> ## Documentation Index > Fetch the complete documentation index at: https://docs.orbit.devotel.io/llms.txt > Use this file to discover all available pages before exploring further. # Quality leaderboard: read the ranking and act on it > Read the Quality Leaderboard like a supervisor — how points and badges compose from QA scorecards, calls and CSAT, how to normalize for sample size, how to drill into evaluations and recordings, and how to turn a rank into coaching and reporting. # Quality leaderboard: read the ranking and act on it **Quality → Leaderboard** ranks every agent on the points they earned during a window, computed on demand from the QA scorecards, handled calls, and CSAT responses the platform already captures. The ranking is not a spreadsheet export or a score that waits for a weekly job — the page re-computes when you open it, so the board always reflects the data you actually have. This guide covers the reading part: what the numbers are made of, how to compare agents fairly, and where to go next once a rank has your attention. The setup side — authoring the form these scores come from, the sampler that decides what gets reviewed, and the AI auto-scorer that feeds the same ledger — lives in the sibling guides linked at the bottom. ## 1. What the leaderboard shows Each row is one agent in the selected window, ranked by total points with the underlying metrics beside the rank: | Column | What it is | | ---------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **#** | Rank for the window, computed server-side across the full roster. | | **Agent** | The agent's display name. A supervisor sees only the queues they supervise. | | **Points** | Total points in the window. Points accrue from point rules scored against the raw metrics — each rule pays out per unit, scaled to a cap, or as a bonus under a target. | | **Calls** | Calls handled in the window. | | **QA** | Average QA score (0–100) across the evaluations your quality pipeline produced for that agent in the window — human-graded rows from the sampler or manual assignments, plus AI auto-score rows where reviewers have not overridden them. | | **CSAT** | Average customer-satisfaction survey score for the agent's handled calls. | | **AHT** | Average handle time per call. | | **Badges** | Threshold awards earned in the window (for example a QA average at or above a set mark), each with a minimum-sample guard so a single call does not mint a badge. | Points are a composed score, not a source metric. The default ruleset weights QA alongside volume and CSAT so an agent cannot win the board on calls alone. The **Scoring rules** tab on the same page shows the exact rules and badge definitions behind the current board, and lets you apply custom weights that re-rank the rows on the spot — the board announces that with a **Custom weights applied** chip, so nobody reads a tuned ranking as the default one. The board answers `rules_source: "default"` on a normal read; when custom weights are applied it answers `rules_source: "custom"`. Treat the two as different rankings — a custom-weight preview reorders agents and is meant for evaluating a rewards scheme before you adopt it, not for the standing supervision record. ## 2. Open the board against the right window and team 1. Navigate to **Quality → Leaderboard** in the dashboard. Access is role-scoped: owner, admin, and supervisor roles see the board; other roles get a permission notice instead of the table. Agents see their own card under the same page, computed against the same default board — their rank always matches what the supervisor view shows for them, which prevents "your dashboard says one thing" arguments. 2. Pick the window with the **Today / This week / This month** selector. The board re-ranks on each switch. Use **Today** for shift-floor goals and same-shift recognition; **This week** for the standard coaching view; **This month** for compensation and review-conversation inputs, where the sample is large enough to be fair. 3. Narrow to a team with the **queue** filter. Options are the voice queues the board scopes on server-side. **All queues** is your full supervised roster; picking one queue turns the page into a team competition — useful for a tier-one vs. tier-two contest or for isolating a new-hire cohort. Over the API the same reads are `POST /api/v1/quality/gamification/leaderboard` (body: `period`, optional `queue_id`) and `GET /api/v1/quality/gamification/config` for the rule and badge definitions behind the board. A supervisor requesting a queue outside their assignment gets a `403`. ## 3. Read the ranking fairly — normalize for sample size Two agents' points are comparable only when their windows and scopes match, and even then the raw rank over-credits volume. Read the columns together: * **Points vs. Calls.** A high-call agent accumulates volume-based points mechanically. Before concluding anything about quality, split the rank by the **QA** column: an agent sitting at rank 3 with a 94 QA average is doing different work from one at rank 3 on volume alone. * **QA average vs. coverage.** The QA column averages *evaluated* calls, not all calls. If your [sample quota](/guides/qa-sampling-settings) reviews three of an agent's conversations per week, an 88 average over three calls carries more noise than an 84 over twelve from a higher-coverage peer. When a rank surprises you, widen the window — the month board normalizes what the day board exaggerates. * **CSAT vs. AHT.** Fast handle times that drag CSAT down are not a win; the badge definitions usually encode a floor on one metric to gate the other, precisely so a speed-at-all-costs style cannot farm awards. * **Sample-size guards on badges.** Badges earn out only over a minimum number of samples, but the *points* behind them have no such floor — treat a top-ranked agent on a thin window as an unconfirmed signal until the week or month view confirms it. A fair comparison rule of thumb: compare within the same queue, same window, and similar call counts, and weight the QA and CSAT columns over total points when the inputs differ in coverage. ## 4. Drill from a rank into the underlying evidence A rank is a summary; supervision happens on the interactions beneath it. Two drill paths: * **Quality → Agents → your-agent** — the per-agent view carries the QA score, compliance-flag count, CSAT, and adherence for the window, plus the recent-evaluated-interactions list. This is the fastest path from "why is this agent at rank 14?" to the specific calls and scores underneath. * **Quality → Evaluations** filters the evaluation ledger by agent; **Quality → Recordings** holds the call recordings those evaluations grade. Work one low outlier per session — read the scorecard, listen to the recording, and check whether the score survives your own listen. If it does not, that is a calibration problem with your reviewers or the [AI auto-scorer flag threshold](/guides/qa-autoscore-settings), not an agent problem. Auto-sampled and AI-flagged rows land in the same ledger: a score you drill into may carry an override from a reviewer who already corrected the AI's first pass. The corrected total is what the leaderboard counts. ## 5. Act on it — coaching and reporting The board's purpose is targeting, not decoration: * **Coaching targeting.** A sustained bottom-decline in QA or CSAT is the queue for a coaching plan — author one under **Voice → Coaching → Coaching plans**, attach training modules, and close it with a measured before/after report when the agent recovers. Scores below your coaching threshold also auto-assign a plan against the same surface, so a hand-authored plan and an auto-assigned one share one workflow. The full loop is in [Voice coaching plans](/guides/voice-coaching-plans). * **Recognition.** The top of the month board, read with the sample-size fairness above, is your recognition shortlist — badges make a good secondary filter because their minimum-sample guards screen out thin-window spikes. * **Monthly reporting.** For the organization-level service view that pairs with per-agent QA, the [monthly SLA and availability report](/guides/monthly-sla-availability-report) covers delivery health and availability; the leaderboard supplies the per-agent quality side of the same monthly review packet. ## 6. Caveats — run it without poisoning the floor A leaderboard can improve a team or quietly wreck one, and the difference is how supervisors talk about it: * **Never shame the bottom.** Read bottom-rank agents in private and coach from the per-agent view, not from the board in a team channel. The visible use of the board should be recognition at the top and trend-checking in the middle. * **Use the trend, not the absolute rank.** A single window's rank is noise at low sample sizes; a three-window trajectory — rising, flat, or declining — is a coaching signal. An agent climbing from rank 18 to 11 over a month is winning even though 11 never appears on anyone's trophy line. * **Custom weights are a preview, not a verdict.** Re-ranking on tuned weights is for evaluating a scoring scheme before adoption. Switching weights mid-competition rewrites history; pin one ruleset per competition period and change it only between periods. * **The board rewards what the ruleset measures.** If agents start optimizing the measured columns at the expense of unmeasured ones — shorter handle time with colder closes, for example — the rules need attention, not the agents. ## See also * [Build a contact-center QA program](/guides/quality-management-program) — the full program from rubric to leaderboard * [QA sampling: weekly sample assignment](/guides/qa-sampling-settings) — the sampler that feeds the human-review side of the QA average * [AI Auto-QA: flag threshold and where auto-scores land](/guides/qa-autoscore-settings) — the auto-score feed into the same ledger * [Voice coaching plans](/guides/voice-coaching-plans) — turn a low trend into a measured coaching plan * [Monthly SLA and availability report](/guides/monthly-sla-availability-report) — the monthly organization-level service report * [Quality Management API](/api-reference/quality) — endpoint reference for the board, config, and per-agent card