Conversation Analytics
The Analytics module provides AI-powered evaluation of every conversation, helping you understand performance, identify issues, and continuously improve your assistants.
Overview
FineGuide Analytics automatically evaluates each conversation across multiple dimensions, giving you actionable insights without manual review.
Key Metrics Explained
Score (1-10)
A single number representing the overall quality and effectiveness of the conversation.
| Score | Interpretation |
|---|---|
| 9-10 | Excellent — User's needs fully met |
| 7-8 | Good — Solid interaction with minor issues |
| 5-6 | Average — Room for improvement |
| 1-4 | Poor — Significant problems identified |
Performance
Indicates how well the assistant handled the conversation:
| Rating | Meaning |
|---|---|
| Good | Consistently accurate and helpful responses |
| Average | Adequate performance with room for improvement |
| Poor | Struggled to provide appropriate responses |
| Unknown | Not enough data to evaluate |
Sentiment
The emotional tone detected throughout the conversation:
| Sentiment | Description |
|---|---|
| Positive | User shows satisfaction and positive emotions |
| Neutral | Balanced, professional tone maintained |
| Negative | User expresses dissatisfaction or frustration |
| Unknown | Sentiment cannot be determined |
Satisfaction
How content the user was with the interaction:
| Level | Description |
|---|---|
| Very Satisfied | High satisfaction expressed |
| Satisfied | Generally content with the interaction |
| Neutral | Neither satisfied nor dissatisfied |
| Unsatisfied | Some disappointment expressed |
| Very Unsatisfied | Strong disappointment or frustration |
| Unknown | Cannot be determined |
Resolution
Whether the user's query was successfully addressed:
| Status | Description |
|---|---|
| Resolved Successfully | Query fully addressed |
| Resolved Partially | Some aspects resolved, others remaining |
| Not Resolved but Relevant | Relevant info provided but issue not fully solved |
| Not Resolved and Irrelevant | Failed to provide relevant assistance |
| Unknown | Resolution status cannot be determined |
Improvement
Assessment of whether the assistant's responses need enhancement:
| Status | Description |
|---|---|
| Need Improvement | Significant changes required |
| Need Some Refinement | Minor adjustments could help |
| No Need for Improvement | Performing optimally |
| Unknown | Unable to determine |
Using the Analytics Dashboard
Filtering Conversations
Narrow down conversations to find specific patterns:
- Use the filter dropdowns at the top of the table
- Filter by any metric (Score, Performance, Sentiment, etc.)
- Combine filters to isolate specific scenarios
Example filters:
- Sentiment = "Negative" → Find frustrated users
- Improvement = "Need Improvement" → Find problem areas
- Resolution = "Not Resolved" → Find unanswered questions
Viewing Details
Click any conversation row to expand and see:
- Improvement suggestions — Specific recommendations from AI
- Conversation summary — Key points covered
- Link to full conversation — Access complete chat history
Acting on Insights
For each conversation needing improvement:
- Read the AI-generated improvement suggestions
- Click through to the full conversation for context
- Identify patterns across similar conversations
- Take action:
- Add content to Learning Context
- Create FAQ entries for common questions
- Adjust assistant personality settings
Common Use Cases
1. Identifying Problem Areas
Goal: Find conversations where the assistant struggled
Steps:
- Filter by Improvement = "Need Improvement"
- Review the improvement suggestions
- Look for patterns (same question types, missing information)
- Update training content accordingly
2. Performance Monitoring
Goal: Track quality over time
Steps:
- Set a date range for the period you want to analyze
- Review average Score and Performance metrics
- Compare to previous periods
- Identify trends (improving or declining)
3. Sentiment Analysis
Goal: Understand user satisfaction
Steps:
- Filter by Sentiment = "Negative"
- Review what caused frustration
- Identify if issues are:
- Knowledge gaps (add training content)
- Response style (adjust personality settings)
- Missing features (enable actions or escalation)
4. Quality Improvement Workflow
Recommended regular workflow:
- Weekly: Filter for "Need Improvement" conversations
- Review: Expand rows to see AI suggestions
- Investigate: Click through to full conversation logs
- Fix: Either:
- Click "Provide a better answer" to add FAQ
- Add documents or URLs to Learning Context
- Adjust assistant configuration
- Monitor: Check if metrics improve in following weeks
Tips for Better Analytics
Keep Conversations Meaningful
- Encourage users to provide context
- Use suggested questions to guide conversations
- Enable follow-up questions for clarification
Review Regularly
- Set aside time weekly to review analytics
- Focus on "Need Improvement" conversations first
- Track changes after making updates
Use Multiple Metrics
- Don't rely on a single metric
- A high Score but negative Sentiment may indicate issues
- Low Resolution with good Performance might mean knowledge gaps
Next Steps
- Configure Assistant Features to improve interactions
- Add Learning Context to fill knowledge gaps
- Set up Webhooks to alert on low scores