Documentation
Back
  • Connecting a site and reading your first audit
  • How Noverva reports numbers
  • Reading your audit
  • How opportunity scores work
  • The REST API
  • Connecting an AI client over MCP
  • Plans and limits

How Noverva reports numbers

Every figure says where it came from, every rate shows what it was counted over, and anything we could not check is excluded rather than counted as zero.

Most SEO tools present every number the same way, whether it was counted, estimated, or bought from somebody who estimated it. Noverva does not, and this page explains what the labels mean so you can argue with them.

Provenance

Every figure carries one of four labels, and the interface shows it.

Measured. Counted directly from something we observed. Pages crawled, findings raised, internal links found. If a measured number is wrong, the underlying record is wrong and you can go and look at it.

Modelled. Estimated by Noverva from a rule we wrote. Opportunity scores are modelled. So is anything describing what might happen.

Reported. Stated by a third party and passed through without verification. Search volume from a data provider is reported: we did not count those searches and neither, in most cases, did they.

Unknown. We could not establish where it came from. Rendered so you can see it, never aggregated into anything else.

The distinction matters most when the numbers sit next to each other. A page showing "142 pages crawled" beside "search volume 8,100" is showing one thing we counted and one thing somebody estimated, and a tool that renders them identically is inviting you to trust them equally.

Rates always show their denominator

A percentage without the count behind it is not a measurement.

"You are mentioned in 40% of answers" means something very different at four prompts than at four hundred, and it means nothing at all if half the engines failed to respond. So every rate in Noverva is shown with the number it was computed over.

What did not answer is not a miss

This is the rule the visibility feature turns on, and it is worth stating plainly because it is the one most commonly got wrong.

If an answer engine is rate-limited, errors, or is not configured, it is excluded from the denominator. It is not counted as an answer that failed to mention you.

The alternative — counting an unreached provider as a non-mention — produces the most damaging possible error: a report telling you your visibility fell, when what actually happened is that a key expired at our end. You would then go and change something about your business in response to a number that measured our outage.

A run where some engines did not answer says so on its face.

Scores explain themselves

No score appears without the reasoning that produced it. Open any opportunity score and you see each signal that contributed, what it contributed, and where it came from. Every audit finding cites the rule that raised it, quotes the evidence, and says what to do.

This is enforced rather than encouraged: the database rejects a finding with an empty explanation, and a score with no breakdown cannot be stored.

What Noverva will not tell you

  • A guaranteed ranking or traffic outcome. Nobody can honestly promise one.
  • Projected traffic from AI citations. The conversion rate from a citation to a visit is unknown, and inventing one would be a modelled number wearing a measured number's clothes.
  • A single score out of a hundred. One number hides the denominator, and the denominator is usually the interesting part.

If a figure you need is missing, it is more likely that we refused to estimate it than that we forgot.