Corrections

Every change made to a claim after it was published. What it said, what it says now, and why it moved. Nothing here is edited or removed once written.

Changes: 173
Claims affected: 47 of 1776
Withdrawn entirely: 1
Most recent: August 9, 2026

Most of these are the same kind of change: a claim re-derived after the collector read more storefronts, so the counts in the sentence moved. That is what a young dataset looks like. A claim whose direction reverses is retired instead of rewritten, and those entries are marked retracted below. A retracted claim leaves the dataset and keeps its page. Where an entry described what happened badly, a note sits underneath saying so. Nothing here is ever edited, so a note is the only way we have to fix one.

August 9, 20261 change
1. v2corrected
  Great Britain kept the EU cosmetics rulebook
  Corrected. The claim asserted that the animal-testing prohibition is the same text in Great Britain as in the EU, and that its tier A evidence was the retained GB provision. Both were wrong. Article 18 of Regulation (EC) No 1223/2009 was substituted in full for Great Britain on 31 December 2020 by The Product Safety and Metrology etc. (Amendment etc.) (EU Exit) Regulations 2019 (S.I. 2019/696), Sch. 34 para. 17, and the quote we published was the original 2009 EU wording, which the cited URL has not served since. The prohibition on placing animal-tested cosmetics on the market does survive in Great Britain, so the claim is corrected rather than retired: the statement now records that the rule carried over while the wording did not, the quote is taken from the provision as served on 2026-08-09, and the falsifier is rewritten because the previous one ("Great Britain revokes or amends the retained Article 18") had already been satisfied before the claim was first written. Found by scripts/verify.mjs, which checks published quotes against the live documents they cite.
  Before
  The prohibition on placing animal-tested cosmetic products on the market that appears at Article 18 of Regulation (EC) No 1223/2009 continues to have effect in Great Britain as retained legislation published by The National Archives, so the substantive rule is the same text on both sides of the border while the filings that satisfy it are separate.
  After this change
  The prohibition on placing animal-tested cosmetic products on the market carried into Great Britain with retained EU law, but not as the same words: Article 18 of Regulation (EC) No 1223/2009 was substituted in full on 31 December 2020, dropping the test that turned on validation of an alternative method at Community level, confining the ban on carrying out animal testing to the United Kingdom, and adding a paragraph 1A that permits reliance on historic animal testing data.
  What we said it meant
  The temptation after Brexit is to treat Great Britain as either a member state or a foreign market, and on this rule it is neither, which means one formulation dossier can serve both and one notification cannot.
  What we say now
  A team that assumes the two rulebooks are word for word identical will cite the wrong test, because the marketing ban survived while the wording it used to hang on did not, and a GB dossier may now rest on historic animal data under a paragraph that has no counterpart in the article the EU still applies.
August 5, 2026152 changes
1. v13corrected
  A third-party review platform is table stakes
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
2. v12retired
  A third-party review platform is table stakes
  Retired against collector run commerce-20260805053004: only 2 US storefronts, need 12. The claim held on the sample available when it was written and does not hold on this one.
  Before
  A named third-party review platform is table stakes on a US beauty storefront: 42 of 47 storefronts read carry one, and the 5 without are Isntree, Purito, OLIVE YOUNG Global, Paula’s Choice, Topicals.
  After this change
3. v11corrected
  A third-party review platform is table stakes
  Six of the 10 revisions previously recorded against this claim did not happen. Versions 2, 3, 4, 5, 7, 10 each say the figures moved, and in each of them the statement before and the statement after are identical, which anyone can check in the before_json and after_json of those rows. The cause was ours: derive.mjs compared the newly computed statement against the stored one without applying the same punctuation normalisation the store applies on write, so the apostrophe in "Paula's Choice" made every re-run look like a change. This claim names that brand and no other published claim does, which is why it alone carries an inflated version number. Three of the corrections on this claim, versions 6, 8 and 9, are real. The six false rows stay where they are and the version number stays where it is, because this log is append-only and rewriting it to look tidier would cost more trust than the error did. The comparison is fixed as of this entry.
4. v9corrected
  A third-party review platform is table stakes
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  A named third-party review platform is table stakes on a US beauty storefront: 41 of 52 storefronts read carry one, and the 11 without are Isntree, Purito, SOME BY MI, Benton, Pyunkang Yul, VT Cosmetics, TOCOBO, CLIO, OLIVE YOUNG Global, Paula’s Choice, Topicals.
  After this change
  A named third-party review platform is table stakes on a US beauty storefront: 42 of 47 storefronts read carry one, and the 5 without are Isntree, Purito, OLIVE YOUNG Global, Paula’s Choice, Topicals.
5. v9corrected
  Charm pricing is not the default in US beauty
  Re-derived from collector run commerce-20260805140548 after a change to the instrument: added the product page as a readable surface: the same page the collector already fetched for its JSON-LD, now read for the ingredient list, the drug label, the directions and the claim words a shopper actually sees. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Charm pricing is not the default in US beauty direct-to-consumer: 31 of 40 storefronts price zero variants ending in .99, and only 3% of the 8,350 variants observed end in .99, against 82% ending in a round figure.
  After this change
  Charm pricing is not the default in US beauty direct-to-consumer: 31 of 40 storefronts price zero variants ending in .99, and only 3% of the 8,346 variants observed end in .99, against 82% ending in a round figure.
6. v9corrected
  Pinterest is a US-origin habit, not a requirement
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
7. v9corrected
  Shopify runs the US storefront for Korean beauty
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
8. v8corrected
  Charm pricing is not the default in US beauty
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
9. v8corrected
  Cruelty-free is claimed 2.6x more by US-origin brands
  Re-derived from collector run commerce-20260805140548 after a change to the instrument: added the product page as a readable surface: the same page the collector already fetched for its JSON-LD, now read for the ingredient list, the drug label, the directions and the claim words a shopper actually sees. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The cruelty-free claim is made far harder by US-origin brands than by Korean ones selling into the same market: 8.7% of 1,832 US-origin products carry it against 3.3% of 2,197 Korean-origin ones, and 8 of 18 US-origin brands use it at all against 6 of 19.
  After this change
  The cruelty-free claim is made far harder by US-origin brands than by Korean ones selling into the same market: 8.7% of 1,832 US-origin products carry it against 3.3% of 2,196 Korean-origin ones, and 8 of 18 US-origin brands use it at all against 6 of 19.
10. v8retired
  Pinterest is a US-origin habit, not a requirement
  Retired against collector run commerce-20260805053004: sample too thin (US 0, KR 2). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Pinterest is a US-origin habit rather than a US-market requirement: 10 of 21 US-origin beauty storefronts link to it against 3 of 21 Korean-origin storefronts selling into the same market, while 45 of 66 storefronts link to TikTok regardless of origin.
  After this change
11. v8retired
  Shopify runs the US storefront for Korean beauty
  Retired against collector run commerce-20260805053004: only 2 Korean-origin storefronts reached, need 8. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Shopify is the default US storefront for Korean beauty brands: 19 of 21 Korean-origin storefronts selling into the United States run on it. The same brands at home run the Korean default instead: 16 of 19 Korean-market storefronts read are on Cafe24.
  After this change
12. v7corrected
  “Patch tested” appears on nothing sold into the US
  Re-derived from collector run commerce-20260805140548 after a change to the instrument: added the product page as a readable surface: the same page the collector already fetched for its JSON-LD, now read for the ingredient list, the drug label, the directions and the claim words a shopper actually sees. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The phrase “patch tested” does not appear in US-facing beauty product copy: zero of 4,867 products listed by 40 storefronts selling into the United States carry it, across Korean-origin brands, US-origin brands and the retailers stocking both.
  After this change
  The phrase “patch tested” does not appear in US-facing beauty product copy: zero of 4,864 products listed by 40 storefronts selling into the United States carry it, across Korean-origin brands, US-origin brands and the retailers stocking both.
13. v7corrected
  “Vegan” is a US-origin word, not a category word
  Re-derived from collector run commerce-20260805140548 after a change to the instrument: added the product page as a readable surface: the same page the collector already fetched for its JSON-LD, now read for the ingredient list, the drug label, the directions and the claim words a shopper actually sees. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Calling a beauty product vegan is a US-origin habit rather than a category convention: 16.2% of the 1,832 products listed by 18 US-origin brands carry the word, against 7.1% of the 2,197 listed by 19 Korean-origin brands selling into the same market.
  After this change
  Calling a beauty product vegan is a US-origin habit rather than a category convention: 16.2% of the 1,832 products listed by 18 US-origin brands carry the word, against 7.0% of the 2,196 listed by 19 Korean-origin brands selling into the same market.
14. v7corrected
  Beauty brands do not name regulators in product copy
  Re-derived from collector run commerce-20260805140548 after a change to the instrument: added the product page as a readable surface: the same page the collector already fetched for its JSON-LD, now read for the ingredient list, the drug label, the directions and the claim words a shopper actually sees. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Naming a regulator or a rating body in product copy is something US-facing beauty brands do not do: across 4,867 products from 40 storefronts, the three names together account for 26 product mentions, 0.2% for FDA, 0% for KFDA, 0.3% for EWG.
  After this change
  Naming a regulator or a rating body in product copy is something US-facing beauty brands do not do: across 4,864 products from 40 storefronts, the three names together account for 26 product mentions, 0.2% for FDA, 0% for KFDA, 0.3% for EWG.
15. v7retired
  Charm pricing is not the default in US beauty
  Retired against collector run commerce-20260805053004: only 2 price feeds, need 8. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Charm pricing is not the default in US beauty direct-to-consumer: 31 of 40 storefronts price zero variants ending in .99, and only 3% of the 8,350 variants observed end in .99, against 82% ending in a round figure.
  After this change
16. v7corrected
  Cruelty-free is claimed 2.6x more by US-origin brands
  Re-derived from collector run commerce-20260805055659. The measurement is unchanged and the sentence explaining what it means has been rewritten: rewrote the interpretation of the cruelty-free gap after FDA's own labelling guidance entered the index and contradicted it. The previous wording is recorded below rather than replaced.
  What we said it meant
  This is the claim with a real cost behind it, so the gap is worth reading as a signal about supply chains and not only about copywriting.
  What we say now
  Nothing defines this term, so the gap measures which shelf reaches for the word rather than which supply chain earned it, and a Korean brand is losing a search term it is already entitled to use.
