See who they are, and where they actually cluster
Raw audience counts mostly rediscover where the population is. Dorqa reports concentration per capita, marks how far each segment over-indexes, and says which of its eight signal components returned data and which returned none.
Every number on this page came from a real run.
Concentration and size are different questions, and so are a signal that was measured and one that was inferred. The product labels both rather than presenting a single blended read.
Age bands, reported as concentration
Each age band is shown with its concentration score and how far it over-indexes against the base population. That second number is the one that tells you something, and it is printed rather than left for you to work out.
- Concentration score per band, not a raw headcount
- Over-index multiplier shown alongside every band
- Baseline stated so the ratio means something
- Source labelled on the panel itself
On a real run for yoga teacher in the United States, the 25-34 band scores 70 and the 18-24 band scores 69. The younger band over-indexes at 1.3x; the higher-scoring one sits below baseline at 0.9x.
The raw score and the multiplier point in opposite directions. Read only the score and you pick the wrong band.
94/100
with all seven returning dataSeven components, each one named
The confidence figure is not a mood. It is built from seven named components, each labelled by the kind of signal it is, and any that returns nothing is shown as empty rather than quietly excluded from the total.
- Seven components listed by name on the panel
- Four read off audience activity, two mixed, one a source list
- Components with no data are displayed, not dropped
- So a high score and a complete score are distinguishable
Named accounts, and the share that follows each
For an occupational audience the module builds a follow graph: it finds people who describe themselves as that job on X or Instagram, reads who they follow, and ranks the accounts by the share of that set following each one. Two ways to use it — target the followers of a large account, or approach the account itself.
- Accounts ranked by share of the practitioner set, not by follower count
- Accounts that appear across unrelated audiences are filtered out
- When no graph clears the bar, the named practitioners are shown instead
- Labelled as what it is: a list of practitioners is not a follow graph
On the same run, Instagram returned no graph. The card says why: only 12 accounts describe themselves that way, too few to find what they follow in common — and that no credits were charged. The twelve are still listed, under a heading saying they are practitioners rather than a follow graph.
An empty card would have read as a broken tool. A named reason reads as a measurement.
- Audience snapshotA written read, with the signal count it drew on
- Who they areAge and regional concentration, with the over-index
- Where this topic surfacesPlatforms ranked by appearance, labelled as such
- Search intelligenceWhat this audience searches around the topic
- Active conversationsThreads, videos and titles, per platform
- Who they followX and Instagram accounts, by share of the practitioner set
- Where they discussCommunities, with member counts
- Language they useThe audience's own phrasing, verbatim
- Trusted sourcesSites, channels, podcasts and newsletters
- Confidence scoreWhich components returned data, and which did not
Where the conversation is, and how it sounds
Communities are found by searching what practitioners write rather than the name of the job, which is the difference between an occupation's trade subreddits and a list of the cities its customers live in.
- Subreddits carry member counts, not an invented activity grade
- Verbatim phrasing pulled from real threads
- On an optometrist run: chart from home, diplopia, patient recall
- Useful input for briefs and for prompt selection
What this actually means
Over-indexing is how much more concentrated a topic's audience is within a segment than that segment's share of the base population would predict. A 1.3x figure means the interest is thirty per cent denser there than baseline. It is a ratio, so a small segment can over-index heavily while contributing few people in total.
It is not audience size
Sorting regions by raw audience mostly reproduces a population ranking, which you already knew. Over-indexing removes population from the answer and leaves the part that is actually about your topic.
Search behaviour is behaviour
What this audience typed into Google, wrote on Reddit, bought on Amazon and watched on YouTube are things they did. The confidence panel separates signals about people from signals about material — a list of trusted sites is a different kind of evidence from a thread someone wrote — and names any component that returned nothing at all. Where a panel ranks platforms by how often a topic appears rather than by activity, it says so on the panel.
The audience, or the people looking for one
Search volume for an occupation is dominated by the people hiring one, not the people doing it. Dorqa detects which side a run is describing and labels every panel accordingly — the demographic cards read “age of searchers” rather than “age of this audience” when the evidence is customer-side, and the ambiguous case resolves to the more cautious label rather than the flattering one.
One of six layers
They share one project, one subscription and one set of credits. See what Dorqa replaces.
Before you start
That interest in the topic is thirty per cent denser in that segment than the segment's share of the base population would predict. 1.0x is exactly baseline.
Because sorting by size mostly reproduces a population ranking. Removing population from the calculation leaves the part of the answer that is about your topic.
Behavioural. It is built from what people searched, asked, bought and watched. The confidence panel groups the seven components by the kind of evidence each one is: four are read directly off audience activity, two combine activity with topic signals, and one is a source list. Any component that returns nothing is named rather than dropped from the total.
Not on its own. A segment can over-index strongly and still be small in absolute terms. The index locates affinity; sizing the opportunity is a separate judgement.
Whoever typed the queries the panel is built from. For an interest audience that is the audience. For an occupation it is usually the people looking to hire one, and the cards say “age of searchers” rather than “age of this audience” when that is the case. Where it cannot tell, it uses the more cautious label.
It ranks by the share of the practitioner set that follows an account, not by that account's own follower count, so a large account everyone follows anyway does not win by size. Accounts that recur across unrelated audiences are filtered out as generic. It does not find a graph for every audience, and when it cannot it names the platform and the reason instead of showing an empty card.
Demographic concentration comes from DataForSEO Trends. Communities, conversations, language and sources come from Reddit, YouTube, Amazon and Google SERP. The follow graph reads public social profiles. Each panel prints its own source rather than leaving it unattributed.
Be first. Not just found.
Free tier, no card. Run your first check in a few minutes.