Skip to main content
The Analytics tab gives you a time-filtered, at-a-glance view of how your evaluation is performing across all test runs. Open any evaluation and click the Analytics tab to access it. Evaluation Analytics tab

​
Overall Health

The top of the Analytics page displays a summary card with four key metrics:

​
Average Evaluation Score Across Time

A line graph plots the normalized evaluation score over time. Each data point represents one test run. Use this chart to:
  • Spot trends — a declining line indicates gradual quality degradation
  • Identify regressions — a sharp drop after a specific run pinpoints the change that introduced the issue
  • Validate improvements — confirm that score increases after prompt or model updates
The graph supports time-range filtering to focus on specific periods.

​
Evaluator Performance

Below the score graph, an evaluator performance card breaks down results per evaluator: Each evaluator is listed as a separate row, letting you quickly identify which evaluator contributes most to failures or costs.

​
How to Use Analytics

​
Track Quality Over Time

  1. Run the same evaluation after each deployment or model change
  2. Open the Analytics tab to see the score trend
  3. Investigate any dips or regressions in the graph

​
Identify Weak Evaluators

  1. Review the Evaluator Performance card
  2. Focus on evaluators with the lowest pass rates
  3. Tune those evaluators or investigate the underlying AI behavior

​
Monitor Costs

  1. Check the Eval Cost in the overall health card
  2. Review per-evaluator costs to identify expensive evaluators
  3. Optimize prompt size or evaluator frequency if costs grow unexpectedly
Last modified on August 28, 2026