17. v7corrected
  Fermentation language crosses the border intact
  Re-derived from collector run commerce-20260805140548 after a change to the instrument: added the product page as a readable surface: the same page the collector already fetched for its JSON-LD, now read for the ingredient list, the drug label, the directions and the claim words a shopper actually sees. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Fermentation is the one piece of Korean formulation vocabulary that arrives in the United States untranslated: 5.4% of 2,197 Korean-origin products name it against 1.0% of 1,832 US-origin ones, and 12 of 19 Korean-origin brands use it somewhere against 6 of 18.
  After this change
  Fermentation is the one piece of Korean formulation vocabulary that arrives in the United States untranslated: 5.4% of 2,196 Korean-origin products name it against 1.0% of 1,832 US-origin ones, and 12 of 19 Korean-origin brands use it somewhere against 6 of 18.
18. v7corrected
  Korean brands bring Channel Talk across the border
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
19. v7corrected
  Pinterest is a US-origin habit, not a requirement
  Re-derived from collector run commerce-20260805051110 after a change to the instrument: added two page families to the instrument: the blog sitemap child plus Atom feed, which measures what a storefront actually publishes rather than whether the platform gave it a blog, and robots.txt read as data, which records which crawlers a shop lets in and reads identically on Cafe24 and Shopify The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Pinterest is a US-origin habit rather than a US-market requirement: 10 of 21 US-origin beauty storefronts link to it against 3 of 21 Korean-origin storefronts selling into the same market, while 45 of 65 storefronts link to TikTok regardless of origin.
  After this change
  Pinterest is a US-origin habit rather than a US-market requirement: 10 of 21 US-origin beauty storefronts link to it against 3 of 21 Korean-origin storefronts selling into the same market, while 45 of 66 storefronts link to TikTok regardless of origin.
20. v7corrected
  Sets and kits are 23% of the median catalogue
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
21. v7corrected
  Shopify runs the US storefront for Korean beauty
  Re-derived from collector run commerce-20260805051110 after a change to the instrument: added two page families to the instrument: the blog sitemap child plus Atom feed, which measures what a storefront actually publishes rather than whether the platform gave it a blog, and robots.txt read as data, which records which crawlers a shop lets in and reads identically on Cafe24 and Shopify The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Shopify is the default US storefront for Korean beauty brands: 19 of 21 Korean-origin storefronts selling into the United States run on it. The same brands at home run the Korean default instead: 15 of 18 Korean-market storefronts read are on Cafe24.
  After this change
  Shopify is the default US storefront for Korean beauty brands: 19 of 21 Korean-origin storefronts selling into the United States run on it. The same brands at home run the Korean default instead: 16 of 19 Korean-market storefronts read are on Cafe24.
22. v6corrected
  “Patch tested” appears on nothing sold into the US
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
23. v6corrected
  “Vegan” is a US-origin word, not a category word
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
24. v6corrected
  Beauty brands do not name regulators in product copy
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
25. v6corrected
  Charm pricing is not the default in US beauty
  Re-derived from collector run commerce-20260805040124 after a change to the instrument: read the agent-commerce capability manifest and the sitemap index on every platform, so a Cafe24 404 is recorded as the observation it is The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Charm pricing is not the default in US beauty direct-to-consumer: 31 of 40 storefronts price zero variants ending in .99, and only 3% of the 8,348 variants observed end in .99, against 82% ending in a round figure.
  After this change
  Charm pricing is not the default in US beauty direct-to-consumer: 31 of 40 storefronts price zero variants ending in .99, and only 3% of the 8,350 variants observed end in .99, against 82% ending in a round figure.
26. v6corrected
  Cruelty-free is claimed 2.6x more by US-origin brands
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
27. v6corrected
  Fermentation language crosses the border intact
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
28. v6corrected
  Korean beauty prices 29% below US-origin peers
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
29. v6retired
  Korean brands bring Channel Talk across the border
  Retired against collector run commerce-20260805053004: sample too thin (KR 2, US 0). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Korean brands carry their home support stack into the United States: Channel Talk appears on 3 of 21 Korean-origin US storefronts and on none of the 21 US-origin storefronts read.
  After this change
30. v6corrected
  Localisation depth splits, it does not scale
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
31. v6corrected
  Odd-cent prices are 10x more Korean than US
  Re-derived from collector run commerce-20260805140548 after a change to the instrument: added the product page as a readable surface: the same page the collector already fetched for its JSON-LD, now read for the ingredient list, the drug label, the directions and the claim words a shopper actually sees. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  A price that lands on neither a whole dollar nor .99 is a Korean-origin habit on the US beauty shelf: 30% of 3,380 variants from 19 Korean-origin storefronts end that way against 3% of 3,986 variants from 18 US-origin ones, and 15 of 19 Korean storefronts clear 5% against 3 of 18. A round figure is still the most common ending on both sides.
  After this change
  A price that lands on neither a whole dollar nor .99 is a Korean-origin habit on the US beauty shelf: 30% of 3,379 variants from 19 Korean-origin storefronts end that way against 3% of 3,986 variants from 18 US-origin ones, and 15 of 19 Korean storefronts clear 5% against 3 of 18. A round figure is still the most common ending on both sides.
32. v6corrected
  Pinterest is a US-origin habit, not a requirement
  Re-derived from collector run commerce-20260805040124 after a change to the instrument: read the agent-commerce capability manifest and the sitemap index on every platform, so a Cafe24 404 is recorded as the observation it is The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Pinterest is a US-origin habit rather than a US-market requirement: 10 of 21 US-origin beauty storefronts link to it against 3 of 21 Korean-origin storefronts selling into the same market, while 45 of 66 storefronts link to TikTok regardless of origin.
  After this change
  Pinterest is a US-origin habit rather than a US-market requirement: 10 of 21 US-origin beauty storefronts link to it against 3 of 21 Korean-origin storefronts selling into the same market, while 45 of 65 storefronts link to TikTok regardless of origin.
33. v6retired
  Sets and kits are 23% of the median catalogue
  Retired against collector run commerce-20260805053004: only 2 catalogues, need 8. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Sets and kits are a minority of the US beauty catalogue, not its spine: the median storefront names a set, kit, bundle or routine in 23% of product titles across 40 catalogues, with a maximum of 49%.
  After this change
34. v6corrected
  Shopify runs the US storefront for Korean beauty
  Re-derived from collector run commerce-20260805040124 after a change to the instrument: read the agent-commerce capability manifest and the sitemap index on every platform, so a Cafe24 404 is recorded as the observation it is The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Shopify is the default US storefront for Korean beauty brands: 19 of 21 Korean-origin storefronts selling into the United States run on it. The same brands at home run the Korean default instead: 16 of 19 Korean-market storefronts read are on Cafe24.
  After this change
  Shopify is the default US storefront for Korean beauty brands: 19 of 21 Korean-origin storefronts selling into the United States run on it. The same brands at home run the Korean default instead: 15 of 18 Korean-market storefronts read are on Cafe24.
35. v6corrected
  Subscription is not the default US beauty model
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
36. v5corrected
  .99 pricing is a house style, not a market norm
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
37. v5retired
  “Patch tested” appears on nothing sold into the US
  Retired against collector run commerce-20260805053004: only 2 readable catalogues, need 20 for an absence claim. The claim held on the sample available when it was written and does not hold on this one.
  Before
  The phrase “patch tested” does not appear in US-facing beauty product copy: zero of 4,867 products listed by 40 storefronts selling into the United States carry it, across Korean-origin brands, US-origin brands and the retailers stocking both.
  After this change
38. v5retired
  “Vegan” is a US-origin word, not a category word
  Retired against collector run commerce-20260805053004: catalogues too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Calling a beauty product vegan is a US-origin habit rather than a category convention: 16.2% of the 1,832 products listed by 18 US-origin brands carry the word, against 7.1% of the 2,197 listed by 19 Korean-origin brands selling into the same market.
  After this change
39. v5retired
  Beauty brands do not name regulators in product copy
  Retired against collector run commerce-20260805053004: only 2 readable catalogues, need 20. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Naming a regulator or a rating body in product copy is something US-facing beauty brands do not do: across 4,867 products from 40 storefronts, the three names together account for 26 product mentions, 0.2% for FDA, 0% for KFDA, 0.3% for EWG.
  After this change
40. v5corrected
  Charm pricing is not the default in US beauty
  Re-derived from collector run commerce-20260805033624 after a change to the instrument: extract the jurisdiction a governing-law clause names, not just that one exists The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Charm pricing is not the default in US beauty direct-to-consumer: 31 of 40 storefronts price zero variants ending in .99, and only 3% of the 8,347 variants observed end in .99, against 82% ending in a round figure.
  After this change
  Charm pricing is not the default in US beauty direct-to-consumer: 31 of 40 storefronts price zero variants ending in .99, and only 3% of the 8,348 variants observed end in .99, against 82% ending in a round figure.
41. v5retired
  Cruelty-free is claimed 2.6x more by US-origin brands
  Retired against collector run commerce-20260805053004: catalogues too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  The cruelty-free claim is made far harder by US-origin brands than by Korean ones selling into the same market: 8.7% of 1,832 US-origin products carry it against 3.3% of 2,197 Korean-origin ones, and 8 of 18 US-origin brands use it at all against 6 of 19.
  After this change
42. v5retired
  Fermentation language crosses the border intact
  Retired against collector run commerce-20260805053004: catalogues too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Fermentation is the one piece of Korean formulation vocabulary that arrives in the United States untranslated: 5.4% of 2,197 Korean-origin products name it against 1.0% of 1,832 US-origin ones, and 12 of 19 Korean-origin brands use it somewhere against 6 of 18.
  After this change
43. v5corrected
  Free shipping in US beauty clusters at $50
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
44. v5corrected
  Instalment payment is a minority option in US beauty
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
45. v5retired
  Korean beauty prices 29% below US-origin peers
  Retired against collector run commerce-20260805053004: sample too thin (KR 2, US 0; need 6 and 4). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Korean beauty brands selling into the United States sit below their US-origin peers on price: a median storefront price of $22 across 19 Korean-origin storefronts against $31 across 18 US-origin storefronts, a gap of 29%.
  After this change
46. v5corrected
  Korean brands bring Channel Talk across the border
  Re-derived from collector run commerce-20260805015420; figures moved.
  Before
  Korean brands carry their home support stack into the United States: Channel Talk appears on 3 of 22 Korean-origin US storefronts and on none of the 23 US-origin storefronts read.
  After this change
  Korean brands carry their home support stack into the United States: Channel Talk appears on 3 of 21 Korean-origin US storefronts and on none of the 21 US-origin storefronts read.
