Analyze
Monitor how the Agent performs and review every conversation it has with your users. Use these insights to improve your knowledge base, spot unanswered questions, and measure the Agent's impact.
Overview
The Analyze section has two tabs:
Statistics: High-level KPIs and usage trends over time.
Conversations: Full conversation logs with search, filtering, and export capabilities.
Dashboard path: Agent > Analyze
Statistics
The Statistics tab gives you a dashboard view of how users interact with the Agent.
Date range
Use the date picker (top-right) to filter all statistics by time period.
Available ranges: All time, Last week, Last month, and Custom (pick a start and end date).
KPI cards
Four summary cards at the top of the page show your key metrics for the selected period. Click any card to update the chart below with that metric's data.
Agent Opened
Total number of times users opened the Agent chat widget during the selected period.
Total Messages
Total number of messages sent by users to the Agent. One user session can generate multiple messages.
Unique Users
Number of distinct users who interacted with the Agent at least once.
Success Rate
Percentage of Agent responses that were not marked as "Not Helpful" by users. A high success rate indicates the Agent is answering questions effectively.
Charts
The chart below the KPI cards visualizes the selected metric over time.
Agent Opened / Total Messages / Unique Users: Displayed as an area chart showing daily volume. Hover any data point to see the exact count and date, plus a View users > link to jump to the users who interacted on that day.
Success Rate: Displayed as a bar chart with a three-color legend:
Messages (blue): Total messages received that day.
Helpful (green): Messages where the user rated the response positively (thumbs up).
Not Helpful (red): Messages where the user rated the response negatively (thumbs down).
Users reached
Below the chart, a table lists the users who interacted with the Agent during the selected period.
Name / Id
The user's display name and unique identifier.
The user's email address (if available).
Last reached
How recently the user last interacted with the Agent.
Tags
User tags. Click add tag to assign a tag directly from this table.
Click See all users at the bottom to open the full Users reached modal.
The modal displays the complete list of users and includes an Export as CSV button (top-right) to download the full user list for further analysis.
Conversations
The Conversations tab is your complete log of every Agent interaction. Use it to review what users are asking, how the Agent responds, and where it falls short.
Conversation list (left panel)
The left panel displays all conversations sorted by most recent. Each entry shows:
User avatar and name: The user who started the conversation.
Preview text: The first few words of the user's message.
Timestamp: When the conversation took place (e.g. "10 hours ago", "3 days ago").
Feedback icons: Thumbs up (👍) and thumbs down (👎) counts indicating how the user rated the Agent's responses.
Filters:
Search by keywords: Free-text search to find conversations containing specific terms.
Date range: Filter by time period (All time, Last week, Last month, Custom).
Response quality: Filter by response type (All Responses, Helpful, Not Helpful).
Export as CSV: Click the button (top-right of the list) to export all conversations matching the current filters.
Conversation detail (right panel)
Click any conversation in the list to view its full content in the right panel. The detail view shows:
User info: Display name, email, and unique user ID.
Date and time: When the conversation occurred.
Full message thread: The complete back-and-forth between the user and the Agent, including user messages, Agent responses, source citations, and CTA buttons.
Three-dot menu (⋮): Available in the top-right of the conversation detail panel:
View user profile: Opens the full user profile in the Users section.
Export as CSV: Download this individual conversation as a CSV file.
Share conversation: Generate a shareable link to this conversation.
Using conversations to improve your Agent
Conversations are the most valuable feedback loop for your knowledge base. Here is what to look for:
Unanswered questions: If the Agent says it does not have enough information, this is a signal to add a new Source, Custom Answer, or connect a relevant Tour.
Low-rated responses: Filter by "Not Helpful" to find responses users marked as unhelpful. Review whether the answer was wrong, incomplete, or just poorly worded.
Repeated questions: If many users ask the same thing, create a Custom Answer to ensure a precise, consistent response every time.
Action opportunities: If users frequently ask "How do I do X?", consider connecting a Tour or Action so the Agent can guide them hands-on instead of just describing the steps.
Quick-create a Custom Answer from a conversation
When reviewing a conversation, hover over any user message to reveal the 💬 Answer this... button on the right side.
Clicking it opens the Custom Answer creation modal pre-filled with the user's question.
This is the fastest way to turn a real user question into a curated Custom Answer: you only need to write the answer and optionally attach an action (Open Post, Launch Experience, Navigate To, or Run Action), then save.
Best practices
Last updated