Analytics¶
In Studio, Analytics gives you an aggregate view of how an AI Analyst is being used: success rate, durations, conversations over time by channel, top users, and the metrics, models and glossary terms the analyst uses most. To inspect individual conversations, use Tasks instead.
Availability
- Who can do this: members with the Edit capability in the active Workspace.
Test runs are excluded
Every number, chart and export on this page is calculated from real conversations only. Conversations run from the analyst's Test your analyst panel in Studio are always excluded. See Testing your AI Analyst.
Open Analytics¶
- In Studio, go to AI Analysts and select the analyst you want to inspect.
- Click Monitoring in the analyst's header and select Analytics.
If you are already in Tasks, use the Tasks / Analytics toggle at the top right.
Choose a range¶
Use the Range control at the top to show the last 7 days, 30 days (the default) or 90 days. The range applies to every tile, chart and table on the page, and to the export.
Statistics tiles¶
| Tile | What it shows |
|---|---|
| Success rate | The share of finished turns (questions and follow-ups) that completed. Finished turns are those that completed, failed or were cancelled. |
| Avg duration | The average run time of a turn. |
| Last run | The date the most recent question started. |
| Total queries | The number of questions and follow-ups (turns) asked in the range. |
| Tasks | The number of conversations started in the range. |
| Artifacts | The number of charts, tables, reports and queries created in the range. |
| Follow-up rate | The share of conversations started in the range that have more than one turn. |
Success rate, Avg duration, Last run and Follow-up rate show — when there is no data for the range. The count tiles show 0.
A high follow-up rate can mean that users are exploring their questions in depth, or that first answers did not fully satisfy them. Use Tasks to read the conversations and find out which.
Conversations over time¶
A daily bar chart of new conversations, split by channel: Explorer, Slack and Teams. Days are calendar days in UTC. Use it to see which channels drive usage. A channel appears in the chart only when it has conversations in the range. To make the analyst available in Slack or Teams, see Messaging apps.
Top users¶
A bar chart of up to 10 users who asked the most questions and follow-ups (turns) in the range. Use it to see who gets the most value from the analyst and who to ask for feedback.
When the range has no data, the chart reads No data for this range yet.
Top used¶
The five metrics, models or glossary terms the analyst used most often in the range, with Name and Uses columns. Use the Metrics / Models / Terms switch to change the list. Metrics is shown first.
A use is counted each time a query from the analyst that references the item runs successfully against your data. Failed queries are not counted. An item that has since been deleted is shown as (deleted). Counts are recorded in the background, so they can lag slightly behind the latest conversations.
Low counts can point to parts of your semantic layer the analyst underuses.
Export the statistics¶
Click Export CSV to download the statistics for the selected range. The file is named <analyst name>-analytics-<range>.csv (for example, Sales Analyst-analytics-30d.csv).
The file has three columns: section, label and value. It contains:
- Every statistic from the tiles. Rates are fractions between 0 and 1, durations are in milliseconds, and dates are ISO timestamps.
- The change in the number of conversations compared with the previous period of the same length, as a percentage. This value is only in the export. It is blank when the previous period had no conversations.
- The number of artifacts by type (chart, table, report and query).
- The top users, metrics, models and terms.
The export does not include the daily data behind Conversations over time.
Usage on the analyst's Overview tab¶
The analyst's Overview tab also has a Usage section with two counters. Both exclude test runs:
- Tasks run: the number of questions and follow-ups (turns) users have run with the analyst, across all time.
- Active users (30d): the number of distinct users who asked the analyst a question in the last 30 days.
Analyze usage with Steward¶
To dig deeper, open the Steward panel and ask it to review this analyst's recent conversations. For example: Review the last week of conversations with Sales Analyst and suggest semantic layer improvements. Steward can search and read the analyst's conversations to answer.
Troubleshooting¶
| Message | What to do |
|---|---|
| Could not load analytics. | Reload the page or try again. |
| Could not export analytics. | Click Export CSV again. If it keeps failing, reload the page. |