  Note added August 5, 2026
  The revision above is labelled as a re-derivation and it was more than that. Between the previous run and this one the collector was changed: four storefronts that had been counted as brands are retailers, and a retailer's catalogue is other companies' product copy. Soko Glam is a US company whose 601 products are Korean, which made it the largest single catalogue in the US-origin group and put Korean listings inside the control set for every origin comparison. Those four are now held out of both origin groups, which is why the storefront counts in this sentence fell without the crawl reaching fewer sites. The run also re-crawled, so any part of the change that came from the market rather than from the fix cannot be separated out here. The earlier figure was wrong when we published it, not merely out of date, and the correction moves every corridor gap in the same direction: the contaminated control understated all of them. Recorded after the fact because the run was not tagged as a methodology change at the time, which was our error and not a limit of the record.
47. v5corrected
  Korean brands cross into a customer support gap
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
48. v5corrected
  Korean storefronts publish no product data, US ones do
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
49. v5retired
  Localisation depth splits, it does not scale
  Retired against collector run commerce-20260805053004: only 1 storefronts declare hreflang. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Localisation depth splits sharply rather than scaling: among the 12 US beauty storefronts that declare alternate locales at all, the median declares 15.5 of them, 5 declare twenty or more, and the deepest declares 189.
  After this change
50. v5corrected
  No review platform is universal in US beauty
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
51. v5corrected
  Odd-cent prices are 10x more Korean than US
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
52. v5corrected
  Pinterest is a US-origin habit, not a requirement
  Re-derived from collector run commerce-20260805015420; figures moved.
  Before
  Pinterest is a US-origin habit rather than a US-market requirement: 12 of 23 US-origin beauty storefronts link to it against 3 of 22 Korean-origin storefronts selling into the same market, while 45 of 66 storefronts link to TikTok regardless of origin.
  After this change
  Pinterest is a US-origin habit rather than a US-market requirement: 10 of 21 US-origin beauty storefronts link to it against 3 of 21 Korean-origin storefronts selling into the same market, while 45 of 66 storefronts link to TikTok regardless of origin.
  Note added August 5, 2026
  The revision above is labelled as a re-derivation and it was more than that. Between the previous run and this one the collector was changed: four storefronts that had been counted as brands are retailers, and a retailer's catalogue is other companies' product copy. Soko Glam is a US company whose 601 products are Korean, which made it the largest single catalogue in the US-origin group and put Korean listings inside the control set for every origin comparison. Those four are now held out of both origin groups, which is why the storefront counts in this sentence fell without the crawl reaching fewer sites. The run also re-crawled, so any part of the change that came from the market rather than from the fix cannot be separated out here. The earlier figure was wrong when we published it, not merely out of date, and the correction moves every corridor gap in the same direction: the contaminated control understated all of them. Recorded after the fact because the run was not tagged as a methodology change at the time, which was our error and not a limit of the record.
53. v5corrected
  Sets and kits are 23% of the median catalogue
  Re-derived from collector run commerce-20260805033624 after a change to the instrument: extract the jurisdiction a governing-law clause names, not just that one exists The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Sets and kits are a minority of the US beauty catalogue, not its spine: the median storefront names a set, kit, bundle or routine in 22% of product titles across 40 catalogues, with a maximum of 49%.
  After this change
  Sets and kits are a minority of the US beauty catalogue, not its spine: the median storefront names a set, kit, bundle or routine in 23% of product titles across 40 catalogues, with a maximum of 49%.
54. v5corrected
  Shopify runs the US storefront for Korean beauty
  Re-derived from collector run commerce-20260805015420; figures moved.
  Before
  Shopify is the default US storefront for Korean beauty brands: 19 of 22 Korean-origin storefronts selling into the United States run on it. The same brands at home run the Korean default instead: 16 of 19 Korean-market storefronts read are on Cafe24.
  After this change
  Shopify is the default US storefront for Korean beauty brands: 19 of 21 Korean-origin storefronts selling into the United States run on it. The same brands at home run the Korean default instead: 16 of 19 Korean-market storefronts read are on Cafe24.
  Note added August 5, 2026
  The revision above is labelled as a re-derivation and it was more than that. Between the previous run and this one the collector was changed: four storefronts that had been counted as brands are retailers, and a retailer's catalogue is other companies' product copy. Soko Glam is a US company whose 601 products are Korean, which made it the largest single catalogue in the US-origin group and put Korean listings inside the control set for every origin comparison. Those four are now held out of both origin groups, which is why the storefront counts in this sentence fell without the crawl reaching fewer sites. The run also re-crawled, so any part of the change that came from the market rather than from the fix cannot be separated out here. The earlier figure was wrong when we published it, not merely out of date, and the correction moves every corridor gap in the same direction: the contaminated control understated all of them. Recorded after the fact because the run was not tagged as a methodology change at the time, which was our error and not a limit of the record.
55. v5corrected
  Software can buy from the US shelf and not the Korean one
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
56. v5retired
  Subscription is not the default US beauty model
  Retired against collector run commerce-20260805053004: only 2 US storefronts, need 12. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Recurring purchase is not how US beauty direct-to-consumer is built: 10 of 47 storefronts run a subscription platform, and a further 10 mention subscribing in copy without one installed.
  After this change
57. v5corrected
  US brands cite a person, Korean brands cite a test
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
58. v4retired
  .99 pricing is a house style, not a market norm
  Retired against collector run commerce-20260805053004: only 2 price feeds, need 8. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Where .99 pricing does appear in US beauty direct-to-consumer it is a house style rather than a market convention: 64% of all .99 variants observed came from 2 of the 40 storefronts read, and 9 storefronts show any at all.
  After this change
59. v4corrected
  “Patch tested” appears on nothing sold into the US
  Re-derived from collector run commerce-20260805040124 after a change to the instrument: read the agent-commerce capability manifest and the sitemap index on every platform, so a Cafe24 404 is recorded as the observation it is The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The phrase “patch tested” does not appear in US-facing beauty product copy: zero of 4,864 products listed by 40 storefronts selling into the United States carry it, across Korean-origin brands, US-origin brands and the retailers stocking both.
  After this change
  The phrase “patch tested” does not appear in US-facing beauty product copy: zero of 4,867 products listed by 40 storefronts selling into the United States carry it, across Korean-origin brands, US-origin brands and the retailers stocking both.
60. v4corrected
  “Vegan” is a US-origin word, not a category word
  Re-derived from collector run commerce-20260805040124 after a change to the instrument: read the agent-commerce capability manifest and the sitemap index on every platform, so a Cafe24 404 is recorded as the observation it is The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Calling a beauty product vegan is a US-origin habit rather than a category convention: 16.2% of the 1,830 products listed by 18 US-origin brands carry the word, against 7.1% of the 2,196 listed by 19 Korean-origin brands selling into the same market.
  After this change
  Calling a beauty product vegan is a US-origin habit rather than a category convention: 16.2% of the 1,832 products listed by 18 US-origin brands carry the word, against 7.1% of the 2,197 listed by 19 Korean-origin brands selling into the same market.
61. v4corrected
  Beauty brands do not name regulators in product copy
  Re-derived from collector run commerce-20260805040124 after a change to the instrument: read the agent-commerce capability manifest and the sitemap index on every platform, so a Cafe24 404 is recorded as the observation it is The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Naming a regulator or a rating body in product copy is something US-facing beauty brands do not do: across 4,864 products from 40 storefronts, the three names together account for 26 product mentions, 0.2% for FDA, 0% for KFDA, 0.3% for EWG.
  After this change
  Naming a regulator or a rating body in product copy is something US-facing beauty brands do not do: across 4,867 products from 40 storefronts, the three names together account for 26 product mentions, 0.2% for FDA, 0% for KFDA, 0.3% for EWG.
62. v4corrected
  Charm pricing is not the default in US beauty
  Re-derived from collector run commerce-20260805015420; figures moved.
  Before
  Charm pricing is not the default in US beauty direct-to-consumer: 31 of 40 storefronts price zero variants ending in .99, and only 3% of the 8,346 variants observed end in .99, against 82% ending in a round figure.
  After this change
  Charm pricing is not the default in US beauty direct-to-consumer: 31 of 40 storefronts price zero variants ending in .99, and only 3% of the 8,347 variants observed end in .99, against 82% ending in a round figure.
  Note added August 5, 2026
  The revision above is labelled as a re-derivation and it was more than that. Between the previous run and this one the collector was changed: four storefronts that had been counted as brands are retailers, and a retailer's catalogue is other companies' product copy. Soko Glam is a US company whose 601 products are Korean, which made it the largest single catalogue in the US-origin group and put Korean listings inside the control set for every origin comparison. Those four are now held out of both origin groups, which is why the storefront counts in this sentence fell without the crawl reaching fewer sites. The run also re-crawled, so any part of the change that came from the market rather than from the fix cannot be separated out here. The earlier figure was wrong when we published it, not merely out of date, and the correction moves every corridor gap in the same direction: the contaminated control understated all of them. Recorded after the fact because the run was not tagged as a methodology change at the time, which was our error and not a limit of the record.
63. v4corrected
  Cruelty-free is claimed 2.6x more by US-origin brands
  Re-derived from collector run commerce-20260805040124 after a change to the instrument: read the agent-commerce capability manifest and the sitemap index on every platform, so a Cafe24 404 is recorded as the observation it is The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The cruelty-free claim is made far harder by US-origin brands than by Korean ones selling into the same market: 8.7% of 1,830 US-origin products carry it against 3.3% of 2,196 Korean-origin ones, and 8 of 18 US-origin brands use it at all against 6 of 19.
  After this change
  The cruelty-free claim is made far harder by US-origin brands than by Korean ones selling into the same market: 8.7% of 1,832 US-origin products carry it against 3.3% of 2,197 Korean-origin ones, and 8 of 18 US-origin brands use it at all against 6 of 19.
64. v4corrected
  Fermentation language crosses the border intact
  Re-derived from collector run commerce-20260805040124 after a change to the instrument: read the agent-commerce capability manifest and the sitemap index on every platform, so a Cafe24 404 is recorded as the observation it is The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Fermentation is the one piece of Korean formulation vocabulary that arrives in the United States untranslated: 5.4% of 2,196 Korean-origin products name it against 1.0% of 1,830 US-origin ones, and 12 of 19 Korean-origin brands use it somewhere against 6 of 18.
  After this change
  Fermentation is the one piece of Korean formulation vocabulary that arrives in the United States untranslated: 5.4% of 2,197 Korean-origin products name it against 1.0% of 1,832 US-origin ones, and 12 of 19 Korean-origin brands use it somewhere against 6 of 18.
65. v4retired
  Free shipping in US beauty clusters at $50
  Retired against collector run commerce-20260805053004: only 0 published thresholds, need 5. The claim held on the sample available when it was written and does not hold on this one.
