Read the scope before the score
Start with the selected brand, date range, market, surface, topics and platforms. Filters change the retained population you are looking at. A percentage from five selected answers cannot be compared directly with a percentage from a different question set or source mix. Save a named view when the same scope will be used for a monthly stakeholder report.
Overview leads with a finding and a suggested next action. Treat that summary as a route into the evidence. The chart’s definition, “How this is measured” and “View data” explain the population and reveal the underlying values. Missing values mean unmeasured or unavailable, not automatically zero.
Separate mentions, citations and position
Visibility describes the share of retained usable answers that name the brand, with a count such as 4/5. Owned-site citation asks whether the answer cites the brand’s website. A third-party article mentioning the business is a different source relationship. Position must have a defined basis; a printed numbered position is not the same as the row number in an alphabetically ordered table.
Read the answer log to see how each answer treats each compared brand. “Named without a number,” “not named” and “unavailable answer” have different meanings. A website appearing among retrieved pages is also not necessarily a citation in the final response. Use Domains and URLs to inspect that distinction.
Understand share of observed mentions
Share of voice distributes observed brand-answer mentions across the compared brands in the current sample. Suppose there are 20 such mentions and your brand accounts for five: its share is 25%. This hypothetical example explains the denominator; it is not an IndexHalo customer result. Adding another compared brand can change the share even when your own mention count is unchanged.
This measure is not market share, revenue share or the proportion of all AI conversations about your industry. For a defensible trend, keep question wording, markets, source set, compared brands and inclusion rules stable. If the population changes, show the new baseline explicitly.
Read Perception as attributed language
Perception groups what the retained answers say about the brand. Read the supporting passages, not only a tone label or attribute summary. Positive language about a feature does not establish customer satisfaction, verified product quality or a provider’s private preference. Unknown or unclassified tone should remain visible as such.
Insights helps compare patterns across sources and brands; Ranking provides the brand table. Use them to identify a question worth examining, then open the original passage. A one-answer source column is useful evidence about that answer, but weak evidence of a stable pattern.
Write a conclusion someone can check
A useful summary names the question set, dates, sources, numerator and denominator, then states the next investigation. For example: “In this saved sample, the brand was frequently named but its own website was cited less often. Review the cited comparison pages before changing our pricing page.” Avoid declaring improvement from a single snapshot or attributing a later change to one edit without a suitable experiment.
Your next step
Choose one dashboard percentage and explain its numerator, denominator and practical implication in a sentence.
Continue to the next guide → · All Academy lessons