Skip to main content
The AI Chatbot dashboard combines conversation volume, containment, handoffs, satisfaction, public actions, usage/cost, languages, widget comparisons, and optional Topics analytics. Use it to answer not only how much Chatbot is used, but also where visitors still need a human and which public workflows fail.

​
Headline metrics

The current dashboard includes:

​
Conversations

Total Chatbot conversations in the selected reporting range, with the percentage closed.

​
Containment

The share of conversations handled without a form handoff, together with the form-handoff rate. Containment is not synonymous with correctness. Review CSAT, action failures, open/escalated conversations, and actual transcripts too.

​
Chatbot CSAT

Normalized satisfaction score on a 0–100 scale plus the number of responses used. The underlying widget can collect thumbs, 1–5 stars, or 1–5 satisfaction; the dashboard normalizes those styles for comparison.

​
Action success

Success rate for public action attempts, with succeeded and failed counts. A high answer-containment rate combined with poor action success can mean the bot understands the question and still fails at the part the customer came to accomplish.

​
Needs attention

Open or escalated conversations currently unresolved in the reporting range.

​
Attachments

Validated visitor uploads in the selected period.

​
AI cost

Committed AI-provider cost, when cost display is enabled in dashboard preferences.

​
Trend widgets

The dashboard currently supports charts for:
  • Conversation volume over time
  • Containment and handoff trend
  • Chatbot CSAT over time
  • Public action outcomes
  • AI cost over time
  • Languages

​
Widget comparison

Compare widgets using:
  • status;
  • conversation count;
  • handoffs; and
  • CSAT / number of CSAT responses.
This is especially useful when one organization has separate widgets for different brands, products, regions, or public knowledge Collections.

​
Topics analytics

When Topics is enabled, additional widgets can include:
  • Topic distribution
  • Sentiment trend
  • Emerging smart tags
  • Approximate geography
  • Device/browser mix
  • Entity attribution
These dimensions depend on the corresponding Topics/Visitor Data policy. An empty chart can correctly mean collection was never enabled rather than that every customer visited from nowhere using no browser at all. Topics →

​
Geography

Approximate geography is derived server-side from the configured GeoIP data path when those visitor fields are enabled. Treat this as analytics context, not verified identity or precise household location.

​
Reporting period

The Chatbot dashboard uses the shared configurable dashboard timeframe system. The fallback timeframe is 30 days until a saved user/organization preference is applied. Users can select predefined or custom date ranges. The dashboard uses the viewer’s browser timezone for range/bucket presentation.

​
Customize the dashboard

As with the Assistant dashboard, Chatbot supports persistent preferences such as:
  • visible widgets;
  • widget order;
  • layout;
  • default timeframe;
  • chart presentation; and
  • whether cost is shown.
Permissions determine whether you are changing your own preferences or organization defaults.

​
Large ranges and truncation

The dashboard response records whether a reporting query was truncated by its bounded row scan. If a very broad range produces a truncation warning in the UI, narrow the timeframe before treating the displayed counts as exact totals. The limit exists to keep an analytics request from becoming an accidental database endurance competition.

​
How to investigate a metric

When a number changes:
  1. compare the same time range to the previous period;
  2. segment by widget;
  3. check handoffs and Needs attention;
  4. inspect action outcomes;
  5. review languages/topics if relevant;
  6. open representative Conversations; and
  7. verify whether a configuration or public-knowledge change happened at the same time.

​
Useful metric combinations

Containment + CSAT helps distinguish successful self-service from merely low handoff. Action success + handoff rate shows whether transactional workflows are actually completing. Topic distribution + Needs attention identifies support areas where the bot struggles. Widget comparison + Collections helps validate whether a narrower knowledge scope improves outcomes. AI cost + conversation volume helps interpret model/provider changes without ignoring customer quality.