  Before
  US beauty direct-to-consumer storefronts that publish a free shipping threshold cluster at a median of $50, across a range of $35 to $100 in a sample of 15.
  After this change
66. v4retired
  Instalment payment is a minority option in US beauty
  Retired against collector run commerce-20260805053004: only 2 storefronts declare payment methods, need 12. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Instalment payment is a minority option rather than table stakes on a US beauty storefront: 19 of 43 storefronts that advertise their payment methods offer Klarna, Afterpay, Affirm or Sezzle, while every one of them offers PayPal.
  After this change
67. v4corrected
  Korean beauty prices 29% below US-origin peers
  Re-derived from collector run commerce-20260805015420; figures moved.
  Before
  Korean beauty brands selling into the United States sit below their US-origin peers on price: a median storefront price of $22 across 19 Korean-origin storefronts against $31 across 20 US-origin storefronts, a gap of 29%.
  After this change
  Korean beauty brands selling into the United States sit below their US-origin peers on price: a median storefront price of $22 across 19 Korean-origin storefronts against $31 across 18 US-origin storefronts, a gap of 29%.
  Note added August 5, 2026
  The revision above is labelled as a re-derivation and it was more than that. Between the previous run and this one the collector was changed: four storefronts that had been counted as brands are retailers, and a retailer's catalogue is other companies' product copy. Soko Glam is a US company whose 601 products are Korean, which made it the largest single catalogue in the US-origin group and put Korean listings inside the control set for every origin comparison. Those four are now held out of both origin groups, which is why the storefront counts in this sentence fell without the crawl reaching fewer sites. The run also re-crawled, so any part of the change that came from the market rather than from the fix cannot be separated out here. The earlier figure was wrong when we published it, not merely out of date, and the correction moves every corridor gap in the same direction: the contaminated control understated all of them. Recorded after the fact because the run was not tagged as a methodology change at the time, which was our error and not a limit of the record.
68. v4corrected
  Korean brands bring Channel Talk across the border
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  Korean brands carry their home support stack into the United States: Channel Talk appears on 5 of 27 Korean-origin US storefronts and on none of the 23 US-origin storefronts read.
  After this change
  Korean brands carry their home support stack into the United States: Channel Talk appears on 3 of 22 Korean-origin US storefronts and on none of the 23 US-origin storefronts read.
69. v4retired
  Korean brands cross into a customer support gap
  Retired against collector run commerce-20260805053004: only 0 brands read on both sides, need 6. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Korean beauty brands drop their home support channel at the border without replacing it: across 13 brands whose Korean and US storefronts were both read, 9 run Kakao support at home and 0 still do in the United States, but only 3 install a US help desk such as Gorgias, Zendesk or Intercom, against 11 of 21 US-origin storefronts.
  After this change
70. v4corrected
  Korean shelves are 4.2 months younger
  Re-derived from collector run commerce-20260805140548 after a change to the instrument: added the product page as a readable surface: the same page the collector already fetched for its JSON-LD, now read for the ingredient list, the drug label, the directions and the claim words a shopper actually sees. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Korean-origin beauty brands keep a visibly newer shelf in the United States than their US-origin peers: the median product on a Korean-origin storefront was published 461 days ago against 585 days for US-origin ones, across 19 and 18 catalogues.
  After this change
  Korean-origin beauty brands keep a visibly newer shelf in the United States than their US-origin peers: the median product on a Korean-origin storefront was published 461 days ago against 586 days for US-origin ones, across 19 and 18 catalogues.
71. v4retired
  Korean storefronts publish no product data, US ones do
  Retired against collector run commerce-20260805053004: sample too thin (US 2, KR home 0). The claim held on the sample available when it was written and does not hold on this one.
  Before
  The Korean beauty shelf and the US one are not equally readable by software: 39 of 47 storefronts selling into the United States publish a complete machine-readable product feed, on the back of 41 running Shopify, while 0 of 19 storefronts read in the Korean home market publish one at all, 0 run Shopify, and 16 run Cafe24, which serves prices as rendered pages rather than as data.
  After this change
72. v4corrected
  Localisation depth splits, it does not scale
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  Localisation depth splits sharply rather than scaling: among the 16 US beauty storefronts that declare alternate locales at all, the median declares 8 of them, 5 declare twenty or more, and the deepest declares 189.
  After this change
  Localisation depth splits sharply rather than scaling: among the 12 US beauty storefronts that declare alternate locales at all, the median declares 15.5 of them, 5 declare twenty or more, and the deepest declares 189.
73. v4retired
  No review platform is universal in US beauty
  Retired against collector run commerce-20260805053004: only 2 storefronts carry a review platform, need 12. The claim held on the sample available when it was written and does not hold on this one.
  Before
  No single review platform owns US beauty: the most common, Yotpo, appears on 20 of the 42 storefronts that carry reviews at all, 7 distinct platforms are in use across the set, and 13 storefronts run more than one at once.
  After this change
74. v4corrected
  Nobody on this shelf has written an AI crawler rule
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
75. v4retired
  Odd-cent prices are 10x more Korean than US
  Retired against collector run commerce-20260805053004: price feeds too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  A price that lands on neither a whole dollar nor .99 is a Korean-origin habit on the US beauty shelf: 30% of 3,380 variants from 19 Korean-origin storefronts end that way against 3% of 3,986 variants from 18 US-origin ones, and 15 of 19 Korean storefronts clear 5% against 3 of 18. A round figure is still the most common ending on both sides.
  After this change
76. v4corrected
  Pinterest is a US-origin habit, not a requirement
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  Pinterest is a US-origin habit rather than a US-market requirement: 12 of 23 US-origin beauty storefronts link to it against 3 of 27 Korean-origin storefronts selling into the same market, while 42 of 63 storefronts link to TikTok regardless of origin.
  After this change
  Pinterest is a US-origin habit rather than a US-market requirement: 12 of 23 US-origin beauty storefronts link to it against 3 of 22 Korean-origin storefronts selling into the same market, while 45 of 66 storefronts link to TikTok regardless of origin.
77. v4corrected
  Sets and kits are 23% of the median catalogue
  Re-derived from collector run commerce-20260805025415 after a change to the instrument: Adds the Shopify return-policy family: four corridor claims on sealed-goods conditions, the 30-day window, transit-liability topics and policy length. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Sets and kits are a minority of the US beauty catalogue, not its spine: the median storefront names a set, kit, bundle or routine in 23% of product titles across 40 catalogues, with a maximum of 49%.
  After this change
  Sets and kits are a minority of the US beauty catalogue, not its spine: the median storefront names a set, kit, bundle or routine in 22% of product titles across 40 catalogues, with a maximum of 49%.
78. v4corrected
  Shopify runs the US storefront for Korean beauty
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  Shopify is the default US storefront for Korean beauty brands: 18 of 27 Korean-origin storefronts selling into the United States run on it. The same brands at home run the Korean default instead: 8 of 11 Korean-market storefronts read are on Cafe24.
  After this change
  Shopify is the default US storefront for Korean beauty brands: 19 of 22 Korean-origin storefronts selling into the United States run on it. The same brands at home run the Korean default instead: 16 of 19 Korean-market storefronts read are on Cafe24.
79. v4retired
  Software can buy from the US shelf and not the Korean one
  Retired against collector run commerce-20260805053004: sample too thin (US 2, KR home 0). The claim held on the sample available when it was written and does not hold on this one.
  Before
  41 of 47 storefronts selling into the United States publish a machine-readable commerce manifest at “/.well-known/ucp” declaring which operations an autonomous shopping agent may run against them, 41 of those declaring a checkout capability, while 0 of 19 storefronts read in the Korean home market publish one at all.
  After this change
80. v4corrected
  Subscription is not the default US beauty model
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  Recurring purchase is not how US beauty direct-to-consumer is built: 10 of 52 storefronts run a subscription platform, and a further 10 mention subscribing in copy without one installed.
  After this change
  Recurring purchase is not how US beauty direct-to-consumer is built: 10 of 47 storefronts run a subscription platform, and a further 10 mention subscribing in copy without one installed.
81. v4corrected
  The K-beauty price gap holds at every rung of the ladder
  Re-derived from collector run commerce-20260805140548 after a change to the instrument: added the product page as a readable surface: the same page the collector already fetched for its JSON-LD, now read for the ingredient list, the drug label, the directions and the claim words a shopper actually sees. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Korean-origin beauty brands sell into the United States below their US-origin peers by roughly the same margin at every price point rather than only at the top or the bottom: $16.50 against $24 at the cheap quarter, $22 against $31 at the middle, $32.80 against $42 at the expensive quarter, gaps of 31%, 29%, 22% across 19 Korean-origin and 18 US-origin storefronts.
  After this change
  Korean-origin beauty brands sell into the United States below their US-origin peers by roughly the same margin at every price point rather than only at the top or the bottom: $16.99 against $24 at the cheap quarter, $22 against $31 at the middle, $32.80 against $42 at the expensive quarter, gaps of 29%, 29%, 22% across 19 Korean-origin and 18 US-origin storefronts.
82. v4retired
  US brands cite a person, Korean brands cite a test
  Retired against collector run commerce-20260805053004: catalogues too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  The two sides of this corridor borrow authority from different places: US-origin brands invoke a dermatologist on 8.4% of their products against 1.6% for Korean-origin brands, while 17 of 19 Korean-origin brands use clinical-testing language somewhere in the catalogue against 11 of 18 US-origin ones. One names a professional, the other names a procedure.
  After this change
83. v3corrected
  .99 pricing is a house style, not a market norm
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  Where .99 pricing does appear in US beauty direct-to-consumer it is a house style rather than a market convention: 64% of all .99 variants observed came from 2 of the 39 storefronts read, and 9 storefronts show any at all.
  After this change
  Where .99 pricing does appear in US beauty direct-to-consumer it is a house style rather than a market convention: 64% of all .99 variants observed came from 2 of the 40 storefronts read, and 9 storefronts show any at all.
84. v3corrected
  “Patch tested” appears on nothing sold into the US
  Re-derived from collector run commerce-20260805033624 after a change to the instrument: extract the jurisdiction a governing-law clause names, not just that one exists The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The phrase “patch tested” does not appear in US-facing beauty product copy: zero of 4,863 products listed by 40 storefronts selling into the United States carry it, across Korean-origin brands, US-origin brands and the retailers stocking both.
  After this change
  The phrase “patch tested” does not appear in US-facing beauty product copy: zero of 4,864 products listed by 40 storefronts selling into the United States carry it, across Korean-origin brands, US-origin brands and the retailers stocking both.
