Help

Read a scorecard

Every completed run has a scorecard: an overall score, pass or fail, four scored layers with the judge's reasons, call metrics, and the full transcript.

Steps

  1. Open the run. In Studio, go to Tests and press View on a run. Or fetch it with GET /api/tests/<run-id>.
    curl -s https://api.wordis-bond.com/api/tests/$RUN_ID \
      -H "Authorization: Bearer $TOKEN" | jq '.data | {score, passed, scorecard, metrics}'
  2. Start with the overall score, out of 100, and the pass or fail badge.
  3. Read the four layers: infrastructure, execution, reaction and outcome. Each has its own score, pass or fail, and the judge's reason.
  4. Read the metrics. Latency p50 and p95 are the agent's response times per turn, in milliseconds. Script adherence (WER) is how far the agent's words were from the expected lines you pinned; it reads n/a when you pinned none. Barge-ins count the times speech overlapped.
  5. Check the judge line. It names the judge model and its version. Compare scores only between runs with the same judge version.
  6. Read the transcript to see each turn the judge scored.
  7. Share it. On the Tests list, press Share on a completed run. Studio copies a public link to the report. Anyone with the link can read the scorecard without an account. The link expires after 30 days. Or ask for the link with GET /api/reports/<run-id>/link.
    curl -s https://api.wordis-bond.com/api/reports/$RUN_ID/link \
      -H "Authorization: Bearer $TOKEN" | jq '.data | {url, expiresAt}'

What you get

You know whether the run passed, which layer failed and why, and you have a link you can send to someone who has no account.

If something fails

StatusCodeMessageWhat it means
404—Run not foundThe run id is wrong, or the run belongs to another account.
404NOT_FOUNDRun not foundYou asked for a share link for a run id this account does not have.
404NOT_FOUNDReport not foundThe share link is wrong or has expired. Ask for a new one.
401—Session expired; sign in againYour session ended. Sign in again.

Not ready yet

  • A sandbox run carries no audio, so its infrastructure layer is judged from turn timing in the transcript, not from audio quality.

Last checked 2026-10-11. More help: Sign up and run your first test · Run a suite from CI · Score a captured transcript · Schedule a suite.