Analytics gives you a clear picture of how your agent is performing and how visitors are engaging with it. Here's what each number means and how to use it to improve over time.
Getting to Analytics
Click Analytics in the sidebar. The default view shows the last 30 days. Use the date picker in the top right to change the range.
Key metrics
Total Conversations — the number of conversations started across all channels in the selected period. A conversation starts when a visitor sends their first message.
Resolved — how many conversations were marked as resolved (by you or automatically). A high resolution rate generally means your agent is handling things well.
Avg. Response Time — the average time between a visitor's message and your agent's reply. This should be under 5 seconds for most agents. If it's higher, there may be a processing issue.
Unique Visitors — the number of distinct visitors who started a conversation. One visitor can have multiple conversations.
Volume chart
The chart shows conversation volume over time. Look for patterns:
- Spikes often correlate with product announcements, blog posts, or promotions
- Dips during weekends are normal for B2B; for consumer products, weekends can be busier
- Consistent growth in volume as you publish more content and your widget gets more traffic
Conversations by channel
The channel breakdown shows how much of your conversation volume comes from each channel (website widget, Gmail, Telegram, etc.). This tells you where your visitors prefer to reach out and where to focus your training efforts.
Top questions
A list of the most common questions visitors asked during the period. This is one of the most actionable analytics views — if the same question keeps appearing and your agent isn't handling it well, that's the first thing to add to your training data or knowledge sources.
Resolution rate by channel
Some channels tend to have higher resolution rates than others. Website widget conversations often have higher resolution rates because visitors ask simpler questions. Gmail conversations can be more complex and take longer to resolve. Comparing across channels helps you identify which integrations need more tuning.
Using analytics to improve your agent
Analytics works best when you use it as a feedback loop:
- Low resolution rate → go to Conversations, filter by Unresolved, read through them, identify patterns
- High volume on a specific topic → check if that topic is well-covered in your knowledge sources
- Spike in conversations after a change → connect the timing to what changed (new feature, changed pricing, etc.)
- Consistent questions not in your sources → add those topics to your documentation or knowledge base, then re-add that source to your agent
Analytics is a diagnostic tool — the useful action is always what you do with the data.