85. v3corrected
  “Vegan” is a US-origin word, not a category word
  Re-derived from collector run commerce-20260805033624 after a change to the instrument: extract the jurisdiction a governing-law clause names, not just that one exists The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Calling a beauty product vegan is a US-origin habit rather than a category convention: 16.2% of the 1,830 products listed by 18 US-origin brands carry the word, against 7.1% of the 2,195 listed by 19 Korean-origin brands selling into the same market.
  After this change
  Calling a beauty product vegan is a US-origin habit rather than a category convention: 16.2% of the 1,830 products listed by 18 US-origin brands carry the word, against 7.1% of the 2,196 listed by 19 Korean-origin brands selling into the same market.
86. v3corrected
  45% of beauty terms pages are shared documents
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
87. v3corrected
  5 in 15 Korean storefronts sell under Korean law
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
88. v3corrected
  A Korean brand’s US legal terms read like an American one’s
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
89. v3corrected
  Beauty brands do not name regulators in product copy
  Re-derived from collector run commerce-20260805033624 after a change to the instrument: extract the jurisdiction a governing-law clause names, not just that one exists The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Naming a regulator or a rating body in product copy is something US-facing beauty brands do not do: across 4,863 products from 40 storefronts, the three names together account for 26 product mentions, 0.2% for FDA, 0% for KFDA, 0.3% for EWG.
  After this change
  Naming a regulator or a rating body in product copy is something US-facing beauty brands do not do: across 4,864 products from 40 storefronts, the three names together account for 26 product mentions, 0.2% for FDA, 0% for KFDA, 0.3% for EWG.
90. v3corrected
  Charm pricing is not the default in US beauty
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  Charm pricing is not the default in US beauty direct-to-consumer: 30 of 39 storefronts price zero variants ending in .99, and only 3% of the 8,271 variants observed end in .99, against 82% ending in a round figure.
  After this change
  Charm pricing is not the default in US beauty direct-to-consumer: 31 of 40 storefronts price zero variants ending in .99, and only 3% of the 8,346 variants observed end in .99, against 82% ending in a round figure.
91. v3corrected
  Cruelty-free is claimed 2.6x more by US-origin brands
  Re-derived from collector run commerce-20260805033624 after a change to the instrument: extract the jurisdiction a governing-law clause names, not just that one exists The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The cruelty-free claim is made far harder by US-origin brands than by Korean ones selling into the same market: 8.7% of 1,830 US-origin products carry it against 3.3% of 2,195 Korean-origin ones, and 8 of 18 US-origin brands use it at all against 6 of 19.
  After this change
  The cruelty-free claim is made far harder by US-origin brands than by Korean ones selling into the same market: 8.7% of 1,830 US-origin products carry it against 3.3% of 2,196 Korean-origin ones, and 8 of 18 US-origin brands use it at all against 6 of 19.
92. v3corrected
  Deep shade ranges stay on the US side of the corridor
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
93. v3corrected
  Discount posture is set per market, not per brand
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
94. v3corrected
  Fermentation language crosses the border intact
  Re-derived from collector run commerce-20260805033624 after a change to the instrument: extract the jurisdiction a governing-law clause names, not just that one exists The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Fermentation is the one piece of Korean formulation vocabulary that arrives in the United States untranslated: 5.4% of 2,195 Korean-origin products name it against 1.0% of 1,830 US-origin ones, and 12 of 19 Korean-origin brands use it somewhere against 6 of 18.
  After this change
  Fermentation is the one piece of Korean formulation vocabulary that arrives in the United States untranslated: 5.4% of 2,196 Korean-origin products name it against 1.0% of 1,830 US-origin ones, and 12 of 19 Korean-origin brands use it somewhere against 6 of 18.
95. v3corrected
  Free shipping in US beauty clusters at $50
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  US beauty direct-to-consumer storefronts that publish a free shipping threshold cluster at a median of $50, across a range of $35 to $100 in a sample of 14.
  After this change
  US beauty direct-to-consumer storefronts that publish a free shipping threshold cluster at a median of $50, across a range of $35 to $100 in a sample of 15.
96. v3corrected
  Instalment payment is a minority option in US beauty
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  Instalment payment is a minority option rather than table stakes on a US beauty storefront: 19 of 43 storefronts that advertise their payment methods offer Klarna, Afterpay, Affirm or Sezzle, against 42 that offer PayPal.
  After this change
  Instalment payment is a minority option rather than table stakes on a US beauty storefront: 19 of 43 storefronts that advertise their payment methods offer Klarna, Afterpay, Affirm or Sezzle, while every one of them offers PayPal.
97. v3corrected
  Instalment payment is a US-origin habit in beauty
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
98. v3corrected
  Korean beauty prices 29% below US-origin peers
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  Korean beauty brands selling into the United States sit below their US-origin peers on price: a median storefront price of $22.5 across 18 Korean-origin storefronts against $31 across 20 US-origin storefronts, a gap of 27%.
  After this change
  Korean beauty brands selling into the United States sit below their US-origin peers on price: a median storefront price of $22 across 19 Korean-origin storefronts against $31 across 20 US-origin storefronts, a gap of 29%.
99. v3corrected
  Korean brands cross into a customer support gap
  Re-derived from collector run commerce-20260805015420; figures moved.
  Before
  Korean beauty brands drop their home support channel at the border without replacing it: across 13 brands whose Korean and US storefronts were both read, 9 run Kakao support at home and 0 still do in the United States, but only 3 install a US help desk such as Gorgias, Zendesk or Intercom, against 12 of 23 US-origin storefronts.
  After this change
  Korean beauty brands drop their home support channel at the border without replacing it: across 13 brands whose Korean and US storefronts were both read, 9 run Kakao support at home and 0 still do in the United States, but only 3 install a US help desk such as Gorgias, Zendesk or Intercom, against 11 of 21 US-origin storefronts.
  Note added August 5, 2026
  The revision above is labelled as a re-derivation and it was more than that. Between the previous run and this one the collector was changed: four storefronts that had been counted as brands are retailers, and a retailer's catalogue is other companies' product copy. Soko Glam is a US company whose 601 products are Korean, which made it the largest single catalogue in the US-origin group and put Korean listings inside the control set for every origin comparison. Those four are now held out of both origin groups, which is why the storefront counts in this sentence fell without the crawl reaching fewer sites. The run also re-crawled, so any part of the change that came from the market rather than from the fix cannot be separated out here. The earlier figure was wrong when we published it, not merely out of date, and the correction moves every corridor gap in the same direction: the contaminated control understated all of them. Recorded after the fact because the run was not tagged as a methodology change at the time, which was our error and not a limit of the record.
100. v3corrected
  Korean brands sell the US shelf on a struck price
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
101. v3corrected
  Korean brands ship a thinner US checkout
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
102. v3corrected
  Korean catalogues are filed by skin concern
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
103. v3corrected
  Korean shelves are 4.2 months younger
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
104. v3corrected
  Korean storefronts arrive with a thinner archive
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
105. v3corrected
  Korean storefronts publish no product data, US ones do
  Re-derived from collector run commerce-20260805051110 after a change to the instrument: added two page families to the instrument: the blog sitemap child plus Atom feed, which measures what a storefront actually publishes rather than whether the platform gave it a blog, and robots.txt read as data, which records which crawlers a shop lets in and reads identically on Cafe24 and Shopify The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The Korean beauty shelf and the US one are not equally readable by software: 39 of 47 storefronts selling into the United States publish a complete machine-readable product feed, on the back of 41 running Shopify, while 0 of 18 storefronts read in the Korean home market publish one at all, 0 run Shopify, and 15 run Cafe24, which serves prices as rendered pages rather than as data.
  After this change
  The Korean beauty shelf and the US one are not equally readable by software: 39 of 47 storefronts selling into the United States publish a complete machine-readable product feed, on the back of 41 running Shopify, while 0 of 19 storefronts read in the Korean home market publish one at all, 0 run Shopify, and 16 run Cafe24, which serves prices as rendered pages rather than as data.
106. v3corrected
  Korean-origin policies argue about the parcel first
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
107. v3corrected
  Korean-origin return policies run 4.3x longer
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
108. v3corrected
  Korean-origin returns have to come back sealed
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
109. v3corrected
  Most US beauty storefronts run no loyalty platform
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
110. v3corrected
  No review platform is universal in US beauty
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  No single review platform owns US beauty: the most common, Yotpo, appears on 20 of the 41 storefronts that carry reviews at all, 7 distinct platforms are in use across the set, and 13 storefronts run more than one at once.
  After this change
  No single review platform owns US beauty: the most common, Yotpo, appears on 20 of the 42 storefronts that carry reviews at all, 7 distinct platforms are in use across the set, and 13 storefronts run more than one at once.
111. v3retired
  Nobody on this shelf has written an AI crawler rule
  Retired against collector run commerce-20260805053004: robots policy read for only 2 storefronts, need 30. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Of 47 beauty storefronts selling into the United States, 3 name any of fourteen known AI crawlers in their robots.txt and 1 disallows one, while 1 publishes no crawler rules at all. The category has not decided to let AI systems read it; it has not considered the question.
  After this change
112. v3corrected
  Odd-cent prices are 10x more Korean than US
  Re-derived from collector run commerce-20260805040124 after a change to the instrument: read the agent-commerce capability manifest and the sitemap index on every platform, so a Cafe24 404 is recorded as the observation it is The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  A price that lands on neither a whole dollar nor .99 is a Korean-origin habit on the US beauty shelf: 30% of 3,379 variants from 19 Korean-origin storefronts end that way against 3% of 3,985 variants from 18 US-origin ones, and 15 of 19 Korean storefronts clear 5% against 3 of 18. A round figure is still the most common ending on both sides.
  After this change
  A price that lands on neither a whole dollar nor .99 is a Korean-origin habit on the US beauty shelf: 30% of 3,380 variants from 19 Korean-origin storefronts end that way against 3% of 3,986 variants from 18 US-origin ones, and 15 of 19 Korean storefronts clear 5% against 3 of 18. A round figure is still the most common ending on both sides.
113. v3corrected
  Sets and kits are 23% of the median catalogue
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  Sets and kits are a minority of the US beauty catalogue, not its spine: the median storefront names a set, kit, bundle or routine in 22% of product titles across 39 catalogues, with a maximum of 49%.
  After this change
  Sets and kits are a minority of the US beauty catalogue, not its spine: the median storefront names a set, kit, bundle or routine in 23% of product titles across 40 catalogues, with a maximum of 49%.
