Direct answer
Use Cockpit > Compare with one funnel stage at a time. Compare a defined metric and its raw components, then inspect the prompts and answers behind the difference. No single metric represents rank across the full funnel.Data required
- Answered prompts with consistent funnel-stage assignments
- Correct SELF and DIRECT competitor classifications
- A fixed provider, model, country, language, topic, tag, and period scope
- Raw components for Mention Rate, Share of Voice, Coverage, and Avg Position
- Full answers contributing to each stage
Workflow
- In Workspace > Prompts, review funnel-stage assignments and unassigned prompts.
- In Cockpit > Compare, select one stage and one metric.
- Record the SELF and DIRECT competitor values with their numerators, denominators, or eligible counts.
- Repeat for the remaining stages without changing other filters.
- Use Cockpit > GEO Matrix and Monitor > Prompt Tracking to locate the contributing prompts.
- Read those answers in Monitor > Response Analysis to verify recommendation context.
Interpretation
Share of Voice is occurrence-level, Mention Rate is response-level, Coverage is prompt-level, and Avg Position is conditional on recorded positions. Each can produce a different ordering. Define rank as the selected metric within the selected stage. Do not average these metrics together or treat a small stage sample as equivalent to a larger one.Possible next actions
- Test whether one decision criterion explains a competitor advantage at a specific stage.
- Correct inconsistent funnel assignments before changing content.
- Investigate provider or topic concentration behind the aggregate.
- Track a fixed funnel cohort after an experiment.
Limitations
- Funnel stages depend on your taxonomy.
- DIRECT competitor configuration affects eligible metrics.
- Answer volumes can differ between stages and providers.
- Observed rank does not establish market share, revenue impact, or causality.