherit it in full: Hankook Cosmetics Manufacturing is a South Korean ODM/OEM cosmetics manufacturer, founded in 1962, the group’s production entity (the distribution entity, 한국화장품, is listed separately under 123690 and is itself 20%-owned by 003350). No own brand, no guidance, a single factory, an extremely lean cost structure, and operating margins of 17… at a K-beauty cycle peak. Operating figures are unchanged from August, because no new quarter has appeared: TTM (Q3 2025 – Q2 2026) revenue …, EBIT 38,586 mil (margin 17.84%), net profit 32,026 mil, operating cash flow … mil. The next report (Q3 2026) comes in November.

What changed are three things, in order of importance.

1. The price rose 27.6% without a single new operating figure. At … the median of the five-model triangulation was …. At …, the same models, updated, give a median of …. The margin of safety wasn’t consumed by a deterioration in the business; it was consumed by the quote. The key day was 09.11.2026: … (13,160 → …) on a volume of 11,798,448 shares — 1.7 times the entire free float of ~6.99 million, with two volatility halts, on a day when KOSPI was falling 2.4%. There is no company announcement that day explaining the move.

2. The 20.01% overhang started materializing, and I can document it. DART filing 20260914800640 shows: on 09.11 an exchange was requested for … face value (77,978 shares), on 09.14 another 2.0 bn (155,957 shares), cumulatively 233,935 shares = 1.03% of share capital. Remaining balance unchanged: …, i.e. 4,301,030 shares at the exchange price of … — equivalent to 61.5% of the current free float, in the hands of 18 financial institutions with a cost basis of 12,824 and the first put option on 10.14.2027. Every day spent above 12,824 is a day those shares have an economic reason to leave.

3. The semi-annual filing puts a physical ceiling on growth. The company discloses, for H1 2026, a production capacity of …, actual production of 89,149 mil, and an average utilization rate of 87.2% (838 of 960 hours). Even more important: the average selling price per unit is … in H1 2026, versus 1,823 in 2025 and 1,939 in 2024 — so the ASP is below the level of two years ago, while ethanol has gotten 14.0% more expensive and glycerin 39.7%. All of the 34.5% growth in H1 2026 is volume, not a single won comes from price. A factory at 87% utilization, with a real-terms declining selling price and rising input cost, doesn’t compound 12% a year for ten years running — and the new factory, whose land was purchased in April 2026, still had, as of 08.13.2026, neither content, nor budget, nor an approved investment timeline.

Estimated value. Five models, all on 22.66 million shares (fully diluted):

Median of the five: …. Monte Carlo across 20,000 scenarios gives median …, P10 …, P90 …, and a probability of undervaluation of …% — practically a coin flip, versus …% in August.

Verdict: HOLD / STOP BUYING. No adding above …. This isn’t a sell: the business is real, earnings quality remains good, the balance sheet is solid, and the base case doesn’t say the stock is expensive — it says it’s fairly valued. But all three anchors that don’t depend on extrapolating the peak margin (EPV …, historical multiples …, DCF bear …) are now below the price, and in August only one was. When the only model that still gives a decent return is the one that assumes the peak margin holds for ten years, you no longer have a margin of safety — you have an opinion. The disciplined re-buy threshold remains below … (below the bond’s exchange price, where the overhang extinguishes automatically), and the position-trim threshold is above …, where even the base model turns negative.