COMPLETED RUNS / EXPLAINABLE EVIDENCE
Compare the same rules.
Declare the time semantics. Select the measurement phase. Hand over a review someone else can recalculate.
First step: open the original demo, or import your baseline and candidate below.
Runs & phases
Selection is always explicitChoose individual-request labels
Exclude controller and summary labels. Names cannot reliably identify them. Mixed rows that cannot be separated are unsupported for formal review.
Labels appear after importing a run.
Error rate always uses every selected sample. No outlier removal. No automatic deduplication.
Diagnostics may include unphased samples. Formal handoff requires both modes, units, windows, and raw-request confirmations.
Same-rule comparison
Nearest-rank · millisecondsYour sample comparison will appear here.
Define both runs, include labels, then calculate. Missing samples remain unknown.
Review & handoff
No overall pass or release verdictEach label’s optional p95 and error thresholds are observations of this sample only. The report preserves unknown conditions and declared mismatches.
No review prepared. Any edit makes an existing review stale.
Preview the outward report
Explicit full-data saving
These choices include both original files, filenames, settings, notes and selection rules. Nothing is saved automatically.
Browser storage is limited and may be cleared. Download a full backup for durable recovery.