Methodology
How the ranking is calculated
KoReADAR measures an increase in attention relative to a topic’s usual level, together with how long it lasts. The goal is to distinguish a brief spike from sustained attention across sources.
1. Put different sources on a comparable scale
We monitor public signals from Google Trends, Signal.bz, news and entertainment rankings on Naver, Daum and Nate, communities including DCInside, TheQoo, Nate Pann, Instiz and FMKorea, and X, TikTok and YouTube. Collection schedules and access vary; every source is not necessarily present in every update. A source that fails to collect is left out of that update without raising the weight of the others, and the page shows how many sources were collected normally.
Most inputs are popularity lists already selected by each service. KoReADAR does not see raw search logs or user counts, so the ranking describes rising attention captured on major public Korean platforms, not a definitive measure of the whole country’s interest. Since 2025 YouTube’s popular list mainly reflects music, film and gaming charts, so it does not represent video attention as a whole.
For each source, we calculate a topic’s share of attention and compare it with its usual share on that source. Views, likes and list positions are different measures, so raw values are not added across platforms. Sources with too little history receive less weight.
2. Group related wording into topics
Keywords and related expressions from titles are grouped into topics. Posts that share only a name, or only common words such as “rookie” and “debut”, are not treated as the same event. Automatic grouping can still merge different events or split one event into several topics. Topics containing only generic words without a specific subject are excluded from the ranking.
3. Combine persistence, corroboration and freshness
The ranking is normally calculated at 15-minute intervals. Each calculation averages attention strength over the last two hours, with recent observations receiving more weight. Under the current settings, an observation one hour old has half the weight of the current observation. The ranking uses this score accumulated over calculations. The accumulated score halves every 90 minutes, so topics that draw attention steadily for hours stay near the top, and topics whose attention fades drop within a few hours. Topics no source lists now are not ranked.
Score of each calculation = 100 × sustained attention × evidence strength × freshness × trend factor × campaign factor
- Sustained attention: source-weighted attention above usual levels, averaged over time.
- Evidence strength: an empirical value combining support from search, news, entertainment, community, social and video sources with persistence. One article listed in several portal rankings counts as one confirmation. It is not the probability that a topic is nationwide or that a story is true.
- Freshness: the score halves one day after a surge began and quarters after two days.
- Trend factor: changes from an hour earlier adjust the score by up to about 15%.
- Campaign factor: certain fan campaign patterns seen only on social platforms receive less weight.
New spikes without an established history of persistence may appear under “Rising now.” Scores express relative attention; they cannot be converted to search volume or numbers of people.
Read the numbers on a card
The chart runs from past on the left to present on the right. Source bars show the strength of attention on each source. News, search and community percentages describe the relative contribution of three display groups, not population shares. Entertainment is displayed with news; social and video signals are displayed with communities. Rank arrows compare with the previous ranking.
Limitations and verification
Baselines use up to seven days of recent history, which may not adequately distinguish weekly patterns or longer seasonal effects. Grouping, translation and previews can contain errors. Verify important claims against the originals and other sources. Titles, images and excerpts belong to their respective rights holders.
Last revised: