Skip to main content

​
The Quality hub: leaderboard, scorecards, and evaluations from one page

The Quality section of the dashboard groups every contact-center quality surface behind one hub. This page is the map: the sections below walk each surface end to end — what it answers, which metrics feed it, and how to run it — and point at the in-depth guide behind each page. The reading path is supervisor-first; agents open their own card on the same surfaces. Use The Quality hub: the supervisor loop for the narrative walkthrough; this page is the orientation map — pick the right surface first, then open its in-depth guide.

​
1. What the Quality hub surfaces

The hub’s own landing page (/quality) renders the org-wide scorecard rollup — conversations scored, pass rate, judge confidence, a by-channel and by-rubric breakdown, and a recent-failures triage list — plus an AI auto-QA coverage card and the QA score trend bars. Every row below is one hop off that hub: Agents get a narrower path on the same surfaces: Me → My scorecard is the agent-facing half of the leaderboard (rank, points, badges, plus per-rubric pass rates, the daily trend, and top failure modes), and the My evaluations view on the evaluations page lists only their own rows, scoped on the server. See Read your own QM scorecard.

​
2. How the leaderboard ranks agents — and what feeds it

The board on /quality/leaderboard turns counters the voice and QA pipelines already write into a point total. The default ruleset weights quality over raw volume so call count alone does not win: Badges add threshold awards — Top Performer (rank 1), Podium Finish (top 3), QA Ace (QA average ≥ 90 on at least 3 evaluations), Centurion (100+ handled calls), Customer Champion (CSAT ≥ 4.5 on at least 5 responses). Average-based badges carry a sample guard so a thin window cannot mint them; points pay out regardless of sample size. Compare like-for-like — same queue, same window, similar call counts — and weight the QA and CSAT columns over total points when coverage differs. Rows queue only on official score rows — reviewer-graded or AI auto-scored once a reviewer has graded them; pending and self-scored rows stay out — so the board counts whatever the evaluations ledger corrected. Ties sort deterministically on QA average, then calls handled, then agent id.

​
Leaderboard row shape

Over the API, one ranked entry carries rank, agentId, agentName, totalPoints, the per-rule breakdown, earned badges, and the raw metrics every rule or badge can reference:
Ranked from POST /api/v1/quality/gamification/leaderboard (or the Leaderboard page itself); the six metrics keys and the badge definitions are on GET /api/v1/quality/gamification/config. Visibility is role-scoped — owner and admin see the full roster, a supervisor sees only the queues they supervise, and an agent’s self-view (GET /gamification/me) ranks them against the same default board so their rank always matches the supervisor view. Tuning point_rules in the request previews a custom scheme and flips rules_source to custom — a per-request preview, never the standing record.

​
3. Running an evaluation and reading the scorecard

An evaluation scores one conversation for one agent against one weighted scorecard form. The form is the rubric: weighted sections, each holding weighted criteria, each criterion with a max_score (default 100). Mark a section auto_fail to force the whole evaluation to zero when any criterion in it scores zero — the compliance-gate pattern. The reviewer submits a raw score per criterion and the server re-derives the weighted 0–100 total_score; a client can never set the total directly.

​
Evaluation rubric fields

Every scored row moves through the ack/appeal lifecycle — pending → acknowledged, or pending → appealed → resolved when the evaluated agent appeals and a reviewer closes the appeal. Two system-created row types join the same ledger: auto-sampled rows queued un-scored by a form’s sample_rate_pct and claimed by a reviewer, and AI auto-scored rows (auto_scored: true) whose below-threshold scores carry flagged_for_review until a reviewer corrects them. Work the evaluation flow end to end in QA Evaluations, and stand the whole program — rubric, blind calibration, ack/appeal, leaderboard — up in Build a contact-center QA program. The Agent scorecards page composes the same ledger per agent with adherence, AHT, FCR, and CSAT beside the rubric breakdown; read it in Agent scorecard.

​
4. Practice Studio: coach the pattern before go-live

Practice Studio flips the normal QA loop: instead of scoring an agent after a real call, it hands the agent a safe sandbox to rehearse the hard conversation before one happens. A supervisor authors a scenario — a difficult close, a price-objection customer, an angry escalation — sets a passing score, and assigns it. An AI stays in character as the customer; the agent rehearses inside the dashboard until their best session clears the bar. When the session ends, an automatic score with written coaching feedback shows what landed and what to work on. No real customer is ever contacted — sessions are in-app simulations, so no calls, messages, or charges result. The pairing that works: a rank drop or rubric drift names the coaching target from the leaderboard or scorecard, then a scenario assigned in Practice Studio gives the rehearsal. Practice sessions count toward practice metrics only — they never re-enter the evaluation ledger, so the drill cannot cancel a bad score on an already-graded conversation. Set up scenarios, assign them, and read session feedback in Practice Studio roleplay.

​
5. Wire the hub into your docs flow

The hub is a map, not a second pipeline — reach for the in-depth guide the moment a question goes past “which page”: Every surface in the hub is tenant-owned — form definitions, evaluation rows, practice scenarios, and recordings live in your workspace only, and reviewer surfaces (leaderboard, evaluations, recordings) render to owner, admin, and supervisor seats while agents get their own scoped views.

​
See also