114. v3corrected
  Software can buy from the US shelf and not the Korean one
  Re-derived from collector run commerce-20260805051110 after a change to the instrument: added two page families to the instrument: the blog sitemap child plus Atom feed, which measures what a storefront actually publishes rather than whether the platform gave it a blog, and robots.txt read as data, which records which crawlers a shop lets in and reads identically on Cafe24 and Shopify The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  41 of 47 storefronts selling into the United States publish a machine-readable commerce manifest at “/.well-known/ucp” declaring which operations an autonomous shopping agent may run against them, 41 of those declaring a checkout capability, while 0 of 18 storefronts read in the Korean home market publish one at all.
  After this change
  41 of 47 storefronts selling into the United States publish a machine-readable commerce manifest at “/.well-known/ucp” declaring which operations an autonomous shopping agent may run against them, 41 of those declaring a checkout capability, while 0 of 19 storefronts read in the Korean home market publish one at all.
115. v3corrected
  The K-beauty price gap holds at every rung of the ladder
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
116. v3corrected
  The two origins buy opposite review software
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
117. v3corrected
  Thirty days is a US default, not a Korean one
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
118. v3corrected
  TikTok is an export behaviour for Korean beauty
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
119. v3corrected
  US beauty settled on one helpdesk, Korean brands did not
  Correction. The retirement recorded above was an operator error, not a finding. The derivation was run against collector run commerce-20260805053004 while that crawl was still in flight and had reached 4 of 66 storefronts, so every claim whose storefronts it had not yet read failed its sample gate and was retired. 44 claims were retired in one command for that reason. The claim was re-derived the same day from collector run commerce-20260805051110, a complete 66 of 66 crawl, and holds. Nothing about the market changed between the two records. The retirement stays in this log because the log is not edited; this entry is what it means. The instrument now refuses to derive from a run missing more than a tenth of the seeded storefronts, and a claim returning from retired now files its own status revision, which is why this correction had to be written by hand rather than appearing on its own.
120. v3corrected
  US brands cite a person, Korean brands cite a test
  Re-derived from collector run commerce-20260805033624 after a change to the instrument: extract the jurisdiction a governing-law clause names, not just that one exists The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The two sides of this corridor borrow authority from different places: US-origin brands invoke a dermatologist on 8.4% of their products against 1.5% for Korean-origin brands, while 17 of 19 Korean-origin brands use clinical-testing language somewhere in the catalogue against 11 of 18 US-origin ones. One names a professional, the other names a procedure.
  After this change
  The two sides of this corridor borrow authority from different places: US-origin brands invoke a dermatologist on 8.4% of their products against 1.6% for Korean-origin brands, while 17 of 19 Korean-origin brands use clinical-testing language somewhere in the catalogue against 11 of 18 US-origin ones. One names a professional, the other names a procedure.
121. v2corrected
  “Patch tested” appears on nothing sold into the US
  Re-derived from collector run commerce-20260805025415 after a change to the instrument: Adds the Shopify return-policy family: four corridor claims on sealed-goods conditions, the 30-day window, transit-liability topics and policy length. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The phrase “patch tested” does not appear in US-facing beauty product copy: zero of 4,853 products listed by 40 storefronts selling into the United States carry it, across Korean-origin brands, US-origin brands and the retailers stocking both.
  After this change
  The phrase “patch tested” does not appear in US-facing beauty product copy: zero of 4,863 products listed by 40 storefronts selling into the United States carry it, across Korean-origin brands, US-origin brands and the retailers stocking both.
122. v2corrected
  “Vegan” is a US-origin word, not a category word
  Re-derived from collector run commerce-20260805025415 after a change to the instrument: Adds the Shopify return-policy family: four corridor claims on sealed-goods conditions, the 30-day window, transit-liability topics and policy length. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Calling a beauty product vegan is a US-origin habit rather than a category convention: 16.2% of the 1,830 products listed by 18 US-origin brands carry the word, against 7.1% of the 2,185 listed by 19 Korean-origin brands selling into the same market.
  After this change
  Calling a beauty product vegan is a US-origin habit rather than a category convention: 16.2% of the 1,830 products listed by 18 US-origin brands carry the word, against 7.1% of the 2,195 listed by 19 Korean-origin brands selling into the same market.
123. v2retired
  45% of beauty terms pages are shared documents
  Retired against collector run commerce-20260805053004: only 2 terms pages read, need 20. The claim held on the sample available when it was written and does not hold on this one.
  Before
  The terms of service on a direct-to-consumer beauty storefront is usually not that brand’s document: 15 of 33 read here share at least 60% of their text with another brand’s, in 3 pools of which the largest covers 9 storefronts and one pool spans both origins.
  After this change
124. v2retired
  5 in 15 Korean storefronts sell under Korean law
  Retired against collector run commerce-20260805053004: only 2 Korean-origin storefronts name a governing law, need 8. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Some Korean-origin brands selling to American shoppers put the sale under Korean law rather than US law: 5 of the 15 that name a governing law name South Korea or the Republic of Korea, while 0 of 6 US-origin storefronts name any law outside the United States.
  After this change
125. v2retired
  A Korean brand’s US legal terms read like an American one’s
  Retired against collector run commerce-20260805053004: terms reads too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Nothing in the legal clauses of a US-facing storefront’s terms separates Korean-origin brands from US-origin ones: arbitration on 9 of 18 Korean-origin and 6 of 15 US-origin, jury waiver on 6 of 18 Korean-origin and 5 of 15 US-origin, class waiver on 6 of 18 Korean-origin and 4 of 15 US-origin, governing law on 15 of 18 Korean-origin and 12 of 15 US-origin, every gap under 10 points.
  After this change
126. v2corrected
  Beauty brands do not name regulators in product copy
  Re-derived from collector run commerce-20260805025415 after a change to the instrument: Adds the Shopify return-policy family: four corridor claims on sealed-goods conditions, the 30-day window, transit-liability topics and policy length. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Naming a regulator or a rating body in product copy is something US-facing beauty brands do not do: across 4,853 products from 40 storefronts, the three names together account for 26 product mentions, 0.2% for FDA, 0% for KFDA, 0.3% for EWG.
  After this change
  Naming a regulator or a rating body in product copy is something US-facing beauty brands do not do: across 4,863 products from 40 storefronts, the three names together account for 26 product mentions, 0.2% for FDA, 0% for KFDA, 0.3% for EWG.
127. v2corrected
  Cruelty-free is claimed 2.6x more by US-origin brands
  Re-derived from collector run commerce-20260805025415 after a change to the instrument: Adds the Shopify return-policy family: four corridor claims on sealed-goods conditions, the 30-day window, transit-liability topics and policy length. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The cruelty-free claim is made far harder by US-origin brands than by Korean ones selling into the same market: 8.7% of 1,830 US-origin products carry it against 3.3% of 2,185 Korean-origin ones, and 8 of 18 US-origin brands use it at all against 6 of 19.
  After this change
  The cruelty-free claim is made far harder by US-origin brands than by Korean ones selling into the same market: 8.7% of 1,830 US-origin products carry it against 3.3% of 2,195 Korean-origin ones, and 8 of 18 US-origin brands use it at all against 6 of 19.
128. v2retired
  Deep shade ranges stay on the US side of the corridor
  Retired against collector run commerce-20260805053004: catalogues with an option-axis reading too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Shipping a product in twelve or more shades is a US-origin convention that Korean-origin storefronts selling here have not adopted: 9 of 18 US-origin catalogues carry at least one such product against 4 of 19 Korean-origin ones, with the deepest range on each side at 48 shades (Rare Beauty) and 40 (TIRTIR).
  After this change
129. v2retired
  Discount posture is set per market, not per brand
  Retired against collector run commerce-20260805053004: only 0 brands with anchor data on both sides, need 6. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Whether a Korean beauty brand prices against a struck-through anchor is decided per storefront, not carried across the border: of 8 brands read on both sides, 6 changed their anchored share by at least 25 points, a median 51 points, and they moved both ways, 4 anchoring more in the US than at home and 2 anchoring more at home.
  After this change
130. v2corrected
  Fermentation language crosses the border intact
  Re-derived from collector run commerce-20260805025415 after a change to the instrument: Adds the Shopify return-policy family: four corridor claims on sealed-goods conditions, the 30-day window, transit-liability topics and policy length. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Fermentation is the one piece of Korean formulation vocabulary that arrives in the United States untranslated: 5.4% of 2,185 Korean-origin products name it against 1.0% of 1,830 US-origin ones, and 12 of 19 Korean-origin brands use it somewhere against 6 of 18.
  After this change
  Fermentation is the one piece of Korean formulation vocabulary that arrives in the United States untranslated: 5.4% of 2,195 Korean-origin products name it against 1.0% of 1,830 US-origin ones, and 12 of 19 Korean-origin brands use it somewhere against 6 of 18.
131. v2retired
  Instalment payment is a US-origin habit in beauty
  Retired against collector run commerce-20260805053004: payment reads too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  The half of US beauty that offers instalment payment is almost entirely one side of the corridor: 13 of 19 US-origin storefronts advertise Klarna, Afterpay, Affirm or Sezzle against 4 of 19 Korean-origin ones, on storefronts where both groups advertise PayPal at the same rate.
  After this change
132. v2corrected
  Korean brands cross into a customer support gap
  Re-derived from collector run commerce-20260805005712; figures moved.
  Before
  Korean beauty brands drop their home support channel at the border without replacing it: across 9 brands whose Korean and US storefronts were both read, 8 run Kakao support at home and 1 still does in the United States, but only 2 install a US help desk such as Gorgias, Zendesk or Intercom, against 12 of 23 US-origin storefronts.
  After this change
  Korean beauty brands drop their home support channel at the border without replacing it: across 13 brands whose Korean and US storefronts were both read, 9 run Kakao support at home and 0 still do in the United States, but only 3 install a US help desk such as Gorgias, Zendesk or Intercom, against 12 of 23 US-origin storefronts.
133. v2retired
  Korean brands sell the US shelf on a struck price
  Retired against collector run commerce-20260805053004: anchor reads too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Korean beauty brands run their US storefronts on a marked-down price in a way their US-origin competitors do not: a median 41% of the Korean-origin catalogue carries a struck-through anchor against 19% of the US-origin one, and 13 of 19 Korean storefronts have at least a third of the shelf marked down against 4 of 18.
  After this change
134. v2retired
  Korean brands ship a thinner US checkout
  Retired against collector run commerce-20260805053004: payment reads too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Korean beauty brands offer fewer ways to pay on their US storefronts than their US-origin competitors: a median 2 advertised payment methods against 3, and 8 of 19 Korean-origin storefronts advertise three or more against 14 of 19 US-origin ones.
