Read run results and logs

Start with run validity

Before interpreting application behavior, confirm the test itself was valid. Runner saturation, missing data files, invalid credentials, or a target maintenance window can all produce misleading failures.

Read by label

The whole-run average is rarely enough. Inspect important labels or transactions separately so a slow checkout path does not hide behind a fast static-asset request.

Preserve the decision

Add tags and notes to explain whether the run passed, failed, or needs a rerun. Future comparisons are much easier when the reasoning lives next to the result.