This page records changes to the figures and claims Parity Vision publishes. We keep our figures current as we confirm new information, and note here what we changed, when and why. The log starts on 13 November 2026; newest first. 한국어
- 14 November 2026 · Korean newsroom post 5 (method and correction note), /en/findings (correction note), /en/checklist (source line)
Previously: post 5's method said a person checked the flagged spots on 3 November "(and on 13 November for the 58 filings added later)", and the correction notes said a person confirmed the three added differences, with no date → Now: the three differences from the 58 added filings were checked on 14 November, and a person confirmed the three differences against the originals on 14 November. The /en/checklist line "checked by a person on 3 and 4 November" now adds 14 November
Why: when we issued the 13 November correction, they had been compared with the originals using an AI tool, but no person had checked them yet. A person opened the Korean and English originals on 14 November.
Checked: on 14 November, against the record of those three checks against the originals. - 14 November 2026 · /en/findings, /en/checklist, Korean newsroom post 5 summary
Previously: re-adding the English tables would have caught five of the six differences in our first sample, and "in every case, the Korean table added up and the English one did not" → Now: re-adding catches the four digit differences, the unit label and the column placement need a side-by-side look at the original, and "in the first sample's four digit differences, the Korean table added up and the English one did not". Post 5's Korean summary no longer says re-adding reveals many year, unit-label and column-placement differences; it says a changed or missing digit often breaks the English total, while years, unit labels and column placement need the original
Why: the six differences in the first sample (235 disclosures) were four digit differences, one unit label and one column placement. Re-adding only reveals the digit differences.
Checked: on 14 November we compared the first sample's record of checks against the originals (3 November) with our list of differences, then corrected. The LinkedIn article with the same sentence was corrected the same day. - 14 November 2026 · /en/findings
Previously: we did not say which English filings we counted → Now: we say we counted only the English filings the companies made themselves, not the Korea Exchange's translations that English KIND also carries. The new /en/faq, made the same day, says so from the start
Why: English KIND carries the Korea Exchange's translations of some Korean filings as well as the companies' own English filings. Our counts cover only the companies' own.
Checked: on 14 November we used an AI tool to open exchange-translated corrections on English KIND and confirmed in our data that every English filing in the large and mid-sized samples was the company's own, then corrected. - 14 November 2026 · /en/findings title and layout
Previously: titled "What we found comparing 235 Korean and English disclosures", with the 235-disclosure sample first → Now: titled "How accurate are Korean companies' English filings? We checked 1,211.", with the 1,211-filing results first and the 235 disclosures later as our first sample. The address did not change
Why: so the opening paragraph directly answers how accurate the English filings are. No figures changed.
Checked: on 14 November, after our figures check and three rounds of review by two AI reviewers. - 14 November 2026 · Korean newsroom posts 5 and 9, Korean Q&A, /en/findings, /en/checklist
Previously: "13 figures" in English; in Korean, "13 numeric differences" (the Q&A said "13 figures"), a post 9 table heading "numeric differences confirmed against the originals", and a post 9 correction note "6 filings with 10 numeric differences (about 0.5%)" → Now: "13 confirmed differences" in English; in Korean, "13 differences", "differences confirmed against the originals" and "6 filings (about 0.5%) with 10 differences", with "about 0.7%" attached to the 9 filings. Korean post 5 now says the differences "mostly" make amounts read several times larger or smaller, or change share counts, and the Korean Q&A says they are "mostly" in the digits of amounts or in share counts
Why: two of the 13 (one unit label, one column placement) are not differences in a figure's value.
Checked: on 14 November we re-counted the types in our list of 13 differences, then corrected. - 13 November 2026 · /en/timeliness
Previously: large companies, next business day 15.2%, and the correction note said only that we had added 17 July to our holiday list → Now: 15.3%; we also added 1 May, and the note says that on both days there was no Korean exchange filing among the 270 companies and no English one among the 269 whose English filing lists we collected
Why: 1 May was missing from our holiday list. Adding it moved one large-company filing from two business days to one. We collected English filing lists for 269 companies; for one company we could not find a list on English KIND.
Checked: on 13 November we fixed the holiday list and recalculated, and that evening counted filings on both days in the Korean lists (270 companies) and the English lists (269), then corrected. - 13 November 2026 · Home comparison speed (Korean and English), /en/findings summary
Previously: 6,870 Korean figures, 2 seconds, at most 151 figures in one filing (measured 3 November) → Now: 6,867 figures, 1.2 seconds, 141 (measured 13 November). The /en/findings summary now gives the re-counted 6,867 next to the first sample's 6,870
Why: we measured the same 235 disclosures again with the 13 November engine.
Checked: against the 13 November measurement record, then corrected. - 13 November 2026 · Korean newsroom posts 5 and 9, /en/findings, /en/checklist, /en/timeliness, and other pages that showed this figure (such as the home page news list)
Previously: large sample of 1,153 filings by 28 companies, "10 figures in 6 of 1,153 filings" (about 0.5%) → Now: 1,211 filings by 29 companies, 9 filings (about 0.7%) with 13 confirmed differences
Why: an error in our list of one company's English filings had left 58 of its filings out.
Checked: on 13 November we added the 58 filings and ran the comparison again, then corrected. A person confirmed the three new differences against the Korean and English originals on 14 November (top entry). - 13 November 2026 · /en/timeliness, Korean newsroom post 9
Previously: mid-sized sample of 310 filings by 30 companies: 43.5% same day, 29.0% next business day, 25.5% two to three, 1.9% four or more; large companies: 76.5% same day, 15.8% next business day, 7.3% two to three; related-filing lists shorter in 9.8% and 13.7%; other ratios from the same samples (such as the large-company same-day share from May) were recalculated too → Now: 1,598 filings by 150 companies with end-2025 consolidated assets of KRW 2 to 10 trillion: 44.6%, 26.8%, 27.3%, 1.3%; large companies: 77.5%, 15.2% (corrected again to 15.3% the same day, entry above), 6.9%; 10.0% and 13.2%
Why: we could not confirm that the first mid-sized sample (30 companies) stayed within KRW 2 to 10 trillion in assets. The large sample gained the 58 filings above, and we added 17 July to our holiday list. The main finding did not change: at mid-sized companies, more than half of English filings arrived a day or more after the Korean original.
Checked: on 13 November we selected every company in range using end-2025 consolidated assets and measured again, then corrected.
We keep the correction notes on each page. When we correct a LinkedIn or Remember post, we leave a comment on that post saying what changed, when and why.