Skip to content

Analytics & Dashboard

The Analytics dashboard gives you a complete picture of how your AI bot and support team are performing. Use it to measure effectiveness, identify knowledge gaps, and continuously improve your bot.

The main dashboard shows high-level KPIs at a glance:

KPIDescription
Total ConversationsNumber of conversations started in the selected time period
AI Resolution RatePercentage of conversations resolved by the bot without human handoff
Avg. Response TimeAverage time from user message to first bot or agent reply
Handoff RatePercentage of conversations escalated to a human agent
Active UsersUnique users who interacted with the bot in the period
Positive Feedback RatePercentage of thumbs-up reactions out of all feedback submitted

Use the date range picker to filter all metrics by day, week, month, or a custom range.

The RAG Analytics section shows how the retrieval pipeline is performing:

  • Retrieval Success Rate — Percentage of queries where the system found relevant content in the knowledge base
  • Top Retrieved Documents — Which documents are retrieved most often, indicating high-demand topics
  • Avg. Relevance Score — Average confidence score from the re-ranker across all queries
  • Query Volume by Hour — Message volume heat map to identify peak usage times

Use this data to identify documents that are rarely retrieved — they may need better titles, more keywords, or a content review.

The Unanswered Analytics section surfaces questions the bot could not answer:

  • Unanswered Questions List — A chronological list of questions that resulted in a low-confidence response or handoff
  • Topic Clustering — Questions are automatically grouped by topic so you can spot patterns
  • Topics Without Docs — Topics that appear frequently in unanswered questions but have no matching document in the knowledge base

Use the Export button to download the unanswered questions list as CSV for offline review or to share with subject matter experts.

Users can rate bot responses directly in the chat widget:

  • 👍 Thumbs Up — Response was helpful
  • 👎 Thumbs Down — Response was not helpful

The Analytics dashboard aggregates this feedback and shows:

  • Feedback volume over time
  • Most downvoted responses — with the original question and bot answer, so you can investigate and improve
  • Feedback trend — Whether satisfaction is improving or declining over time
v1.6.0

Sentiment Analytics

The Sentiment Analytics section provides ML-powered analysis of customer sentiment across conversations:

  • Sentiment Distribution — Breakdown of conversations by detected sentiment (positive, neutral, negative)
  • Sentiment Trend — How overall customer sentiment changes over time
  • Negative Sentiment Alerts — Conversations flagged with negative sentiment are surfaced for priority review
  • Sentiment by Topic — Which topics tend to generate negative sentiment, helping you identify problematic areas in your knowledge base

Sentiment data feeds into the sentiment-aware routing system (see Human Handoff) — conversations with detected negative sentiment can be automatically prioritized for faster human agent response.

v1.6.10

The dashboard includes an interactive Conversations Over Time area chart that visualizes daily conversation volume as a smooth gradient area graph. The chart uses your theme’s primary color and provides:

  • Date axis — Shows dates along the horizontal axis
  • Conversation count — Vertical axis displays the number of conversations per day
  • Hover tooltip — Hover over any data point to see the exact date and conversation count
  • Responsive layout — Automatically resizes to fit the dashboard width

Use this chart to identify trends in conversation volume, spot peak usage days, and correlate volume changes with events like product launches or marketing campaigns.

Admins have access to additional operational metrics:

  • Agent Response Time — How quickly human agents respond after taking over a conversation
  • SLA Breach Rate — Percentage of handoff conversations that exceeded the SLA target (see Human Handoff)
  • Conversation Volume by Channel — Breakdown of conversations by Web Widget, LINE OA, and other channels
  • Knowledge Base Coverage — Document count, total chunks indexed, and last updated timestamps
  • Cost per Conversation — Estimated AI inference cost per resolved conversation (Enterprise plan)

All admin metrics can be exported as CSV or viewed as charts in the dashboard.

v1.6.1

The Language Detection Analytics page (Analytics → Language Detection) gives admins visibility into how visitor languages are being detected across conversations.

Key metrics:

MetricDescription
Top LanguagesDistribution of detected languages across all conversations in the selected period
Detection Method BreakdownHow each language was identified — Unicode script analysis, browser language header, diacritics, or conversation continuity
Detection RatePercentage of conversations where language was successfully detected vs. defaulting to org language
Daily Language TrendVolume of conversations per language over time

Date range: Use the selector in the top right to filter by last 7, 14, 30, or 90 days.

Use this data to:

  • Verify that the language detection is working correctly for your customer base
  • Decide whether to enable the Widget Language Selector for specific locales
  • Identify whether customers are using languages that are underrepresented in your knowledge base