Two markets,two experiments

Problem

Japanese consumers read every ingredient list. It was unknown how long the efficacy-led angle that worked in Korea would take to earn trust, or which evidence would be accepted first.

Action

320 videos · 14 days · TikTok + Reels. Deployed in four hook families (ingredient-list read · comparison · after-use · expert quote). Reaction ranking and dynamics judgment sealed before deployment.

Observation

In 11 videos using the “30 seconds reading the ingredient list” format, saves rose before shares, and higher. Question comments concentrated on ingredient source and concentration. Efficacy angles drew views but saves did not follow.

Result

Judged Trust · high confidence. Forecast agreement 84% (baseline 91%). The answer to “what do they trust enough to buy?” was ingredient evidence, not efficacy.

Problem

No reviews, so ads don’t convert; no conversions, so reviews don’t accumulate. There was not even a hypothesis about where the first buyers would come from.

Action

318 videos · 21 days · TikTok. Three hook families (“zero reviews” confession · derm vs. drugstore comparison · ingredient demo). The sealed forecast ranked comparison first.

Observation

On the 7 videos that used “zero reviews” as the hook, question comments arrived before likes — comparison requests like “how is this different from what my derm recommends?” The comparison family under-performed its forecast; the confession family opened exploration.

Result

Judged Exploration · medium confidence. Forecast agreement 76% — the sealed forecast was partly wrong, and that error became the round-two hypothesis.