Insights & Analytics

What Insights Measure

Insights are calculated from assistant activity within the selected time range.

A conversation represents one user message and the assistant’s response.

Insights are designed to help you:

• Track usage over time

• Monitor response performance

• Observe changes in response quality

Unique Users

The number of distinct users who initiated at least one conversation.

What it represents

• The breadth of assistant usage

• Whether activity comes from many users or repeated interactions

Average Response Time

The average time between a user message and the assistant’s response, measured in milliseconds.

What it represents

• Overall system responsiveness

• The impact of configuration, model choice, and knowledge size

This metric is best used for relative comparison rather than absolute evaluation.

Average Response Score

The average quality score of assistant responses, expressed as a percentage.

What it represents

• How well user questions are matched against the assistant’s knowledge base

• How knowledge alignment changes over time as content is added or updated

Scores are intended for comparison across periods or after training updates.

Conversation Trends

A time-based visualization of conversation volume.

What it shows

• Growth or decline in usage

• The impact of deployments, updates, or integrations

Hourly Activity

A distribution of conversations by hour of the day.

What it shows

• When users interact with the assistant

• Daily and weekly usage patterns

Using Insights

Insights are most effective when metrics are analyzed together and compared over time.

Common uses include:

• Evaluating the effect of new training data

• Identifying changes in usage patterns

• Detecting performance regressions after configuration changes

Limitations

Insights provide aggregated metrics. They do not replace message-level analysis.

For detailed review of individual conversations and responses, use the Conversations page.