Give every metric a clear definition
A mention rate could mean answers that name the brand divided by successfully observed answers. An owned-site citation rate could mean answers linking to the client’s domain divided by the same denominator. Both need an explicit query scope and collection period.
Do not treat an unanswered query as “brand absent” unless that treatment is part of a documented method. Show failed or unavailable observations separately so changes in data collection do not look like changes in visibility.
THE WORKING SEQUENCE
- 01Define the query set
- 02Record the evidence
- 03Prioritize the work
Separate the score from the underlying evidence
A composite score can make a large audit easier to discuss, but it compresses different kinds of information. Citation observations, technical access checks, and content quality judgments are not the same measure. Keep their individual results visible.
For a weighted model, publish the section weights and the scoring rubric. The StemsStack sample score is illustrative; it does not represent a customer or a measured result for StemsStack. A higher rubric score does not guarantee more leads or citations.
Compare like with like
Use the same questions, grouping, language, and collection procedure for a repeat audit. Keep branded and unbranded queries separate. If you add an engine or expand the query set, show the old comparable subset alongside the new scope.
Record visible changes to the search experience where practical. Do not imply statistical certainty from a small convenience sample. Report what you observed and identify alternative explanations, including answer variability and changes in sources.
Turn the report into an operating routine
Maintain a backlog of issues with owners and completion evidence. Track the publication dates of important changes. In the next round, check both implementation completion and the answer observations rather than using one to stand in for the other.
A useful monthly review asks which findings are repeatable, what the client has changed, and what needs a better test. If the data are inconclusive, say so. Honest uncertainty is more useful than a chart that implies progress the evidence cannot support.
Put the method to work. Explore the AEO / GEO audit system in the Stems Agency Vault.