Skip to main content
Answer: “How is this firm represented in the service-selection questions that matter to our audience?” For a private-equity use case, first determine whether the measurement actually includes banker-persona questions.

​
Prepare the scope

Use list_projects and list_runs to choose the project and completed run. Check readiness, then call get_results_stability for the same run and supported filters. For each measurement, preserve the scope, returned confidence interval, and row-quality fields. See run selection and decision-grade measurement. The JSON examples below are MCP tools/call parameter objects. Replace ID placeholders and sample selections with values discovered in your account.

​
Establish which perspectives exist

  1. Read get_project for the firm and competitor context.
  2. Use list_dimension_options and list_queries to inspect measured topics, intents, and any user_profile values.
  3. If a banker persona is present, use that exact value in supported measurement filters and the corresponding stability request. If it is absent, report the existing topic view and state that a banker-specific measurement is unavailable.

​
Retrieve the evidence

Begin with entity-level mention coverage for the configured brand and competitors. Add get_sentiment and list_mention_statements to inspect how those mentions are framed.
The example above is unfiltered. Add a persona or topic filter only after discovering a valid value, using the filter format. Keep the same selection across compatible calls.

​
Build the report

Organize the report by measured service-selection topic or verified persona. Show entity coverage, intervals, sentiment scope, and representative supporting statements. Explain where the evidence is limited. A banker perspective is an interpretation of a measured audience slice, not a new historical dataset created by asking the assistant a question. Where no such slice exists, the report can identify questions for a future measurement without claiming those questions were already tested. Reference: dimensions, queries, mention scores, sentiment.