Review customer outcomes, workload and operating cost together. The fictional returns-support example keeps one card selection through editing, saving and verification.
Before you begin
- Open the project with permission to edit it.
- Have completed customer conversations in the Environment and period you will review. Internal Run tests do not populate live-activity charts.
- To follow the custom-score example, first create and run the policy Evaluation.
Set up the dashboard
Choose the review scope
Open Analytics. Select your Environment, leave All versions selected for a review across releases, and click Apply. Choose the review dates in the calendar.
The worked example uses Main, November 8–10, 2026. Keep this scope when opening individual Evaluation results.
Select a focused set of cards
Click Edit view at the top right. For this review, keep:
- Quality: Resolution rate and Return policy accuracy.
- Usage: Unique users and User messages.
- Spend: Spend.
Turn off the other cards. Use the drag handles to arrange selected cards within their groups. The summary at the top still provides total conversations and cost per conversation.
Save the layout
Click Save. This persists the project version’s card selection and order; Environment and date filters remain separate.
Verify the saved selection
After the dialog closes, reopen Edit view. Confirm that the five cards above remain selected, then close the dialog.
Verify the review is usable
- Check the Environment and dates before interpreting the figures.
- Read the Resolution rate alongside total conversation volume and cost.
- Check the policy Evaluation’s scored-conversation count. An empty card is a scoring-coverage gap, not a zero score.
When a result needs explanation, investigate the customer conversation. For all card and filter behavior, see Analytics.