How this board is built

Didn't Naver kill real-time search rankings?

Correct — Naver retired its real-time rankings in 2021, Daum in 2020. So this board doesn't copy any of those lists. It's built from signals that are still public: Google Trends rising queries, Daum's algorithmic realtime issues, Nate's public issue keywords, ZUM's issue trend, YouTube trending videos, and Naver DataLab search indices.

We have no affiliation, sponsorship, or data-supply relationship with the companies listed as sources. We reference rankings and indices those sites publish openly, and we don't use their logos or trademarks. If a site stops publishing, we drop that source immediately.

If one source stalls, the board still builds from the rest. You can see which sources are live on the status page. Status →

What the KTB Index means

We call this number the KTB Index. When citing, write it as "KTB Index 45.9" so it's clear which measure you mean. The formula is published in full below.

A relative value between 0 and 100 — not an absolute search volume. Each source uses a different unit (search index, view count, rank), and none of them publish raw volume. So we normalize within each source to 0–1, then combine with weights.

We use a log scale, because view counts have a long tail — the gap between #1 and #2 dwarfs the gap between #2 and #10. Raw values would let a single item flatten everything else.

So 45.9 vs 40.6 doesn't mean "13% more searches" — it means "picked up higher and more consistently across sources".

Shopping trends and the K-series (Food, Place, Beauty, Fashion, Wellness, Pop, Drama, Movie) use a different calculation — official NAVER DataLab API values (shopping insight, search trends), shown relative to an anchor keyword. They don't mix with the board score.

• AnchorOne strong keyword per category is fixed at 100 so separate batches stay comparable
• Current weekThe latest week is dropped — an unfinished week holds only a few days and always reads low
• GrowthThe last three weeks against the weeks before. Jumps off a near-zero base are discarded
• K-BrandThis one uses shopping-click change, not search volume — brands measured by search volume only show size, and the ranking never moves. The click index is relative within its category, not actual sales
• Early-warning levelsWe only watch keywords climbing from ranks 11-20 — measured by band, calls at ranks 21-25 hit just 1 in 11, and that band reaches the top 10 within six hours only 4.6% of the time (2026-09-11). Score adds up rate of rise, acceleration, source count and current rank to sort into imminent / likely / watch, and every call is graded against reality six hours later
• Trending nowYouTube trending music, cross-read against our search index. YouTube ranks partly by cumulative views, so we re-measure by views per day — and where plays and searches disagree we say why: high plays with low searches means people are listening without looking it up; the reverse means it is talked about more than played

Which signals we trust, and how much

The exact weights are internal tuning values we don't publish. What we do publish is the ordering and the reasoning — trust comes from the rationale, not the decimals.

• MostPortal-curated issues (Google Trends · Daum Realtime)
• ThenYouTube Trending — real attention, but shaped by the algorithm
• LessNaver DataLab — a daily index, not real-time
• LeastCommunities (DCInside etc.) — heavily skewed samples, used only as a corrective

Community signals are capped so the board reflects Korea, not just its forums. If a source pauses, the board keeps building from the rest — see the status page for what's live.

Cross-source bonus

A keyword picked up by multiple sources earns a bonus that grows with each extra source. A spike in only one place is usually that platform's own dynamics — not real attention.

Click any keyword to see which sources picked it up.

Refresh and change

We re-collect every source every five minutes. The ▲▼ figure is the score change versus one hour ago, not the rank change — near the top, a big score jump often moves the rank not at all.

What you can do here

• One-line takesLeave one line on any keyword page. Takes from everywhere gather in one place, and other languages are auto-translated — with the original one tap away. Signed in, once a day, 140 characters.
• Keyword pagesTap any keyword for one page with its first appearance, best rank, the headline behind it (linked to the source), related coverage, a video, and what rose alongside it.
• GlossaryOnly words people actually search the meaning of — chosen by measuring search volume, and published only after a human checks the definition.
• Archive & daily briefingOpen any past day's board as it stood, with a briefing written strictly from that day's measured data.
• K-seriesRankings by axis — drama, film, webtoon, actors, beauty, fashion, food, places. Tap an item for photos, video, an explainer, and coverage.
• Feed & APISubscribe via RSS (/rss.xml), or pull ranks and history yourself with a free API key.

Takes are hidden automatically once reports accumulate, and we don't open the comment box on sensitive keywords.

What we don't do

• Generated summariesNo — the "why" is an actual headline, quoted as-is
• Paid placementNo — you cannot buy a spot on this board
• Rewriting historyNo — past snapshots stay as recorded

We do merge split spellings (NCT 127 / 엔시티 127) after human review, and we can block abusive terms. Neither ever raises or lowers a score arbitrarily.

How the score is computed

Each source measures something different — views on YouTube, ranks on portals. They can't be added as-is, so we normalize within each source to 0-1, then weight and sum.

unit    = (log10(raw + 1) - min) / (max - min)   # 출처 안에서
total   = Σ ( unit × weight[source] )
score   = total / max(total) × 100              # 1위를 100 으로

Why the log: view counts have a long tail. A million versus a hundred thousand is not ten times more meaningful than a thousand versus a hundred — used raw, one big number would swallow the board.

Source weights

Weights can change; when they do, it goes in the change history above. Changing quietly would mean yesterday's number and today's were measured with different rulers.

Change history

When a rule changes, we record it here. Changing quietly and mixing old numbers with new ones would be deceptive.

Corrections

When we fix a value we collected wrong, we record it here. Changing a number quietly would make yesterday's citation wrong through no fault of the person who made it.

Nothing to correct so far. An empty list is the normal state — the commitment stands either way.