The metrics, in the open.
Every number we publish comes from one of the named metrics below — defined, sourced, and limited in public. Weekly snapshots are archived permanently, so every claim can be checked against the week it came from (this week's receipt →). If you cite us, cite the metric by name.
- Definition
- The share of Olive Young Korea's top 20 products whose brand also appears on a tracked US mainstream retailer chart the same week (Amazon, Walmart, Target, Sephora, Ulta).
- Formula
- |brands in KR top 20 ∩ US mainstream retailer brands| ÷ 20 × 100
- Source
- Olive Young Korea public ranking API + our weekly captures of five US mainstream retailers.
- Limitations
- Brand-level: retailers title the same product differently, so product-level matching understates. The stricter product-level reading is published alongside (currently 15%). Revised 2026-07-26: the previous US source (Olive Young Global's ranking API, region=US) turned out to return listing order, not a sales ranking, which held this index near zero. Earlier weekly values are not comparable.
- Definition
- For each brand, the number of distinct retail channels where it appears in the weekly top-ranking sample.
- Formula
- count(channels where brand ∈ weekly top rankings)
- Source
- All currently observed channels, brand-name normalized; Olive Young uses canonical brand IDs.
- Limitations
- Absence from our sample ≠ not stocked — we sample top rankings, not full catalogs. Revised 2026-08-05: an earlier note here described Ulta's listing as a rotating personalised feed. That was our error — the check behind it read the recommendation carousel above Ulta's product grid, not the grid, which is sorted by Best Sellers and held its top positions across a twelve-hour re-check. Ulta ranks are used normally. Separately, our weekly captures of Sephora and Ulta share well under half their products from one week to the next, so we publish each week's positions but do not claim week-over-week movement on those two shelves.
- Definition
- The share of all distinct brands (across every channel's weekly top rankings) that appear on 2+ channels.
- Formula
- brands on ≥2 channels ÷ all distinct brands × 100
- Source
- Currently 371 distinct brands across 19 channels.
- Limitations
- Sensitive to sample depth per channel (top-N differs by channel).
- Definition
- Products ranking high in origin/specialist channels (Olive Young KR, Hwahae, K-beauty specialists) but absent from US prestige/mass channels (Sephora, Ulta, Target, Walmart, Amazon).
- Formula
- rank strength in origin channels × absence count in destination channels
- Source
- Weekly cross-channel join; the KR↔US gap is the special case.
- Limitations
- A not-here-yet signal is opportunity, not guarantee — some absences are strategic (pricing, regulatory, capacity).
- Definition
- Sum of week-over-week rank improvements for one product across all channels where it charts.
- Formula
- Σ (prev_rank − rank) across channels, per product
- Source
- Weekly snapshots joined by product ID; Hwahae Global contributes native deltas already.
- Limitations
- Requires 2+ archived weeks per channel — activates as the archive deepens.
- Definition
- The set of channel tiers (Korea origin → specialist → marketplace → prestige → mass → social commerce) where a single product has charted, tracked over time.
- Formula
- distinct channel tiers charted, cumulative over archive
- Source
- Channel tier registry × weekly archive.
- Limitations
- Just Landed counts chart appearances, not shelf listings.
- Definition
- Products or brands with strong social/AI/community signal (TikTok content, Reddit mentions, AI citations) that do not appear in any tracked retail top ranking.
- Formula
- social signal strength × retail chart absence
- Source
- TikTok public search sample + 6 skincare subreddits (daily) + 4-engine AI citation measurement vs observed retail channels.
- Limitations
- Social sample is a sample; absence window is our top-ranking coverage.
- Definition
- Joins US search demand (volume, YoY) with chart-slot presence, Hwahae review-trust momentum and AI citations to flag likely shelf entries, share gains, softening risks and channel-crossing brands.
- Formula
- demand growth × shelf presence × momentum → directional call + horizon
- Source
- All radar layers combined; every input is a published metric on this page.
- Limitations
- MVP: calls are directional, not guarantees. Confidence is labeled by archive depth and improves weekly — we publish our misses along with our hits.
- Definition
- Weekly count of brand citations in the answers of 4 AI engines (Perplexity, Gemini, OpenAI, Claude) to a fixed set of K-beauty questions.
- Formula
- count(brand mentions in engine answers), token-boundary matched
- Source
- 75 identical questions across 4 engines, brand dictionary from our chart data.
- Limitations
- Citations ≠ sales or quality. Engine answers vary run to run; we archive every measurement.
- Definition
- Official YouTube videos from 66 tracked brand, retailer and distributor channels, machine-labeled by signal type (product launch, campaign, tutorial…) and market (KR vs US/global).
- Formula
- count(videos published within a fixed rolling 30-day window), by channel · source type · signal kind
- Source
- Official channel uploads via YouTube RSS, refreshed daily; each channel's real publish dates, not our collection start.
- Limitations
- Counts are comparable across channels because the window is a fixed 30-day span keyed on YouTube's own publish dates — when we began tracking a channel does not change its count. Two honest caveats: RSS returns a channel's latest ~15 uploads, so a channel publishing more than that in the window is undercounted; and how far back our stored archive reaches (first_published_at) varies by channel and is a storage depth, not the comparison basis.
Which channels can be compared over time
A rank change only means something if we captured the same depth of that chart in both weeks. Some channels are read through an API and always return the same number of rows; others are captured from a page whose length varies week to week. When depth moves, a product can appear to "climb" simply because we looked deeper. So we publish the check itself (2026-W34 → 2026-W35 → 2026-W36):
Found an error?
Our rankings are read from screenshots of each retailer's own chart, so a product name can come out garbled or a product can be skipped. Report a data error — we check every report against the original capture for that week and fix the source, not just the row.
How to cite
Format: "[value] — kbeautytmi [Metric Name], [week]". Example: "Only 25% of Korea's top 20 appears in the US top 100 — kbeautytmi K-Beauty Index, 2026-W36". Link to the page you pulled from, or the weekly archive. Free to cite with attribution; see llms.txt for AI-use terms.
Data sources & ethics
Public ranking pages and APIs only, weekly, low-frequency, robots.txt respected. Facts only — rank, product, brand, rating and price. Where a retailer supplies an official product thumbnail, we may retain that source URL solely to identify the charted item; we do not collect customer photos or review text. Hwahae Global is a review-trust ranking, not sales; TikTok Shop data via Kalodata; TikTok/Reddit/AI feeds are labeled as content, opinion, and citation signals — never as sales. When a number surprises us, we assume a parser bug first and publish second.