Metrics, denominators, and confidence intervals
Definitions for every Signal metric, how each is calculated, and why every rate is shown with its sample size and a 95% interval.
Updated
Every rate in Signal is shown with its denominator (n) and a 95% confidence interval. These are the definitions used throughout the product, in reports, and in exports.
Mention rate
The share of answers that name your brand or one of its alternate names. Denominator: all answers in the selected period, engines, and topics. Mentions are matched on whole words, so "Acme" does not match "Acmeville."
Citation rate
The share of answers that cite at least one URL on your domain or its subdomains. Denominator: all answers. A citation is stronger evidence of influence than a mention because the engine relied on your content.
Share of voice
Your brand's mentions divided by all brand and competitor mentions. Denominator: total mentions of tracked organizations. Share of voice only covers the competitors you track, so adding a competitor changes it.
First-position rate and average position
Position is the order in which tracked organizations are first named in an answer (1 = named first). First-position rate is the share of answers that mention you in which you are named first. Denominator: answers that mention you.
Sentiment
A lexicon heuristic applied to the sentences that mention your brand, labeled positive, neutral, or negative. It is designed to surface answers worth reading, not to certify tone. Open the answer to read the full context.
Accuracy
The share of reviewed claims about your brand marked accurate. Claims are sentences that mention your brand and contain a checkable statement such as a date, price, location, or capability. Claims that exactly restate a value on your fact sheet are confirmed automatically; everything else waits for review.
Confidence intervals
Signal uses the Wilson score interval at 95% confidence. For 30 mentions in 100 answers, the mention rate is 30% with an interval of 21.9% to 39.6%, shown as "30% ±8.8 pts." When two periods' intervals overlap, Signal does not label the change as meaningful. Small samples produce wide intervals; that width is information, not a defect.
What these metrics do not measure
Visibility in answers is not the same as traffic, conversions, or revenue, and an engine citing you does not mean a user clicked. Signal reports what was observed in answers. Connect outcomes in your analytics tools and compare timing rather than assuming causation.