Skip to main content

​
My scorecard reference

This page is the field reference for /me/scorecard — the per-agent, self-service quality scorecard every operator sees from their user menu. It covers what the tile aggregates, how the window works, who can see whose scorecard, and the first move when a rubric turns red. For the walkthrough narrative, see Read your own QM scorecard.

​
What the scorecard is

A single-member view of the LLM-judged quality pipeline. The page scopes every metric to one signed-in operator: the conversations assigned to them, scored against the organization’s rubrics. It is not a team average and it is not a queue rollup — one member at a time, self-serve. The page answers three questions for an agent: which of my rubrics is failing, is that failure a trend or a blip, and where do I rank this window.

​
Metrics on the page

All figures below come from the quality evaluations the judge writes for conversations assigned to you — scoped to the signed-in user, never an org aggregate.
  • Total scored — rubric evaluations in the window.
  • Overall pass rate — passed evaluations over total, with the raw counts underneath.
  • Average confidence — how sure the quality judge was on average, as a percentage.
  • By rubric — one row per rubric: name, scope badge (inbox or org), passed/total counts, and the pass-rate bar. Sorted by lowest pass rate first; the list is capped at the twenty weakest rubrics, so the KPI totals always come from the full population rather than the visible rows.
  • Top failure modes — the rubrics pulling the overall rate down, sorted worst-first. The coaching focus is the card’s top entry.
  • Daily trend — one bar per day of evaluation volume, colored by that day’s pass rate. Use it to separate a steady failure from a single bad day.
The page also mounts the My performance gamification panel: points earned, a per-rule breakdown, badges, and your anonymised rank on the team board.

​
Scorecard vs the leaderboard

/me/scorecard and Quality → Leaderboard read the same outcomes but answer different questions: For the org-side board — scoring rules, badge thresholds, and supervisor navigation — see Quality leaderboard: scoring rules, badges, and thresholds.

​
Window, refresh, and how bands are read

  • Window — a trailing 7, 30, or 90 days, selected at the top of the page. Default is 30. The same window drives every metric: KPI totals, by-rubric rows, and the daily trend always aggregate the same population.
  • Channel filter — All channels or Inbox only. A Voice entry exists in the API but produces no outcomes today, so the page does not offer it; leave the filter on All unless you are isolating chat work.
  • Refresh — the scorecard is a live read over the evaluation tables. New judge outcomes appear as they land — there is no nightly rollup to wait on.
  • Color bands — the same thresholds apply to every pass rate on the page: green at 80% and above, amber from 60%, red below that.
  • Org position — an agent sees exactly one cross-agent comparison, and it is bounded: the anonymised rank out of the roster on the gamification panel. Rubric-level outcomes are never compared across agents on this page; org-aggregate percentile views belong to the leaderboard’s reviewer-scoped surface.

​
Access — who sees whose scorecard

Scope is enforced on the API, not hidden in the UI:
  • Agent role — reads their own scorecard. Any authenticated operator with inbox access can open theirs; no extra role is required.
  • Supervisor, admin, owner — read any member’s scorecard, normally through Quality → Agents → member, which renders the same data plus supervisor context.
  • Viewer / other agents — see none. Requesting a scorecard that is not yours returns 403, and agents never see each other’s rubric results. The only cross-member comparison visible to an agent is the anonymised rank.

​
Acting on it

Work the weakest rubric first — the top entry in Top failure modes — and convert a failing evaluation into training before the next live conversation:
  1. Check the passed/total count, not just the percentage — a 40% rate on five evaluations is noise.
  2. Compare 7 days against 90 days to tell a recovering rubric from a worsening one.
  3. Open the failed evaluation in Quality → Evaluations and acknowledge it, or appeal it with a written note if you disagree — the appeal returns to a reviewer to resolve.
  4. Turn the finding into coaching material: a supervisor can attach a coaching card from Quality coaching cards, and you can rehearse the scenario in Practice Studio roleplay before your next live conversation.

​
Troubleshooting

  • “No scored conversations yet” — quality management is enabled, but nothing you handled was scored in the current window. Evaluations appear after you start handling conversations; widen the window to 90 days before concluding something is broken.
  • “Your scorecard isn’t available” (403) — quality management is not enabled for your organization, or you opened another member’s scorecard without a supervisor, admin, or owner role. Ask your admin.
  • Empty by-rubric list with counts in the KPI tiles — your filters narrowed the visible rows to zero. The by-rubric list shows at most the twenty weakest rubrics; clear the channel filter or widen the window.
  • A rubric you know you were scored on is missing — only the twenty lowest pass rates render. A healthy rubric sits above the cap and is still counted in the KPI totals.

​
See also