Direct answer
Filter Cockpit > Compare to one provider and model at a time. Define outperform using a specific metric, preserve a matched prompt cohort, and trace the result to answers before ranking competitors.Data required
- Matched answered prompts across the providers or models being compared
- Correct SELF and DIRECT competitor classifications
- Fixed country, language, topic, tag, funnel, and period filters
- Raw components for the selected metric
- Representative answers and exposed citations
Workflow
- In Cockpit > Compare, select one provider, model, and metric.
- Confirm comparable completed-answer and answered-prompt counts.
- Record SELF and DIRECT competitor values with raw components.
- Repeat for another provider without changing the remaining filters.
- In Cockpit > GEO Matrix, locate topic and prompt cells driving the gap.
- Read the underlying answers in Monitor > Response Analysis and inspect exposed sources when relevant.
Interpretation
Product Share of Voice counts mention occurrences. Product Mention Rate is response-level. Product Coverage is prompt-level. Avg Position is conditional on recorded mention positions. A competitor can outperform on one metric and not another. A provider with a smaller or different prompt sample is not directly comparable. State the provider, model, metric, filters, and unit with every conclusion.Possible next actions
- Test whether a provider-specific content or evidence gap persists on matched prompts.
- Investigate one topic or model rather than applying a broad competitive response.
- Review aliases and DIRECT relationships when denominator components look incomplete.
- Collect additional comparable answers before acting on a small sample.
Limitations
- Provider output varies between runs and models.
- Competitor classification affects eligible product metrics.
- Unequal samples can reverse rankings.
- Observed mentions do not establish recommendation quality or commercial performance.