Every number shows where it came from
Every figure in Dorqa has a named source, a stated time window and one formula on every screen. When we do not have a figure, we say so. When a figure is old, we date it.
Last updated 21 September 2026
Ten rules every screen follows
One number, one meaning
A metric is calculated once and shown the same on every screen. The Trend Score on Trend Engine, Trend Explorer and a topic's own page comes from one shared formula. Different measures get different names: the Trend Score measures momentum, the Decision Score measures the opportunity to act.
Every count has a window
A count means nothing without its time frame, so each one carries it: last 7 days, last 14 days, or lifetime. Search Console reports with a delay of a few days, so Organic Performance states the exact dates it covers and the period it compares against.
Missing data reads no data, never zero
A zero reads as evidence that something does not exist. Across the tools we have audited (Search Console, Analytics, Organic Performance, Site Audit, Domain Overview, Trend and Audience Intelligence), a source that returns nothing shows no data or a dash.
A failed lookup says it failed
When a request times out, the section says it could not load the data and offers a retry. It does not show an empty list in its place, because an empty list reads as an all-clear.
Old data is dated
Search volume and year-over-year figures show when they were fetched. Anything older than 30 days is greyed out, with the date one hover away.
Small movements are not findings
A change counts only when it is bigger than normal variation for its volume, roughly twice the square root of the combined count. A page going from 35 to 42 clicks is ordinary wobble, so Dorqa does not flag it.
Comparisons must be like for like
If Analytics recorded far fewer sessions per search click in one period than the other, tracking changed between them. Dorqa then marks the comparison as not comparable instead of reporting growth that measures the tracking fix.
Partial scores say so
A score built from several parts is scaled to the parts actually measured, and names the ones it could not include. With too little data, there is no score at all. Missing data never reads as a weaker site.
No recommendation from missing data
If Dorqa did not measure something, it does not score it as an opening. A topic with heavy publishing activity is never labelled an early-mover opportunity, however fast it is growing.
Show the working
Scores print their components and weights, and rates print their arithmetic, so you can check the numbers yourself.
How trends are scored
Signals come from named public platforms and publication feeds. Every score prints its parts.
- Sources. GitHub, Hacker News, arXiv, HuggingFace and Product Hunt for developer, research and launch activity; search demand data; and dozens of industry publication feeds across marketing, finance, health, retail, real estate, HR, education, travel, supply chain and sustainability.
- Publishers are counted by owner. Ten titles from one media group are one newsroom, not ten confirmations.
- Platforms, not labels. Some platforms reach us through more than one collector. Each platform counts once, so a topic cannot look cross-confirmed because two collectors saw the same place.
- How the scores are built. Velocity (0 to 10) weights developer activity 35%, community 30%, research 20%, market 10% and search 5%. The Trend Score (0 to 100) weights velocity 40%, signal momentum 25%, source diversity 20% and acceleration 15%. When last week had fewer than 3 signals, acceleration shows New instead of a percentage.
- Stages need more than one source. Exploding, Emerging and Early each require separate platforms for tech topics, or separate publishers for business topics. Cooling means a topic was busy in the last 30 days and attention is now dropping fast.
- Where the coverage comes from. Every publication feed is tagged India, US, UK or global. Each business topic shows its mix for the last 14 days, and signals with no country are shown as unclassified rather than dropped.
Search Console, Analytics, Site Audit and Domain Overview
Your own Google data is the source of truth. Everything modelled is labelled as modelled.
- Headline totals match Search Console. Clicks and impressions come from Google's property totals. Query-level sections can only use the queries Google reports, so they state how much of your clicks they cover. On one site we checked, that was 58%.
- Labels describe what was measured. A page whose clicks fell while its ranking held is labelled Fewer clicks at same rank. Dorqa does not guess the cause.
- Estimated traffic is an estimate. Traffic potential on Domain Overview is ranking positions times search volume times a standard click-through curve. It does not discount for AI Overviews, so treat it as an upper bound.
- A difficulty of 0 is not easy. The difficulty formula clips the bottom fifth of its scale to exactly 0, and many keywords land there. Dorqa shows a dash instead of calling them easy.
- Unscored is not slow. When a page speed test cannot score a page, Site Audit shows a dash, not a red 0.
What the audience numbers mean
- Behaviour, not a census. Audience Intelligence measures what people search for, discuss, read and watch. Figures describe search interest and published content, not a headcount.
- Searchers are not always the audience. For an occupation such as dentists, most searches come from people looking to hire one. Dorqa labels those panels as searchers.
- Demographics are relative. Age and gender are relative search interest scaled to 100, not percentages. Over-index comparisons appear only where a baseline exists: age for India, the US and the UK, gender for the US and the UK.
- Community growth has a floor. Member growth appears only when our readings of a community are at least 28 days apart and it had at least 5,000 members. Every figure carries its window.
- Coverage is not confidence. The Coverage Score counts which sections returned data. It says nothing about whether the data is right, and is named accordingly.
- Exports keep the caveats. Copy for AI carries a caveat under every block that could be misread, and repeats its reading rules at the end. A machine-written summary was switched off after it described things the data did not support.
What Dorqa cannot tell you
We would rather you hear these from us.
- Trend detection is global. Choosing a country changes search volume and video results, not which trends are detected.
- Coverage is deeper for tech than for other industries. Industry coverage relies mainly on publication feeds.
- Publication coverage leans US. As of September 2026, 55 publication feeds are active: 26 US, 16 global, 12 Indian and 1 British. By owner that is 14 US, 15 global, 7 Indian and 1 British. Each topic's country mix is on its page.
- Product launch coverage is partial. Launches are matched to topics by their wording and some are missed, so a low launch count is never read as proof a market is open.
- Search Console lags. Google reports with a delay of a few days, and withholds rare and private queries from query-level reports.
- Very large properties can time out. On long date ranges for the largest Search Console properties, a lookup can occasionally time out. The page says so and offers a retry.
- Merged topics keep their history separate. When duplicate topics are merged, past mentions stay with the archived duplicate and the kept topic collects new ones from the merge date.
- Audience research covers India, the US and the UK. Other markets are not yet supported.
About the data
No. Dorqa measures behaviour: what people search for, discuss and publish, plus your own Search Console and Analytics data. Audience figures are relative search interest, not headcounts, and the page says so beside every panel that could be misread.
Some figures are modelled rather than measured, and every provider models them differently. Estimated organic traffic is ranking positions times search volume times a click-through curve, and keyword difficulty uses each provider's own formula. Dorqa labels modelled figures as estimates and treats estimated traffic as an upper bound.
The section says it could not load the data and offers a retry. It never shows an empty result in its place, because an empty result reads as an all-clear. Scores built from several parts name any part that could not be measured.
Email support@dorqa.io with the page and the number. Every figure traces back to its source, and we will show you where it came from.
Be first. Not just found.
Free tier, no card. Check the numbers yourself.