Skip to main content

​
Per-queue analytics drilldown console

Every queue in Voice → Queues carries its own analytics page. The list view at /voice/queues shows one summary row per queue — offered, answered, service level, abandon across every queue at once. The drilldown — /voice/queues/{queueId}/analytics — is where a supervisor reads one queue in full: live state pushed over a stream, the summary KPI strip, a volume-and-SLA trend, a day-by-hour heatmap, a per-agent leaderboard, and two breakdowns (who hung up, and how each skill performs). Open it from the queues list by selecting a queue row’s analytics action, or navigate directly to the route above.

​
1. What the page aggregates

The page reads one queue end to end. The live section streams six counters over Server-Sent Events: agents logged in, agents available, agents on call, agents in wrap-up, calls waiting, and the longest current wait. Everything below the live section is historical, aggregated over the window you pick (1, 7, or 30 days) and bucketed (30 minutes, 1 hour, or 1 day):
AHT on this page deliberately excludes the time a caller spends waiting in queue before an agent answers. That matches how Five9, Genesys, and NICE CXone define the metric, so a supervisor comparing Orbit’s number to a prior vendor’s wallboard is comparing the same thing. The AHT tile carries a (?) definition on hover and keyboard focus.
Summary numbers are volume-weighted on the server, so the KPI strip agrees with the cross-queue comparison and the list view — a 1-call bucket never skews a 30-day average.

​
2. Reading the tiles, in order

Read top to bottom; each section answers a narrower question than the one above:
  1. Live tiles — four cards (available agents, on call, calls in queue, longest wait) beneath a connection status line. Live means the stream is connected; Reconnecting means it is re-establishing. Under 5 waiting reads normal; over 5 highlights. Over 2 minutes of longest wait flags. If the stream cannot connect, the section renders an error card with a retry — the historical sections below still load.
  2. Summary KPI strip — five aggregate tiles for the selected window: service level, AHT, ASA, abandon %, and total offered. SLA at 80% or higher renders green, abandon above 10% red.
  3. Time-series chart — offered vs answered bars (left axis) overlaid with the service-level line (right axis, 0–100%) across buckets. An empty window renders an explicit empty state rather than a flat chart.
  4. Heatmap — day-of-week × hour grid of offered volume. Use it for staffing: the hour-row you shade agents for is the hour the grid shows dark.
  5. Agent leaderboard — ranked per agent: calls handled, AHT, average CSAT, average NPS. The first 50 rows mount; Load more pages the rest.
  6. Hangup-party breakdown — who ended calls: caller, agent, system, carrier, or unknown, as share bars. Caller-hung-up share rising alongside abandon % points to wait-time pressure, not agent behavior.
  7. Per-skill breakdown — offered, service level, AHT, ASA, abandon per skill tag on the queue. Same coloring rules as the KPIs. Load more pages long skill lists.
Every section handles its own loading, error, and empty states inline, so one failing query never blanks the whole page.

​
3. Comparing across time

Two selectors in the page header shape every historical section at once:
  • Window — 1 day, 7 days, or 30 days ending now.
  • Bucket — the time granularity of the trend chart: 30 minutes, 1 hour, or 1 day.
The workflow that gets comparison out of these two controls:
  1. Set Window: 30 days with Bucket: 1 day — this is the monthly baseline. Note the SLA line’s average and where it dips.
  2. Drop to Window: 7 days with Bucket: 1 hour — zoom into the trough you found in step 1. Same metrics, finer grain.
  3. For intraday diagnosis, Window: 1 day with Bucket: 30 minutes lines a bad stretch up with the heatmap’s busy cells — that pairing answers “did staffing match the load.”
A queue-level SLA dip then localizes to one skill in the per-skill table without leaving the page, and to one agent group in the leaderboard.

​
4. Export to CSV

The KPI strip and tables answer questions inside the dashboard. When the numbers leave the dashboard — a weekly staffing review, a slide, or a spreadsheet — work in the window you already verified on screen:
  1. Set the window and bucket you intend to export.
  2. Copy the KPI strip (five numbers, volume-weighted) directly — paste into a slide as the headline.
  3. For the leaderboards and skill tables, the rendered columns (agent handled / AHT / CSAT / NPS; skill offered / SL / AHT / ASA / abandon) copy cleanly into a CSV spreadsheet — select the table, copy, and use paste-with-formatting. Long tables export the whole set: click Load more until no button remains, then copy.
For programmatic export, the same rollups are served by the per-queue analytics API under /api/v1/voice/queues/{id}/analytics — pull the window server-side and write it out exactly as displayed, with the same volume-weighted totals. The drilldown is one of three queue-observability surfaces; use each for its job:
  • Per-agent scorecard — when the leaderboard flags one agent (handled count far below peers, AHT out of family, CSAT/NPS dropping on normal volume), open that agent’s scorecard for the coaching view. Covered in Agent scorecard.
  • Wallboard — the drilldown’s live tiles answer “right now” for one queue; the wallboard keeps all queues’ real-time SLA up on a screen without interaction. When the live tiles show sustained pressure across several queues, move to the wallboard and its alarm rules rather than refreshing drilldowns. Covered in Wallboard and Wallboard alarm rules.

​
6. How this page relates to Voice → Queues

Pick the surface by scope:
  • /voice/queues (the list) — all queues at once, one summary row each, for “which queue hurts.” See Voice queues.
  • /voice/queues/{id}/analytics (this page) — one queue in depth, for “why does this queue hurt and where.”
The summary metrics are computed from the same volume-weighted rollup as the list view, so a queue whose list row shows SLA 74% and abandon 12% shows exactly those values on this page’s KPI strip for the same window. The queues list is the triage surface; this drilldown is the diagnosis surface.

​
7. Screenshot sequence

Capture the page for a review in this order — each shot stands alone and the set tells one story:
  1. Header + live tiles — includes the window/bucket selectors (so every later shot is dated) and the four live counters.
  2. KPI strip — the five aggregate numbers, with the AHT definition tooltip visible.
  3. Time-series chart — hover the worst bucket so the tooltip shows offered, answered, and SLA% for that interval.
  4. Heatmap — full grid; mark the cells you actually staff.
  5. Leaderboard, top rows — truncate politely (top 10), or name the agent in closed session.
  6. Hangup breakdown + skill table — pair them; caller-share plus one bad skill row is the actionable cell.

​
See also