  After this change
135. v2retired
  Korean catalogues are filed by skin concern
  Retired against collector run commerce-20260805053004: tag vocabularies too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  The tags a storefront publishes on its own products split by origin: a median of 16% of a Korean-origin catalogue’s tag vocabulary names a skin concern against 2% of a US-origin one’s, while 6% of the US-origin vocabulary names a back-office state such as a promotion window or a feature flag against 3% Korean-origin.
  After this change
136. v2retired
  Korean shelves are 4.2 months younger
  Retired against collector run commerce-20260805053004: sample too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Korean-origin beauty brands keep a visibly newer shelf in the United States than their US-origin peers: the median product on a Korean-origin storefront was published 461 days ago against 585 days for US-origin ones, across 19 and 18 catalogues.
  After this change
137. v2retired
  Korean storefronts arrive with a thinner archive
  Retired against collector run commerce-20260805053004: article counts read for KR 2, US 0; need 8 each. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Counted from each storefront’s own blog sitemap, a US-origin beauty brand selling into the United States has published a median of 16.5 articles against 7 for a Korean-origin brand selling into the same market, with 8 of 19 Korean-origin storefronts holding five articles or fewer and Milk Makeup holding 273.
  After this change
138. v2corrected
  Korean storefronts publish no product data, US ones do
  Re-derived from collector run commerce-20260805040124 after a change to the instrument: read the agent-commerce capability manifest and the sitemap index on every platform, so a Cafe24 404 is recorded as the observation it is The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The Korean beauty shelf and the US one are not equally readable by software: 39 of 47 storefronts selling into the United States publish a complete machine-readable product feed, on the back of 41 running Shopify, while 0 of 19 storefronts read in the Korean home market publish one at all, 0 run Shopify, and 16 run Cafe24, which serves prices as rendered pages rather than as data.
  After this change
  The Korean beauty shelf and the US one are not equally readable by software: 39 of 47 storefronts selling into the United States publish a complete machine-readable product feed, on the back of 41 running Shopify, while 0 of 18 storefronts read in the Korean home market publish one at all, 0 run Shopify, and 15 run Cafe24, which serves prices as rendered pages rather than as data.
139. v2retired
  Korean-origin policies argue about the parcel first
  Retired against collector run commerce-20260805053004: policy reads too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  A Korean-origin brand’s US returns page spends its length on the parcel and a US-origin one does not mention it: 14 of 18 Korean-origin policies raise customs, carriers, or a lost or stolen box, against 3 of 13 US-origin policies. Customs appears on 7 Korean-origin pages and 0 US-origin ones.
  After this change
140. v2retired
  Korean-origin return policies run 4.3x longer
  Retired against collector run commerce-20260805053004: policy reads too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Korean-origin beauty brands write a far longer returns page into the United States than their US-origin peers: a median of 4,243 characters across 18 policies against 990 across 13, a factor of 4.3.
  After this change
141. v2retired
  Korean-origin returns have to come back sealed
  Retired against collector run commerce-20260805053004: policy reads too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Korean-origin beauty brands selling into the United States write a sealed-goods condition into their return policy and their US-origin peers do not: 14 of 18 Korean-origin policies require the product unopened or factory-sealed, against 2 of 13 US-origin ones, 6 of which say in writing that opened or gently used product can come back.
  After this change
142. v2retired
  Most US beauty storefronts run no loyalty platform
  Retired against collector run commerce-20260805053004: only 2 storefronts read, need 25. The claim held on the sample available when it was written and does not hold on this one.
  Before
  Running a loyalty programme is a minority choice in US beauty direct-to-consumer rather than the default: 14 of 47 storefronts have a named loyalty platform installed, and even counting every storefront that merely uses the words rewards or points anywhere on the page only reaches 24.
  After this change
143. v2corrected
  Nobody on this shelf has written an AI crawler rule
  Re-derived from collector run commerce-20260805051110 after a change to the instrument: verb agreement in the AI-crawler statement, which computed a count that landed on one and read as 1 publish The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Of 47 beauty storefronts selling into the United States, 3 name any of fourteen known AI crawlers in their robots.txt and 1 disallow one, while 1 publish no crawler rules at all. The category has not decided to let AI systems read it; it has not considered the question.
  After this change
  Of 47 beauty storefronts selling into the United States, 3 name any of fourteen known AI crawlers in their robots.txt and 1 disallows one, while 1 publishes no crawler rules at all. The category has not decided to let AI systems read it; it has not considered the question.
144. v2corrected
  Odd-cent prices are 10x more Korean than US
  Re-derived from collector run commerce-20260805033624 after a change to the instrument: extract the jurisdiction a governing-law clause names, not just that one exists The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  A price that lands on neither a whole dollar nor .99 is a Korean-origin habit on the US beauty shelf: 30% of 3,378 variants from 19 Korean-origin storefronts end that way against 3% of 3,985 variants from 18 US-origin ones, and 15 of 19 Korean storefronts clear 5% against 3 of 18. A round figure is still the most common ending on both sides.
  After this change
  A price that lands on neither a whole dollar nor .99 is a Korean-origin habit on the US beauty shelf: 30% of 3,379 variants from 19 Korean-origin storefronts end that way against 3% of 3,985 variants from 18 US-origin ones, and 15 of 19 Korean storefronts clear 5% against 3 of 18. A round figure is still the most common ending on both sides.
145. v2corrected
  Software can buy from the US shelf and not the Korean one
  Re-derived from collector run commerce-20260805040124 after a change to the instrument: replace markdown backticks with quotation marks in claim prose; the site and every downstream consumer render claim text as plain prose, so a backtick was showing up literally in a sentence meant to be quotable The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  41 of 47 storefronts selling into the United States publish a machine-readable commerce manifest at `/.well-known/ucp` declaring which operations an autonomous shopping agent may run against them, 41 of those declaring a checkout capability, while 0 of 18 storefronts read in the Korean home market publish one at all.
  After this change
  41 of 47 storefronts selling into the United States publish a machine-readable commerce manifest at “/.well-known/ucp” declaring which operations an autonomous shopping agent may run against them, 41 of those declaring a checkout capability, while 0 of 18 storefronts read in the Korean home market publish one at all.
146. v2retired
  The K-beauty price gap holds at every rung of the ladder
  Retired against collector run commerce-20260805053004: sample too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Korean-origin beauty brands sell into the United States below their US-origin peers by roughly the same margin at every price point rather than only at the top or the bottom: $16.50 against $24 at the cheap quarter, $22 against $31 at the middle, $32.80 against $42 at the expensive quarter, gaps of 31%, 29%, 22% across 19 Korean-origin and 18 US-origin storefronts.
  After this change
147. v2corrected
  The SPF page skips the label the law defines
  Re-derived from collector run commerce-20260805140548 after a change to the instrument: corrected subject-verb agreement in the SPF drug-label sentence. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  Of 8 beauty product pages selling into the United States that advertise an SPF value, 1 publish a Drug Facts panel and 2 name an active ingredient, the two elements that distinguish an over-the-counter drug label from marketing copy, while 0 carry any warnings text at all.
  After this change
  Of 8 beauty product pages selling into the United States that advertise an SPF value, 1 publishes a Drug Facts panel and 2 name an active ingredient, the two elements that distinguish an over-the-counter drug label from marketing copy, while 0 carry any warning text at all.
148. v2retired
  The two origins buy opposite review software
  Retired against collector run commerce-20260805053004: review reads too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  Agreement on Shopify does not extend to what gets installed on it: of the storefronts running a named review platform, 11 of 19 Korean-origin ones run Judge.me against 2 of 19 US-origin ones, while 11 of 19 US-origin ones run Okendo against 4 of 19 Korean-origin. Both groups install review software and they pick different software.
  After this change
149. v2retired
  Thirty days is a US default, not a Korean one
  Retired against collector run commerce-20260805053004: stated windows too thin (KR 2, US 0; need 8 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  The United States has one default return window and Korean-origin sellers into it do not share it: 9 of 10 US-origin policies that state a window set it at exactly 30 days, against 5 of 10 Korean-origin ones, which split between 14 and 30 days.
  After this change
150. v2retired
  TikTok is an export behaviour for Korean beauty
  Retired against collector run commerce-20260805053004: only 0 brands read on both sides, need 6. The claim held on the sample available when it was written and does not hold on this one.
  Before
  TikTok is something Korean beauty brands do for the United States rather than something they do: across 13 brands whose Korean and US storefronts were both read, 10 link to TikTok from the US storefront and 4 link to it from the Korean one, while Instagram sits on 13 at home and 11 abroad.
  After this change
151. v2retired
  US beauty settled on one helpdesk, Korean brands did not
  Retired against collector run commerce-20260805053004: helpdesk reads too thin (KR 1, US 0; need 6 each). The claim held on the sample available when it was written and does not hold on this one.
  Before
  US-origin beauty has converged on a single customer support tool and the Korean brands selling beside them have not: every one of the 11 US-origin storefronts running a named helpdesk runs Gorgias, against 3 of the 9 Korean-origin ones, which spread across 3 other tools.
  After this change
152. v2corrected
  US brands cite a person, Korean brands cite a test
  Re-derived from collector run commerce-20260805025415 after a change to the instrument: Adds the Shopify return-policy family: four corridor claims on sealed-goods conditions, the 30-day window, transit-liability topics and policy length. The figures moved for that reason. Any part of the change attributable to the market moving cannot be separated from it in this revision, and the previous statement is recorded below rather than replaced.
  Before
  The two sides of this corridor borrow authority from different places: US-origin brands invoke a dermatologist on 8.4% of their products against 1.6% for Korean-origin brands, while 17 of 19 Korean-origin brands use clinical-testing language somewhere in the catalogue against 11 of 18 US-origin ones. One names a professional, the other names a procedure.
  After this change
  The two sides of this corridor borrow authority from different places: US-origin brands invoke a dermatologist on 8.4% of their products against 1.5% for Korean-origin brands, while 17 of 19 Korean-origin brands use clinical-testing language somewhere in the catalogue against 11 of 18 US-origin ones. One names a professional, the other names a procedure.
August 4, 202620 changes
1. v8corrected
  A third-party review platform is table stakes
  Re-derived from collector run commerce-20260804224019; figures moved.
  Before
  A named third-party review platform is table stakes on a US beauty storefront: 41 of 53 storefronts read carry one, and the 12 without are Isntree, Purito, SOME BY MI, Benton, Pyunkang Yul, VT Cosmetics, Banila Co, TOCOBO, CLIO, OLIVE YOUNG Global, Paula’s Choice, Topicals.
