Metrics, scores, and missing results
Read rubric-level results, inspect metric evidence, and understand why a metric may not have a score.
A rubric is a grouped evaluation dimension. A metric is an individual check within that rubric. On a run, open Metrics Breakdown to expand rubric rows and inspect the metrics beneath them.
Read evidence, not only a number
Select a metric row to see its detailed explanation. Where available, that view connects the result to evidence and technical detail from the interaction.
In a run's Metrics Breakdown, ordinary 0–100 metric rows use these visual bands:
- pass: 80 or above;
- partial: 50–79;
- fail: below 50.
These bands are specific to run-level metric rows. Summary and suite views can use different bands for aggregate scores. Timing and overlap metrics can use seconds and inverse thresholds, and some metrics do not have a pass/fail threshold. A missing score is not automatically a zero.
Understand scoring status
The Metric scoring section uses three public outcomes:
- Eligible — the metric has the information it needs to be scored.
- Needs setup — required context is missing.
- Not scored — the available evidence could not support a score for that metric.
A completed run can still contain metrics that need setup or are not scored. A failed run can show a failure reason while retaining useful metric evidence under Metrics Breakdown.
Use suite signals carefully
Top issues on a suite highlights its lowest-scoring metrics below the displayed threshold of 80. It is a useful place to investigate, not a universal definition of a failing metric.