Outputs
Outputs returned or generated by VOC Insight MCP.
Output Categories
| Output | Examples |
|---|---|
| Brand metrics | Volume, reach, engagement, sentiment, platform distribution |
| Posts | Source posts, links, platform, sentiment, engagement, timestamps |
| Sentiment | Positive/negative/neutral distribution and key drivers |
| Voice share | Brand vs competitor voice-share trend |
| Competitor comparison | Volume, engagement, sentiment, platform performance |
| Risk alerts | Negative spikes, platform risk, emerging issues |
| AI report | AI summary, multi-dimensional analysis, recommendations, evidence |
| Persona | Persona definitions, motivations, demographics, discussion topics, competitors, and quality data |
| Persona interview | Analyst-mode or first-person simulated answers, insights, suggestions, and markdown |
Key Fields
| Data | Common Fields |
|---|---|
| Brand metrics | mentions, reach, engagement, sentiment, platformBreakdown |
| Posts | postId, platform, url, text, author, publishedAt, sentiment, engagement |
| Sentiment | positive, neutral, negative, drivers, examples |
| Competitors | brandId, brandName, voiceShare, sentimentScore, engagementRate |
| Collection progress | jobId, status, progress, startedAt, completedAt |
| Persona overview | reportId, availableSections, activeRun, truncated, omitted |
| Persona module | personaId, nextCursor, items or module-specific data |
| Persona billing | chargedCredits, collectionEstimatedCredits, personaCredits |
Data Readiness
Read and analysis tools should be used after collection has completed and posts are available. The practical gate is collection status completed or partial, with post count greater than 0. A completed collection with 0 posts usually means no available discussion matched the current setup; it is stale data, not necessarily a system failure.
Traceable Evidence
VOC Insight MCP results should preserve raw post evidence where possible, including platform, URL, timestamp, and engagement data. Customers can keep asking which posts support a conclusion, or filter raw content by platform, time, sentiment, and popularity.
Persona generation and interview results are marked simulated: true. They are research simulations based on collected data, not verbatim statements or testimony from real users; external reports should label them accordingly.