Skip to main content

​
Direct answer

A useful prompt set represents the decisions you want to study, not every possible question about your market. Review prompt wording, topic coverage, funnel stages, markets, providers, and models together. There is no universal prompt count that makes the dataset representative.

​
Data required

  • Your active prompts in Workspace > Prompts
  • Topic, tag, and funnel-stage assignments for those prompts
  • The provider, model, country, language, and monitoring settings in Workspace > Monitoring
  • Completed answers for a consistent analysis period
  • A separate list of the products, audiences, and buyer decisions you intend to monitor
Optional discovery inputs include Google Search Console data and Query Fan-Out records. Query Fan-Out contains web queries observed during provider processing; it is not a measure of user search demand.

​
Workflow

  1. In Workspace > Prompts, remove duplicates and flag prompts that name your brand when the purpose is to measure category discovery.
  2. Group the remaining prompts by topic and funnel stage. Compare that structure with the products, audiences, and decisions in your monitoring brief.
  3. In Workspace > Monitoring, verify that the intended providers, models, countries, languages, and schedule apply to the prompt set.
  4. In Monitor > Prompt Tracking, apply one scope at a time and inspect prompts with missing answers, no SELF mention, or materially different results across providers.
  5. Read the corresponding answers in Monitor > Response Analysis before deciding that an absence represents an opportunity.
  6. Use Query Fan-Out or connected search data only to propose candidate prompts. Review each candidate for relevance before adding it.

​
Interpretation

Product Coverage is prompt-level: distinct answered prompts with at least one SELF mention divided by distinct answered prompts. Product Mention Rate is answer-level: answers with a SELF mention divided by answers with at least one SELF or DIRECT brand mention. Keep these units separate. A prompt can be covered because one answer mentions your brand while still having a low answer-level Mention Rate across providers or models. A prompt with no SELF mention may expose a meaningful gap, an irrelevant question, or an incomplete sample.

​
Possible next actions

  • Test whether adding a missing, category-neutral decision prompt changes what you learn about a topic.
  • Test whether splitting an ambiguous prompt into clearer intents produces more interpretable answers.
  • Pause a prompt as a reversible experiment when its answers consistently fall outside the intended category.
  • Add a tag before changing the prompt set when the apparent gap may be a reporting-structure problem.
Treat each action as a hypothesis. Record the prompt-set change so comparisons do not mix different portfolios without explanation.

​
Limitations

  • Qwairy observes answers generated for your configured prompts; it does not prove that the set represents all user demand.
  • Provider and model output can vary between runs.
  • Changing prompts, providers, models, countries, or schedules changes the measured population and can break trend comparability.
  • Missing answers are not evidence of brand absence.