  After this change
  A named third-party review platform is table stakes on a US beauty storefront: 41 of 52 storefronts read carry one, and the 11 without are Isntree, Purito, SOME BY MI, Benton, Pyunkang Yul, VT Cosmetics, TOCOBO, CLIO, OLIVE YOUNG Global, Paula’s Choice, Topicals.
2. v6corrected
  A third-party review platform is table stakes
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  A named third-party review platform is table stakes on a US beauty storefront: 23 of 26 storefronts read carry one, and the 3 without are Isntree, OLIVE YOUNG Global, Paula’s Choice.
  After this change
  A named third-party review platform is table stakes on a US beauty storefront: 41 of 53 storefronts read carry one, and the 12 without are Isntree, Purito, SOME BY MI, Benton, Pyunkang Yul, VT Cosmetics, Banila Co, TOCOBO, CLIO, OLIVE YOUNG Global, Paula’s Choice, Topicals.
3. v3corrected
  Korean brands bring Channel Talk across the border
  Re-derived from collector run commerce-20260804224019; figures moved.
  Before
  Korean brands carry their home support stack into the United States: Channel Talk appears on 5 of 28 Korean-origin US storefronts and on none of the 23 US-origin storefronts read.
  After this change
  Korean brands carry their home support stack into the United States: Channel Talk appears on 5 of 27 Korean-origin US storefronts and on none of the 23 US-origin storefronts read.
4. v3corrected
  Localisation depth splits, it does not scale
  Re-derived from collector run commerce-20260804224019; figures moved.
  Before
  Localisation depth splits sharply rather than scaling: among the 17 US beauty storefronts that declare alternate locales at all, the median declares 7 of them, 5 declare twenty or more, and the deepest declares 189.
  After this change
  Localisation depth splits sharply rather than scaling: among the 16 US beauty storefronts that declare alternate locales at all, the median declares 8 of them, 5 declare twenty or more, and the deepest declares 189.
5. v3corrected
  Pinterest is a US-origin habit, not a requirement
  Re-derived from collector run commerce-20260804224019; figures moved.
  Before
  Pinterest is a US-origin habit rather than a US-market requirement: 12 of 23 US-origin beauty storefronts link to it against 3 of 28 Korean-origin storefronts selling into the same market, while 42 of 64 storefronts link to TikTok regardless of origin.
  After this change
  Pinterest is a US-origin habit rather than a US-market requirement: 12 of 23 US-origin beauty storefronts link to it against 3 of 27 Korean-origin storefronts selling into the same market, while 42 of 63 storefronts link to TikTok regardless of origin.
6. v3corrected
  Shopify runs the US storefront for Korean beauty
  Re-derived from collector run commerce-20260804224019; figures moved.
  Before
  Shopify is the default US storefront for Korean beauty brands: 18 of 28 Korean-origin storefronts selling into the United States run on it, against 12 on Cafe24, the Korean domestic default.
  After this change
  Shopify is the default US storefront for Korean beauty brands: 18 of 27 Korean-origin storefronts selling into the United States run on it. The same brands at home run the Korean default instead: 8 of 11 Korean-market storefronts read are on Cafe24.
7. v3corrected
  Subscription is not the default US beauty model
  Re-derived from collector run commerce-20260804224019; figures moved.
  Before
  Recurring purchase is not how US beauty direct-to-consumer is built: 10 of 53 storefronts run a subscription platform, and a further 10 mention subscribing in copy without one installed.
  After this change
  Recurring purchase is not how US beauty direct-to-consumer is built: 10 of 52 storefronts run a subscription platform, and a further 10 mention subscribing in copy without one installed.
8. v2corrected
  .99 pricing is a house style, not a market norm
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  Where .99 pricing does appear in US beauty direct-to-consumer it is a house style rather than a market convention: 86% of all .99 variants observed came from 2 of the 21 storefronts read, and 6 storefronts show any at all.
  After this change
  Where .99 pricing does appear in US beauty direct-to-consumer it is a house style rather than a market convention: 64% of all .99 variants observed came from 2 of the 39 storefronts read, and 9 storefronts show any at all.
9. v2corrected
  Charm pricing is not the default in US beauty
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  Charm pricing is not the default in US beauty direct-to-consumer: 15 of 21 storefronts price zero variants ending in .99, and only 4% of the 4,993 variants observed end in .99, against 81% ending in a round figure.
  After this change
  Charm pricing is not the default in US beauty direct-to-consumer: 30 of 39 storefronts price zero variants ending in .99, and only 3% of the 8,271 variants observed end in .99, against 82% ending in a round figure.
10. v2corrected
  Free shipping in US beauty clusters at $50
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  US beauty direct-to-consumer storefronts that publish a free shipping threshold cluster at a median of $60, across a range of $35 to $80 in a sample of 7.
  After this change
  US beauty direct-to-consumer storefronts that publish a free shipping threshold cluster at a median of $50, across a range of $35 to $100 in a sample of 14.
11. v2corrected
  Instalment payment is a minority option in US beauty
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  Instalment payment is a minority option rather than table stakes on a US beauty storefront: 10 of 24 storefronts that advertise their payment methods offer Klarna, Afterpay, Affirm or Sezzle, while every one of them offers PayPal.
  After this change
  Instalment payment is a minority option rather than table stakes on a US beauty storefront: 19 of 43 storefronts that advertise their payment methods offer Klarna, Afterpay, Affirm or Sezzle, against 42 that offer PayPal.
12. v2corrected
  Korean beauty prices 29% below US-origin peers
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  Korean beauty brands selling into the United States sit below their US-origin peers on price: a median storefront price of $24 across 13 Korean-origin storefronts against $29 across 7 US-origin storefronts, a gap of 17%.
  After this change
  Korean beauty brands selling into the United States sit below their US-origin peers on price: a median storefront price of $22.5 across 18 Korean-origin storefronts against $31 across 20 US-origin storefronts, a gap of 27%.
13. v2corrected
  Korean brands bring Channel Talk across the border
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  Korean brands carry their home support stack into the United States: Channel Talk appears on 3 of 15 Korean-origin US storefronts and on none of the 9 US-origin storefronts read.
  After this change
  Korean brands carry their home support stack into the United States: Channel Talk appears on 5 of 28 Korean-origin US storefronts and on none of the 23 US-origin storefronts read.
14. v2corrected
  Localisation depth splits, it does not scale
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  Localisation depth splits sharply rather than scaling: among the 9 US beauty storefronts that declare alternate locales at all, the median declares 9 of them, 3 declare twenty or more, and the deepest declares 189.
  After this change
  Localisation depth splits sharply rather than scaling: among the 17 US beauty storefronts that declare alternate locales at all, the median declares 7 of them, 5 declare twenty or more, and the deepest declares 189.
15. v2corrected
  No review platform is universal in US beauty
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  No single review platform owns US beauty: the most common, Yotpo, appears on 12 of the 23 storefronts that carry reviews at all, 6 distinct platforms are in use across the set, and 8 storefronts run more than one at once.
  After this change
  No single review platform owns US beauty: the most common, Yotpo, appears on 20 of the 41 storefronts that carry reviews at all, 7 distinct platforms are in use across the set, and 13 storefronts run more than one at once.
16. v2corrected
  Pinterest is a US-origin habit, not a requirement
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  Pinterest is a US-origin habit rather than a US-market requirement: 6 of 9 US-origin beauty storefronts link to it against 2 of 15 Korean-origin storefronts selling into the same market, while 23 of 30 storefronts link to TikTok regardless of origin.
  After this change
  Pinterest is a US-origin habit rather than a US-market requirement: 12 of 23 US-origin beauty storefronts link to it against 3 of 28 Korean-origin storefronts selling into the same market, while 42 of 64 storefronts link to TikTok regardless of origin.
17. v2corrected
  Sets and kits are 23% of the median catalogue
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  Sets and kits are a minority of the US beauty catalogue, not its spine: the median storefront names a set, kit, bundle or routine in 19% of product titles across 21 catalogues, with a maximum of 48%.
  After this change
  Sets and kits are a minority of the US beauty catalogue, not its spine: the median storefront names a set, kit, bundle or routine in 22% of product titles across 39 catalogues, with a maximum of 49%.
18. v2corrected
  Shopify runs the US storefront for Korean beauty
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  Shopify is the default US storefront for Korean beauty brands: 13 of 15 Korean-origin storefronts selling into the United States run on it, against 2 on Cafe24, the Korean domestic default.
  After this change
  Shopify is the default US storefront for Korean beauty brands: 18 of 28 Korean-origin storefronts selling into the United States run on it, against 12 on Cafe24, the Korean domestic default.
19. v2corrected
  Subscription is not the default US beauty model
  Re-derived from collector run commerce-20260804223321; figures moved.
  Before
  Recurring purchase is not how US beauty direct-to-consumer is built: 4 of 26 storefronts run a subscription platform, and a further 7 mention subscribing in copy without one installed.
  After this change
  Recurring purchase is not how US beauty direct-to-consumer is built: 10 of 53 storefronts run a subscription platform, and a further 10 mention subscribing in copy without one installed.
20. v2retiredretracted
  The K-beauty price gap sits in the tails
  Retired against collector run commerce-20260804223321: quartiles are 30% and 16% apart, the bands do not overlap. The claim held on the sample available when it was written and does not hold on this one.
  Before
  The price gap between Korean-origin and US-origin beauty brands selling into the United States is not a gap in the working range: the middle half of the ladder is effectively the same, $19 to $38.49 for Korean-origin storefronts against $20 to $38 for US-origin ones, while the median top-of-catalogue price is $131.20 against $250.
  After this change

6 entries are not in this list

For two days in August 2026 the derivation compared a statement built with a straight apostrophe against the same statement stored with a curly one. Any claim naming a brand like Paula’s Choice therefore looked changed on every run, and wrote a revision saying its figures had moved when nothing had. Six revisions on one claim were that bug and no change of ours.

They are still on the claim, because this log is not edited, and the revision directly beneath them says which ones and why: a third-party review platform is table stakes. They are kept off this page because a list of corrections we did not make is not evidence of anything.

The same log is machine-readable in entries.json, under revisions on each claim. Every entry carries the statement as it stood at that version, so any past version of any claim can be retrieved from the file. Withdrawn claims move to retired.json rather than disappearing, so a script holding an old id can tell a retraction from a deletion. Found something we got wrong? hello@tryorbt.com, and the correction goes on this page with your name on it if you want it there.

How the Orbt Market Index is made

August 9, 20261 change

August 5, 2026152 changes