How PulseOrbit scores trends
By PulseOrbit. Published .
PulseOrbit is an AI trend radar that spots rising trends early and generates content, product, and niche ideas for creators and sellers. Velocity Score weights search interest at 40%, content volume at 30%, marketplace mentions at 20%, and regional breadth at 10%. The worked example scores 45. Board figures are dated stored snapshots, not a live feed.
Stored snapshots, not a live feed. Figures come from named public sources, Wikimedia pageviews and the GDELT Project, plus YouTube and Google Trends through DataForSEO where shown. Each figure carries its own observation date. Scheduled jobs refresh the stored snapshot when a source responds. If a source does not respond, the older figure stays on the page with its older date and is marked stale. PulseOrbit does not promise accuracy, search rankings, or AI citations. The worked example below stays a worked example of the method. It is not a row on the board.
What do the weights mean?
PulseOrbit is an AI trend radar that spots rising trends early and generates content, product, and niche ideas for creators and sellers. Each 28-day percent change is capped at 100 before weighting. A decline scores 0 on that part in this example method, because the score is built to describe a rise. The mainstream threshold is a search interest index of 25 on a 0–100 scale.
| Component | Weight | What the example measures |
|---|---|---|
| Search interest change | 40% | 28-day search interest change |
| Content volume change | 30% | 28-day content volume change |
| Marketplace mention change | 20% | 28-day marketplace mention change |
| Regional breadth | 10% | which tracked regions are rising |
What does the worked example add up to?
The worked example Velocity Score is 45. Early Signal % is 50. How early is 6 weeks. The snapshot date used here is 2 November 2026. This worked example is not a row on the Rising Now board.
| Input | Example value | Weight | Contribution |
|---|---|---|---|
| Search interest change (28 days) | +60% | 40% | 24 |
| Content volume change (28 days) | +40% | 30% | 12 |
| Marketplace mention change (28 days) | +20% | 20% | 4 |
| Regional breadth | 3 of 6 regions | 10% | 5 |
| Velocity Score | Rounded | 45 |
Parts are shown to one decimal place. The Velocity Score is the sum rounded to the nearest whole number.
How does a stored snapshot fill the same weights?
The worked example above stays an example. A stored snapshot uses the same four weights and the same rounding function, with the measurements below. It is not a live feed. Refresh the snapshot file with pnpm ingest, then rebuild. YouTube view velocity is a separate figure, not a fifth weight. A Vercel cron refreshes it once a day when YOUTUBE_API_KEY is set. This deployment was built with the live data mode, so the public board shows the stored snapshot, with each figure dated by its source. A preview of the file is at /preview.
| Weight | Stored snapshot input |
|---|---|
| 40% attention | 28-day percent change in Wikimedia pageviews for one en.wikipedia article (all-access, user agent). en.wikipedia is a global proxy. It is not a United Kingdom or United States search index, and it is not Google Trends. A prior window under 500 views is shown and not scored. |
| 30% news volume | 28-day percent change in the global GDELT DOC 2.0 timeline for the topic query. Source: GDELT Project. The file records whether the series is raw article counts or volume intensity. A raw prior window under 10 articles, or an intensity prior under 0.05, is shown and not scored. |
| 20% marketplace | Not measured. The input is 0. Etsy and Amazon KDP stay idea text. PulseOrbit does not read a marketplace. |
| 10% breadth | Share of three GDELT series that rose: global, source country US, and source country UK. That is 3 series, not the 6 example countries in the worked example. |
Early Signal on a snapshot row uses the same formula as the worked example. The points come from a pageview index scaled so the busiest week in the 12-week window is 100. The mainstream line is still 25, the same number as the worked example's index of 25. It is not a Google Trends index. How early is the number of weeks from the first week of that window until the scaled index reaches 25. If it never reaches the line, or it starts above the line, how early is 0.
The Wikimedia top pageviews list for the snapshot date is checked. A rank is shown when the article is in the en.wikipedia top 1000 that day. Most niche articles are outside that list. The rank is not part of the Velocity Score. Pageviews data is CC0; the page still credits Wikimedia pageviews. GDELT terms allow commercial use with citation, so the page says Source: GDELT Project and links to the project.
If a fetch fails, ingest keeps the previous snapshot file. The build reads that file and does not call the network. A missing or unreadable file leaves the live board empty and shows an error instead of crashing the build. YouTube is skipped the same way when YOUTUBE_API_KEY is missing, including during the build. One daily refresh searches each tracked topic for region codes GB and US. search.list costs 100 units, so that plan is 3,210 units of the 10,000 unit daily quota. The cron stores that reading and the preview reads it. A page view does not call YouTube. The reading is discarded after 30 days. The page says Data from YouTube and links each video.
Google Trends search interest is a separate labelled figure for the United Kingdom and the United States. It is not a fifth weight, and it does not replace Wikimedia pageviews. The figure is Google Trends via DataForSEO, from the Keywords Data API standard Explore queue. It is not a first-party Google Trends API. Each task sends up to 5 keywords for one location. This list has 16 topics, which is 8 tasks a day. DataForSEO lists that standard task at $0.0027, so a daily run is about $0.0216. The cron is /api/cron/dataforseo at 06:45 UTC, after the YouTube cron. It posts tasks and collects any tasks still open from the previous run. The standard queue can take up to 45 minutes. The cron stores the reading, and the preview reads that store. A page view does not call DataForSEO. A reading is kept until the next successful run and dropped after 30 days. A failed fetch leaves the previous reading in place. If DATAFORSEO_LOGIN or DATAFORSEO_PASSWORD is missing, including during the build, the figure is skipped and Wikimedia, GDELT, and YouTube still render.
How to read a PulseOrbit velocity score
- Read the snapshot date first. This worked example uses the snapshot dated 2 November 2026 and was last updated 5 November 2026.
- Cap each 28-day percent change at 100. In the worked example, search interest change is 60, content volume change is 40, and marketplace mention change is 20.
- Apply the weights: search interest 40%, content volume 30%, marketplace mentions 20%, and regional breadth 10%. 3 of 6 tracked regions are rising, which is a breadth component of 50.
- Add the weighted parts and round to the nearest whole number. The worked example parts are 24, 12, 4, and 5, so the Velocity Score is 45.
- Read Early Signal % from the search-interest points gained before the index crossed 25. The worked example gained 30 points and 15 of them came first, so Early Signal % is 50.
- Read how early as the week gap from the first early signal to the week the index crossed 25. The worked example is 6 weeks.
Questions about the scoring method
How is the Velocity Score calculated?
Each 28-day percent change is capped at 100. Search interest is weighted 40%, content volume 30%, marketplace mentions 20%, and the share of 6 tracked regions that are rising is weighted 10%. The four parts are added and rounded to the nearest whole number.
What is Early Signal %?
Early Signal % is the share of the example search-interest point gain that happened before the index crossed 25. In the worked example, 15 of 30 points came first, so Early Signal % is 50.
What does how early mean?
How early is the number of weeks between the first week the example signal fired while the index was still under 25 and the week it crossed 25. The worked example is 6 weeks.
Which sources does this version use?
This deployment scores stored snapshots from Wikimedia pageviews (en.wikipedia, all-access, user agent) and the GDELT Project news volume, each with its own observation date. YouTube and Google Trends through DataForSEO are shown separately where available. None of these is a live feed.
Browse the stored snapshot boardHow the score is explained in the notes