Skip to main content

Test Results

After a test run completes, Swarm produces several layers of analysis.

Per-persona results

Each persona’s run includes:
  • Step-by-step walkthrough — every action taken, page visited, and form filled
  • Screenshots — captured at each step for visual reference
  • Thoughts and reactions — what the persona was thinking at each step
  • Status — whether they completed the goal, got stuck, or encountered an error

Synthesis report

The synthesis report aggregates findings across all personas:
  • Executive summary — a high-level overview of the test results
  • Common friction points — issues encountered by multiple personas, ranked by severity
  • Positive highlights — things that worked well across the board
  • Tickets — structured issue descriptions ready for your issue tracker

Tickets

Each ticket includes:
  • A descriptive title
  • Steps to reproduce
  • Which personas encountered the issue
  • Suggested severity level
These are formatted for direct import into Linear, Jira, GitHub Issues, or any project management tool.

Judge verdict

An independent AI judge reviews the test evidence and provides:
  • Verdict — pass, fail, or partial
  • Confidence level — how certain the judge is in its assessment
  • Evidence blocks — specific observations supporting the verdict
  • Recommended actions — prioritized list of improvements
The judge acts as a second opinion — it reviews the raw data independently rather than relying on persona self-reports.

Viewing results

Results are available in two places:
  • CLI — streamed in real time during the test, with a summary at completion
  • Dashboard — full detailed view at the live dashboard link shown when the test starts